My recent post on GLP-1s for obesity perfectly illustrates this. Early weight-loss signals appeared in diabetes trials and preclinical data, yet the broader potential was under-pursued for years (10 years, or more…). Champions like Lotte Bjerre Knudsen at Novo Nordisk had to overcome widespread scepticism that obesity was a “lifestyle issue” rather than a treatable chronic disease. Many companies had abandoned obesity programmes entirely, following commercial failures. The result? Multi-year delays in what became a multi-hundred-billion-dollar (and growing) market with profound cardiometabolic benefits.

This pattern repeats across pharma history. Opportunities hide not because they are secret, but because we are not primed to see them.

Classic Cases of Plain-Sight Signals Ignored or Delayed

Sildenafil (Viagra) stands as the textbook example. Pfizer developed it for angina and hypertension. In trials, it showed limited cardiovascular efficacy but a notable side effect: improved erections. Instead of dismissing this as noise or an undesirable off-target effect, observant clinicians and developers recognised its potential for erectile dysfunction (ED). The drug was successfully repurposed and became a blockbuster. The “plain sight” element was the incidental clinical observation that did not fit the original target product profile (TPP) - yet held massive value in an adjacent indication.

Thalidomide offers an even more dramatic arc. Withdrawn in the early 1960s for devastating teratogenic effects after use as a sedative and anti-emetic, the molecule sat in disrepute. Yet its anti-inflammatory and immunomodulatory properties (inhibiting TNF-alpha) were later harnessed for erythema nodosum leprosum and multiple myeloma. Celgene turned a toxic failure into a successful oncology and rare-disease therapy under strict controls. The data on its mechanisms existed in the literature; what changed was someone looking with a different lens.

AZT (zidovudine) was originally synthesised in the 1960s as a potential chemotherapy agent but abandoned for lack of efficacy. During the early HIV/AIDS crisis, its anti-retroviral activity became apparent, leading to rapid repurposing as the first approved HIV treatment through NIH-industry collaboration.

Daptomycin was developed by Eli Lilly as an antibiotic but shelved during a portfolio shift. Cubist later licensed it, recognised its value against resistant Gram-positive infections, and brought it to market successfully.These cases share a thread: early data or mechanisms existed, but corporate priorities, safety concerns, or indication focus obscured the potential until external pressure, serendipity, or determined advocates intervened.

More Recent and Subtle Examples

Modern examples often involve shelved compounds or internal delays rather than outright failures.

• Mirdametinib (for neurofibromatosis type 1) and tovorafenib (for paediatric low-grade glioma) were nearly scrapped after poor performance in their original oncology indications (e.g., melanoma). Patient advocacy groups and smaller biotechs spotted the fit for rarer, genetically defined populations and revived them, leading to approvals.

• At Eli Lilly, internal committees delayed advancement of CDK4/6 inhibitors for breast cancer, allowing Pfizer (Ibrance) and Novartis to reach market first with their own agents. Lilly’s Verzenio eventually arrived but later. Similar conservatism reportedly held back early triple-agonist programmes (GLP-1/GIP/glucagon) at Novo Nordisk due to glucagon-related safety concerns, even as semaglutide advanced.

• Broader industry trends show “shelved” assets gathering dust - compounds with clean safety data but abandoned due to perceived commercial risk or portfolio shifts. Non-profits and academics sometimes mine these for repurposing in rare diseases or new mechanisms.

In each case, the signal was there: mechanistic data, subgroup responses, side-effect patterns, or adjacent disease biology. What was missing was systematic scanning for opportunity profiles beyond the initial hypothesis.

Why Opportunities Stay Hidden - and How “What If” Thinking Helps

Several forces contribute to this blindness:

• Path dependence and TPP rigidity: Teams optimise for a specific indication, patient population, or endpoint. Anything outside risks “scope creep” or regulatory complications.

• Institutional inertia: Committees, risk models, and portfolio priorities favour incrementalism. Obesity was long viewed as non-druggable or behavioural. “Undruggable” targets persist until new modalities prove otherwise.

• Cognitive and cultural biases: Familiarity breeds dismissal. Early weak signals are noise until a champion reframes them. Safety fears (real or perceived) can kill programmes outright.

• Incentive misalignment: Short-term revenue protection (e.g., delaying better formulations to extend patents) or fear of cannibalisation can suppress bolder exploration.

Counterfactual “what if” thinking counters this. Ask systematically:

• What if this side effect is actually the efficacy signal?

• What if this works in a genetically defined subgroup?

• What if obesity/diabetes/neurodegeneration share pathways we are already modulating?

• What if we develop the asset in parallel for multiple profiles rather than sequentially?

Structured processes - cross-functional “opportunity mining” teams, external scouting of shelved assets, real-world evidence mining, and deliberate scenario planning - can surface these. Your blog’s emphasis on asymmetric learning fits here: winners build edges by noticing and acting on signals others filter out.

Implications for Pharma Today

With exploding data (omics, real-world evidence, AI pattern recognition), the volume of “plain sight” signals will only grow. Yet without cultural and process shifts, many will remain invisible. The GLP-1 revolution did not require entirely new science - it required believing the early hints and pursuing obesity as a serious indication. Trillions in value and patient benefit hinged on that reframing.

Leaders who institutionalise curiosity - ruthless truth-seeking, champion protection, and routine “what if” audits of pipelines and failed programmes - will capture the next wave. The opportunities are already hiding in plain sight. The question is whether we train ourselves to see them.

Sources / Further Reading

• GLP-1 obesity post: https://ideapharma.substack.com/p/glp1s-for-obesity-were-hiding-in-plain-sight

• Asymmetric Learning series:

Asymmetric Learning - Pharmaceutical Innovation

Innovations in Pharmaceutical Innovation A blog/ part-work from Mike Rea, CEO of IDEA Pharma, exploring pharmaceutical innovation in general, and more specifically the approach to using the learning process for competitive advantage

By Mike Rea

• Sildenafil repurposing history: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3068895/

• Thalidomide revival: https://www.nejm.org/doi/full/10.1056/NEJMra0905631

• AZT development: https://www.nature.com/articles/d41586-019-03763-2

• Recent repurposing examples (mirdametinib etc.): https://www.fda.gov/drugs and company press releases

• Industry shelved assets discussion: various BioCentury and Fierce Biotech reports

The remarkable success of GLP-1 receptor agonists in obesity and beyond has transformed patient lives. Yet, as I’ve covered before, the path from diabetes-focused development to dedicated obesity indications took nearly a decade or more - despite early signals of weight loss. This wasn’t inevitable. It reflects common pharma challenges: path dependence on primary indications, regulatory focus, and internal skepticism.

My core thesis here: Better, earlier “what iffing” - systematic counterfactual thinking about adjacent opportunities - could enable the industry to lose less time spotting and pursuing high-impact applications. This isn’t about perfect prediction but about building processes to challenge assumptions and explore multi-indication potential proactively.

The GLP-1 Timeline and the Delay

Key quotes from Lotte Bjerre Knudsen (the persistent champion at Novo Nordisk):

“A lot of people didn’t believe it… They didn’t believe this class of drugs would be good for diabetes, but it was good. They didn’t believe it would be good for weight loss, but it was good.”

On early resistance and peptides reaching the brain: “…people just didn’t think that the peptides could impact on the brain.”

Knudsen’s advocacy eventually prevailed, but broader industry buy-in lagged. Weight loss signals existed in diabetes trials and animal data, yet dedicated higher-dose obesity programs came later.

Counterfactual for GLP-1: Time That Could Have Been Gained

If obesity potential had been aggressively “what if’d” in the late 1990s/early 2000s - running parallel higher-dose trials alongside diabetes work - approvals like Saxenda could have arrived 5–10 years earlier in optimistic scenarios (or 3–7 years more realistically). Even on conservative assumptions, those “lost” years add up. Imagine only 5–10 million high-risk people globally gaining access 3–7 years earlier – a fraction of today’s eligible population.

Based on current cost‑effectiveness work for semaglutide‑like GLP‑1s, that kind of early uptake would likely translate into hundreds of thousands of major cardiovascular events avoided, on the order of 1–5 million QALYs gained, and tens to low hundreds of billions of dollars in aggregate health and economic value. These aren’t precise forecasts, but they show that “a few years’ delay” for a platform like GLP‑1s probably corresponds to multi‑million‑QALY and multi‑billion‑dollar consequences.

On the purely financial side, the GLP‑1 wave is already measured in tens of billions of dollars a year for individual products, with class‑wide forecasts in the $150–250 billion‑per‑year range as obesity indications scale. Even a modest 3–7 year acceleration of that trajectory implies hundreds of billions to over a trillion dollars of revenue and health value brought forward. Again, these are directional, not precise, but they underline the stakes: “being late” with a platform like GLP‑1s is not just a scientific missed opportunity – it is a trillion‑dollar, multi‑million‑QALY delay.

Extending “What If” to Other Areas

This pattern repeats:

Oncology: BCR-ABL for imatinib (Gleevec) or CTLA-4/PD-1 pathways for checkpoint inhibitors saw mechanistic insights decades before broad approvals and expansions. Earlier platform thinking (e.g., histology-agnostic trials or aggressive cross-tumor exploration post-1990s) could have delivered years of additional survival benefits.

Immunology/Autoimmune: TNF inhibitors like Humira started narrow but expanded across indications. Faster cross-indication “what if” programs might have relieved more patients sooner.

Repurposing examples (aspirin in cancer prevention, metformin overlaps) show even low-cost opportunities often wait for serendipity or external champions rather than proactive analysis.

Common delays (5–15+ years) stem from indication silos, risk aversion after past failures, and focus on de-risked primary endpoints. Counterfactual exercises could surface these earlier.

Institutionalizing Better “What Iffing” for Foresight

The lesson isn’t hindsight blame, but forward-looking process improvement. Pharma can reduce lost time by embedding systematic scenario planning and counterfactual analysis:

• Early cross-functional reviews: Mandate teams to explicitly model “what if this mechanism works in adjacent diseases?” using existing trial data, animal models, and real-world signals.

• Scenario planning frameworks: Explore optimistic/base/pessimistic futures for multi-indication potential, including regulatory, reimbursement, and competitive angles. This is already used in some long-range planning but could be applied earlier in discovery.

• AI/digital twins for simulations: Emerging tools enable “what if” modeling of biological networks, trial outcomes, or portfolio impacts far more efficiently.

• Incentives and culture: Reward champions like Knudsen and de-stigmatize “risky” adjacent pursuits. Portfolio reviews could assess assets’ “adjacency potential.”

• Regulatory and collaborative nudges: Master protocols, basket trials, or public-private foresight initiatives could accelerate testing.

This approach helps “see ahead more clearly” without derailing core programs - the diabetes foundation still de-risked GLP-1s. But it minimizes blind spots.

Quantifiable stakes: Development delays cost revenue, patent life, and - most importantly - patient outcomes. Even modest accelerations compound into massive health and economic benefits.

Lessons and Open Questions

GLP-1 ultimately succeeded spectacularly. But reflecting on these delays highlights how better institutionalized “what iffing” could optimize R&D systems. For today’s pipelines: Are we fully exploring GLP-1s in neurodegeneration or other metabolic overlaps? What about emerging modalities in oncology or immunology?

This post is a confession of sorts. Ideally, it’s the best one to forward to someone… (That’s why it’s here.)

After 300-plus pieces, here’s the simple question it tries to answer: what exactly is the philosophy that ties all this together? If the archive is the footnotes, this is the argument. The question that started my philosophy a couple of decades ago: if you gave the same molecule to two companies, would it end up in the same place? I’ve never met anyone who thought it would, but we operate in an industry that seems to believe the answer is ‘yes’.

1. The uncomfortable starting point

Pharma innovation matters more than ever. Yet we routinely sabotage it with the way we choose to run it.

The symptoms are everywhere:

• R&D “productivity crises” that are really decision-making, culture, and incentive problems.

• Trials that fail not at readout, but in the design assumptions months or years earlier.

• Launches that feel like acts of faith rather than the logical endpoint of a coherent journey.

• A stubborn preference for certainty theatre over honest uncertainty.

The premise of this blog isn’t that pharma is broken or that people are stupid. It’s that we’ve normalised ways of working that leave huge amounts of value - and learning - on the table. This blog exists to poke at that normal.

2. Asymmetric learning: the quiet edge

The name isn’t just branding; it’s the thesis.

Back in 2020, I wrote about asymmetric learning as the hidden advantage behind some of the biggest biopharma successes. The market has kindly kept supplying fresh case studies ever since.

By it I mean:

• Learning more than your competitors from the same (or less) activity.

• Designing programmes where even “failures” teach you something useful - about biology, patients, endpoints, or your own judgement.

• Structuring decisions so the downside is bounded but the informational upside is essentially unlimited.

In practice this shows up as real-time trials, smart externalisation, partnerships that shift risk while accelerating insight, and launches treated as live experiments rather than one-off events.

Companies that bake this into their culture, portfolio, and commercial approach quietly outperform those still trapped in the prediction-and-de-risk paradigm. The data keeps voting the same way.

3. Premise → Proof → Promise: the development backbone

One of the more explicit posts in the archive lays out the three-word sequence that underpins everything here.

• Premise: What do we actually believe, and why might it matter in the real world? This is where the sharp hypothesis about target, mechanism, patient, and unmet need lives. Not a slogan - a testable story.

• Proof: What evidence would genuinely change our mind, positive or negative? This is about asking real questions, not just generating registrable p-values.

• Promise: If the premise holds, what is the authentic promise to patients, payers, and the business? This is where product profile, access, and launch strategy matter.

Most development gets this sequence backwards: Promise first (peak sales decks and lifecycle fantasies), then a Proof programme reverse-engineered to support it, then frantic fudging of the Premise when reality bites.

The philosophy here is simple: earn your Promise by being ruthless about Premise and Proof. When the sequence is clear, TPPs, trial design, and launch execution stop feeling like art and start looking like disciplined craft.

4. Innovation theatre vs actually wanting to know

The recurring villain is innovation theatre - those glossy labs, patient-centricity videos, AI press releases, and “fail fast” posters that somehow never survive contact with reality.

Fifteen years on, the symptoms are depressingly familiar.

Behind it all sits a harder question: do you actually want to know?

Asymmetric learning only works when the real incentive is discovering the truth, not confirming the PowerPoint. Many studies are effectively dead before they start - not because the science was impossible, but because the design quietly avoided uncomfortable comparisons, the endpoints were chosen to be passable rather than decisive, and the team needed the study to succeed more than it needed to learn.

The most radical thing a pharma company can do is genuinely prefer the right answer over the favourable one.

5. Incentives, Freshness, and the quiet death of innovation

If there’s one explanatory variable running through the archive, it’s incentives. Whenever smart, well-paid people keep doing obviously suboptimal things, the question is always the same: what are they being rewarded (or punished) for?Recurring patterns:

• R&D teams rewarded for starts, not intelligent stops.

• Leaders promoted for ownership of big budgets rather than decision quality.

• Organisations that quietly default to the Compromise Product Profile (CPP) over the differentiated Target Product Profile (TPP), or ideally the Target Opportunity Profile, because compromise keeps more stakeholders happy - even as it kills the product.

Add in the issue of Freshness: portfolios clogged with old, comfortable bets that crowd out sharper new ones. Option sets that collapse too early to “one big bet” - even though one remains the second-worst number of options in pharma.

Opportunity cost is real. Freshness - the rate at which you generate and act on new options - often matters more than any single asset. CPPs are where innovation goes to die politely.

Change the incentives and what you tolerate, and different outcomes become possible.

6. Strategy and launch: discipline, not destiny

Strategy isn’t a mood or a strapline; it’s the set of things you’re willing not to do.

In launch, the same logic applies. First-mover advantage is overrated. The winners are usually those who design development with launch in mind, build an information edge on patients/prescribers/payers before Day 1, and keep learning - and acting - faster than competitors after approval.

Launch isn’t an event. It’s Premise → Proof → Promise continuing in the real world, now with a revenue line attached.

7. Thinking tools, not complicated diagrams

Not every post is about molecules and markets. Some are about how to think better.

You’ll find simple models (Freshness, Premise → Proof → Promise, binoculars vs eyes, the two worst numbers of options), analogies from spies, poker, and aviation, and a consistent push toward Bayesian habits: good priors, rapid updating, refusing to be surprised twice by the same thing.

The meta-philosophy is straightforward: serious work deserves simple, sharp tools you can actually use on Monday morning.

8. Tone: high stakes, low solemnity

The subject is serious - lives, science, billions, careers. The tone deliberately isn’t.

Plain language, mild irreverence when deserved, and a willingness to call expensive nonsense what it is. Underneath the occasional sharpness is stubborn optimism: pharma can be much better than it is, and some companies and teams already prove it.

9. What this blog is actually for

After all these posts, here’s the closest thing I have to a mission:

• To show that asymmetric learning is a practical, repeatable edge - in how you choose, run, and launch programmes.

• To make the hidden costs visible: of CPPs, anaemic trial designs, performative cultures.

• To give serious people in and around pharma ideas and examples that help them make slightly better decisions under uncertainty.

This isn’t neutral corporate thought leadership. It’s a long, evolving argument that you can design for Premise → Proof → Promise, reward wanting to know, and treat every trial, partnership, and launch as an asymmetric learning opportunity.The market keeps rewarding those who do.

10. How to use the archive

If you’re new:

• Start with the philosophical anchors: Premise → Proof → Promise, The Real Reason Pharma Studies Fail Before They Start, Innovation Theatre in Pharma (15 Years Later), Opportunity Cost, and The Market Keeps Voting for Asymmetric Learning.

• Then pick a case study (GLP-1s, CAR-T, real-time trials, biotech M&A) and read it through this lens.

• Finally, look at your own world and ask the practical questions: Where are we choosing CPP over TPP? Where are we pretending to want the answer when we really want the outcome? Where could we design for more learning at lower cost?

If even one decision lands differently because something here shifted your prior, the whole exercise will have been worth it.

Five years later, the data isn’t just supportive - it’s overwhelming.

Pre‑commercial emerging biopharma companies (the “lil’ biotech” in the chart caption) have taken over clinical trial starts across every phase. They’re not just participating. They’re leading.

Here’s the chart, from A16Z:

Key takeaways from the 2016–2025 data (Stifel / Citeline / IQVIA):

• Phase 1 - pre‑commercial EBP (‘emerging biopharma’, gold line) grew heavily from ~500 starts to peaks above 1,600, while Large Pharma (dark blue) declined.

• Phase 2 - similar story: “lil’ biotech” surging, big players flat‑to‑down. (I’m not a fan of the phrase “lil’ biotech’ so will stop using it now…)

• Phase 3 - even in late stage, pre‑commercial EBP now outpaces Large Pharma.

(The chart shows, but A16Z doesn’t speculate on, the dip post 2023… Small pharma seems to have picked up some, but not all, of it…)

Across all phases, pre‑commercial EBP companies accounted for the majority of trial starts in recent years, with their share rising from around the low‑40s percent in the mid‑2010s to roughly two‑thirds by the early 2020s in several datasets.

This probably isn’t a blip. It looks more like a structural shift.

What Asymmetric Learning Looks Like in Practice

Asymmetric learning favours exploration over exploitation in the early phases. You spend time and capital to generate optionality - multiple shots on goal, rapid iteration on signals, and a willingness to kill ideas fast and pivot.

Small biotechs are structurally built for this:

• One big success changes everything (asymmetric upside).

• Limited infrastructure means lower fixed costs and less pressure to protect an existing core.

• Funding models reward bold exploration - until they don’t.

• Decision rights sit closer to the science.

Large Pharma, by contrast, has optimised for symmetric learning: predictable returns, portfolio balancing, quarterly discipline, risk committees, and huge legacy operations that punish failure more visibly. The result - fewer early bets, more in‑licensing later, and a flatter line on the charts above.

This isn’t about “big companies are dumb”. It’s about incentives and architecture. When your organisation is built for reliability at scale rather than exploration, the rational move is to let someone else carry the exploration load - then acquire or partner once the asymmetry has been resolved into clearer probabilities.

The market seems to have noticed. Licensing deals, acquisitions, and platform valuations all reflect exactly this dynamic.

Why This Matters, Beyond the Numbers

The chart isn’t just about trial volume. It’s evidence that the locus of innovation has shifted towards organisations that can tolerate - and even seek - asymmetry.

It raises hard questions for anyone running (or investing in) large R&D organisations:

• How do you rebuild exploration capacity without blowing up efficiency metrics?

• Can you create “biotech” pockets inside a large company that are truly protected from symmetric pressures?

• Or is the winning model now permanent specialisation - EBP explores, Big Pharma scales and commercialises?

Some large players, such as Lilly, are experimenting - venture arms, separate discovery units, aggressive external innovation mandates. But the aggregate data suggests it’s not yet moving the needle on raw trial starts.

The Deeper Lesson for All of Us

Asymmetric learning was never only a biopharma concept. It applies to strategy in any uncertain domain - tech, climate, personal careers. The pattern is the same: early variance creates later advantage. The organisations (and people) that win are often the ones willing to look inefficient for a while.

The biotech surge shows the power of that approach at industry scale. It also shows the limits of pure symmetry.

Big Pharma isn’t going away - its strengths in late‑stage execution, manufacturing, and global access remain critical. Commercial at obesity market scale still needs big organisations and budgets. But the data make one thing clear - if you want to lead in discovery and early development, you need to get comfortable carrying more of the exploration load yourself.

Or get very good at partnering with those who do.

Sources: Stifel Biopharma Update, Citeline Trialtrove, IQVIA Institute reports.

In pharma, the allure of being “first to market” is powerful. Conventional wisdom – and some data – suggest a modest edge: perhaps 6–10% higher market share or 27% higher long‑run sales in certain contexts. But as many developers know, that advantage often erodes quickly without deeper differentiation, and evaporates entirely if the second to market is also ‘best to market’.

The real differentiator isn’t launch timing on its own. It’s asymmetric learning: the ability to learn faster and more deeply from biology, clinical data, patient realities, regulatory signals and market feedback than anyone else in the field.

That means:

• Empirical exploration rather than wishful extrapolation.

• Rapid iteration in trials and in market, not just in decks.

• Avoiding premature lock‑in on target product profiles (TPPs).

• Holding plans lightly enough that you can pivot when the data tell you to.

This stands in stark contrast to the rigid “prediction paradigm”: big bets, made early, on assumptions that are then defended long after the evidence has moved on.

Recent standout launches in novel or high‑unmet‑need spaces illustrate this perfectly. Companies like Merck, Madrigal, BMS (building on Karuna) and Vertex turned first‑mover (or near‑first) positions into rapid, scalable success not simply through speed to market, but through superior learning velocity.

Case studies: learning in action

Merck’s Winrevair (sotatercept) for Pulmonary Arterial Hypertension (PAH, approved 2024)

Winrevair is a first‑in‑class activin signalling inhibitor in a disease with very limited disease‑modifying options. Merck (via Acceleron) brought deep mechanistic insight together with disciplined execution.

Performance: A strong ramp, with hundreds of millions in revenue in its early quarters and impressive scaling through 2025 via label expansions and international growth.

Asymmetric learning edge:

• Rapid trial iteration, including early stops for efficacy.

• Biology‑driven patient selection to really stress‑test the mechanism.

• Supply readiness tuned to realistic uptake scenarios, not wishful thinking.

Existing therapies were largely supportive. Winrevair offered transformative potential – and the team learnt quickly enough to prove it and then build on it.

Madrigal’s Rezdiffra (resmetirom) for MASH (approved March 2024)

Rezdiffra is the first and, so far, only approved therapy for metabolic dysfunction‑associated steatohepatitis with fibrosis – in a field littered with prior failures.

Performance: Explosive growth: hundreds of millions of dollars in the first full year, with strong patient and prescriber uptake in what was previously a “graveyard” category.

Asymmetric learning edge:

• Persistence with hard histology endpoints where others had retreated.

• A clear focus on real‑world relevance, not just tidy trial read‑outs.

• Flawless commercial execution: market education, payer engagement and ongoing data generation to define and own a new category.

After decades of setbacks in NASH/MASH, Madrigal’s edge was not a magic molecule in isolation; it was their ability to keep learning where others had stopped.

BMS’s Cobenfy (xanomeline/trospium, approved September 2024) – building on Karuna’s innovation

Cobenfy brings the first new pharmacological mechanism for schizophrenia in over 50 years: a muscarinic agonist that sidesteps the familiar D2 blockade issues. Its success builds directly on the pioneering work of Karuna Therapeutics – friends whose journey is a textbook in asymmetric learning.

Karuna in‑licensed an abandoned molecule (xanomeline from Eli Lilly), paired it with trospium to tackle tolerability, and persisted with empirical exploration to unlock a novel cholinergic pathway after decades of stagnation in the field.

Performance: A steady build with tens of millions in early quarters, backed by strong coverage and education efforts, and with further potential in adjunctive use and label expansions (for example, in Alzheimer’s psychosis).

Asymmetric learning edge:

• A bold commitment to a non‑dopamine pathway when the field had largely converged elsewhere.

• Rapid iteration on clinical data, including dose, tolerability and positioning.

• Seamless integration of Karuna’s learnings and culture into BMS post‑acquisition.

It is a vivid reminder of how focused, persistent learning on “unfashionable” biology can revive whole therapeutic areas.

A strong parallel in schizophrenia is Intra‑Cellular Therapies (led by the great Sharon Mates), whose Caplyta (lumateperone) has grown impressively through differentiated pharmacology and thoughtful execution. Their trajectory – including expansion into bipolar depression and subsequent $15bn acquisition by Johnson & Johnson – reinforces the point: asymmetric learning and patient‑focused innovation drive real value in neuropsychiatry.

Vertex’s Alyftrek (vanzacaftor/tezacaftor/deutivacaftor, approved late 2024/early 2025)

Alyftrek is a next‑generation, once‑daily CFTR modulator, building on the Trikafta platform.

Performance: Strong contributions within Vertex’s already dominant cystic fibrosis franchise, with label expansions moving towards broader genetic eligibility and more convenient regimens.

Asymmetric learning edge:

• Decades of CFTR biology mastery and systematic variant testing.

• Close, iterative engagement with patients and clinicians on convenience, adherence and access.

• A pipeline designed as a learning system, not a one‑off success.

Vertex’s story in cystic fibrosis is, in many ways, an extended case study in compounded learning.

Patterns of asymmetric learning success

Across these launches, recognisable traits emerge that align with the asymmetric learning framework:

• Novel mechanisms in uncontested or high‑unmet spaces
  Deep biological insight over incremental “me too” strategies. Winrevair, Rezdiffra and Cobenfy did not simply offer marginal improvements; they reframed their categories.

• Rapid leverage of data
  Early trial stops for efficacy, thoughtful label expansions and proactive generation of real‑world evidence. Data were treated as fuel for new decisions, not just as regulatory artefacts.

• Execution as a learning process
  Supply, education, payer engagement and even acquisitions (BMS/Karuna, J&J/Intra‑Cellular) were used as accelerators for more and better feedback, not just as distribution channels.

• Flexibility over rigid prediction
  Teams systematically chose empirical pivots over clinging to early TPPs. Plans were hypotheses to be updated, not contracts to be defended.

By contrast, in crowded spaces like GLP‑1s, we see a different pattern: Novo’s early semaglutide repurposing and learning edge, followed by Lilly’s very strong dual‑agonist “fast‑follower” move. In such markets, learning velocity often trumps pure timing. First‑mover helps when the space is truly novel – but it only compounds if you also have an asymmetric learning edge.

Non‑pharma parallels

This is not a pharma‑specific phenomenon.

• Amazon was not the first e‑tailer, but it learned and iterated relentlessly on logistics, data, and, eventually, Prime.

• Google refined search beyond early engines through superior algorithms and continuous, data‑driven user feedback loops.

• Tesla accelerated EV adoption not just with timing, but with an obsession with battery chemistry, software, manufacturing and over‑the‑air learning.

MySpace faded. Facebook won on adaptation and product iteration. Netscape gave way to Google Chrome. The pattern repeats:

Learning velocity > launch velocity.
The myth of the first‑mover advantage crumbles without asymmetric learning behind it.

Takeaways for pipelines and portfolios

So what do you actually do with this if you are running a portfolio or pipeline?

1. Design for learning, not just for approval

• Prioritise exploration and stage‑appropriate decisions over rigid, long‑range prediction.

• Build in optionality: adaptive designs, exploratory cohorts, early real‑world pilots.

• Treat the TPP as a living hypothesis, not a fixed contract with your former self.

2. Build a cultural and organisational edge

• Blend biotech agility with pharma scale: short feedback loops between discovery, clinical, commercial and HEOR, rather than linear hand‑offs.

• Use AI/ML to compress learning cycles – in target validation, trial design, segmentation and real‑world data – but remember it only amplifies the experiments you are actually willing to run.

• Reward teams for updating their views with data, not for simply sticking to the original plan.

3. Ask sharper portfolio questions

Instead of counting only molecules and milestones, ask:

• Where are we genuinely learning asymmetrically – in biology, in patients, in the market?

• Where are we betting on our capability to learn, rather than just on being first?

• Which programmes are trapped in rigid prediction mode, with TPPs and assumptions that have quietly outlived their evidence?

4. Audit for rigid prediction risk

• Identify teams and programmes that treat early assumptions as sacred.

• Look for launch plans that lack built‑in mechanisms to learn from patients, prescribers and payers in the first 6–12 months.

• Actively favour teams that pivot on emerging data, even when that is politically uncomfortable.

The standout stories of 2024–2025 – and earlier journeys like Karuna’s and Intra‑Cellular’s – show that asymmetric learning is a repeatable competitive advantage. In an era of complex biology and fast‑moving markets, it is how first‑mover positions become enduring ones, rather than brief footnotes in someone else’s success story.

The data suggests the latter. Acquiring or licensing from biotechs often delivers higher clinical success rates and, crucially at today’s lower biotech valuations, better risk-adjusted economics than going it alone from discovery. But it’s not without nuance - especially when examining how individual big players perform against the average and what lessons emerge from contrasting strategies.

Internal vs External: The Success Rate Gap

Clinical development remains a brutal filter. Recent benchmarks put the composite Phase I-to-approval success rate at around 7-10%, though large pharma as a group outperforms the broader industry (often dragged down by smaller biotechs with riskier assets).

External assets - whether licensed or acquired from biotechs - frequently beat internal benchmarks. In-licensed or partnered projects have shown meaningfully higher phase transition probabilities, with some analyses indicating 12-17% advantages in early phases and 11-18% in Phase III. Partnered assets have nearly doubled non-partnered success rates from Phase I to launch in certain multi-year windows (e.g., ~19% vs ~9%).

Selection effects explain much of this: biotechs typically only advance assets with compelling early data to a deal stage, de-risking them for Big Pharma. External origin now dominates approvals, with 60-70%+ of recent NMEs stemming from outside large pharma labs, biotechs being a primary source.

Big Pharma vs Industry Average - and the Standouts

A comprehensive 2006-2022 benchmarking study of 18 leading companies calculated an average Phase I LoA of 14.3% - well above the industry norm. Yet the asymmetry missed in the ‘average’ is telling: Amgen led at ~22.8%, followed by Novo Nordisk (~20.7%) and Eisai (~18.4%). At the lower end sat AbbVie (~8.1%), Astellas (~8.6%), and GSK (~9.1%).

Amgen and GSK illustrate divergent approaches with instructive outcomes.

Amgen: Targeted Acquisition Engine

Amgen has combined a focused internal pipeline with high-impact biotech acquisitions, yielding top-tier success rates. Deals such as Horizon Therapeutics (2023, ~$27.8 billion) brought first-in-class rare disease assets like TEPEZZA (thyroid eye disease), KRYSTEXXA (gout), and UPLIZNA - strengthening its inflammation and nephrology portfolio and delivering immediate revenue and pipeline depth. Earlier moves included ChemoCentryx (TAVNEOS for vasculitis, 2022) and others like Five Prime and Teneobio, emphasising platform technologies and de-risked clinical assets.

This strategy aligns with Amgen’s high LoA: it buys validated shots on goal in areas like oncology (e.g., IMDELLTRA for small cell lung cancer) and rare diseases, where biologics often enjoy ~2x higher approval odds. The result? Stronger pipeline productivity and commercial momentum without shouldering the full early-stage attrition burden.

GSK: Internal Strength with Selective External Augmentation

GSK has historically leaned more on internal R&D, with a lower LoA in the benchmarking data. It maintains a broad clinical trial footprint and invests heavily in organic discovery, particularly in vaccines, respiratory, and infectious diseases. Recent performance shows improvement - multiple Phase III readouts and approvals in 2024-2025 - but it has faced pipeline gaps and patent pressures.

In response, GSK has ramped up targeted external deals: multi-billion alliances (e.g., with Hengrui Pharma for up to 12 programmes in respiratory/immunology/oncology), acquisitions like IDRx, and partnerships in AI, RNA, and fibrotic diseases. This hybrid shift aims to complement internal efforts without a “shopping spree” for late-stage assets, which GSK leadership has viewed sceptically for quality reasons. The question about whether external early stage risk looks different than internal early stage risk remains open (although I suspect we all think the same way about it…).

Lessons from the Contrast

Amgen’s edge comes from disciplined, thematic acquisitions that integrate seamlessly and accelerate in high-value areas (rare diseases, oncology) - importantly areas that they already know well. It demonstrates that external moves, when selective and capability-aligned, boost both LoA and growth without diluting focus. Horizon, despite its size, has enhanced Amgen’s rare disease leadership with early-lifecycle assets offering strong commercial runway.

GSK’s path highlights the strengths (and risks) of internal emphasis: deeper scientific ownership and potentially higher long-term commercial value per approved drug (internally originated NMEs have shown ~40% higher average value in some analyses). Yet lower success rates and pipeline volatility underscore the need for external augmentation amid rising complexity. GSK’s recent deals suggest a maturing hybrid model.

Broader takeaway: Pure internal innovation is rare for sustained output in the industry today. Winners balance deep capabilities with aggressive scouting. External reliance isn’t dilution - it’s asymmetric risk management. Invention is necessary, but it is not sufficient.

The Cost Equation: Cheaper (De-Risked) Shots on Goal?

Full internal discovery-to-launch costs $1-3+ billion per approved drug when failures are fully capitalised - a figure that aligns closely with the $2 billion+ average I highlighted in my recent Opportunity Cost update, where R&D costs per approved asset run from as low as ~$1 billion in the best performers to north of $10 billion in others. Simple industry-wide calculations of total R&D spend divided by approvals push the effective average even higher, towards $6 billion in some analyses of the top 30 companies. Preclinical and early clinical attrition - where most programmes die, through failure or decisions to discontinue - devours budgets before a single asset ever reaches proof-of-concept.

Acquiring a biotech or licensing a mid/ late-stage asset fundamentally changes the economics. You pay a premium for de-risked data (often post-Phase I/II proof-of-concept), but sidestep much of the early-stage sunk cost and gain speed to market. At post-2022 biotech valuations - perhaps more ‘reasonable’ after the 2021 peak - quality assets have become accessible. Big Pharma has deployed hundreds of billions in M&A precisely for this reason: external moves compress the failure stack and improve risk-adjusted returns. I wonder, also, whether the unseen costs - new product planning, portfolio trimming, etc., go uncosted in these analyses, now that they’re taken by the biotech, either by necessity, or the focus of VC-backed investment.

Deal structures (upfront payments plus milestones and royalties) further spread risk. Early-stage clinical deals have often outperformed late-stage deals in historical returns. Biologics and orphan indications - frequent strengths of biotech origin - also enjoy higher approval rates and commercial premiums.

Caveats remain important. Integration risks, cultural mismatches between entrepreneurial biotech and rigid pharma, and post-deal revenue shortfalls are real (many assets miss pre-deal peak sales forecasts, just as most internal pharma forecasts are wrong - its no surprise that M&A forecasts are equally wrong, and in both directions). Internally originated drugs can still deliver stronger long-term commercial upside in certain datasets. When acquisition premiums are fully loaded, external strategies do not automatically guarantee superior productivity. Overpaying on hype or chasing crowded mechanisms remains a common pitfall, as we’re seeing in some spaces today.

Strategic Asymmetry in a Complex Era

The smartest Big Pharma strategies treat external innovation as high-conviction options on validated science - real options that preserve pluripotency rather than locking everything into one narrow path. In a world of complex biology, regulatory hurdles, and finite capital, the hybrid model sustains output where pure internal efforts falter. One option is still the second-worst number of options in pharma.

For biotechs, it underscores focusing on nimble, high-quality science - with licensing or acquisition as a value-maximising path. For patients, more validated shots reach the clinic faster.

Amgen’s success with targeted buys versus GSK’s internal-plus-selective approach shows no single blueprint exists. Execution, therapeutic focus, and timing matter. The data does not say “abandon internal R&D.” It says build profound capabilities and scout relentlessly. In pharma, winning means knowing when to invent - and when to acquire the invention.

Suggested Reading / Sources

• Schuhmacher et al., “Benchmarking R&D success rates of leading pharmaceutical companies” (Drug Discovery Today, 2025) - https://www.sciencedirect.com/science/article/pii/S1359644625000042

• Deloitte, “Measuring the Return from Pharmaceutical Innovation 2025” - https://www.deloitte.com/content/dam/assets-zone3/us/en/docs/industries/life-sciences-health-care/2025/measuring-the-return-from-pharmaceutical-innovation.pdf

• IDEA Pharma analyses on R&D productivity and opportunity cost - https://ideapharma.substack.com/p/opportunity-cost-the-cost-of-opportunity-f73 and https://ideapharma.substack.com/p/invention-necessary-but-not-sufficient

• IQVIA, Nature Reviews Drug Discovery, and Syneos Health reports on licensing/M&A performance (2023-2025 editions)

• Company announcements on Amgen (Horizon) and GSK deals

I used this exact image about 15 years ago. Not sure pharma has got better in general. (I’ve failed 😅)

You keep using that word. Innovation.

I do not think it means what pharma think it means.

Fifteen years on, and the line still lands like a punch in the gut every time I see another press release crowning some incremental tweak or shiny pilot as “transformative innovation.”

The Gap Between Word and Reality

In pharma, we love the word innovation. We hate what it actually demands.

Invention is coming up with something new - a molecule, a target, a platform, a clever bit of tech. Innovation is turning that invention into something that creates real, sustained value for patients and the business. As I’ve said many times: invention is necessary but not sufficient.

Most of what gets labelled “innovation” sits firmly in the invention bucket - or worse, in the pure marketing bucket. A new formulation. A digital app that patients download twice. An AI pilot that generates pretty graphs but never touches a clinical trial. A gleaming innovation lab with bean bags and whiteboards that hosts impressive demos but ships nothing to market.

This is what people now call innovation theatre. (Well, Steve Blank was the person who (I think) first used it, so it’s good that ‘people’ have caught up…)

What Innovation Theatre Looks Like

Innovation theatre is the corporate equivalent of a Hollywood set: impressive from the front, completely hollow when you walk around the back.

You’ve probably seen it (or run it):

• Hackathons and idea challenges that generate thousands of Post-its and zero implemented solutions.

• Innovation labs and accelerators that partner with flashy startups, announce grand collaborations, and quietly sunset the projects two years later.

• Chief Innovation Officers who’ve become professional AI PowerPointers.

• AI/digital health initiatives that move from hype to inspection-grade reality far slower than promised.

These activities create activity. They create great photos for the annual report. What they rarely create is new revenue from new drugs or meaningfully better patient outcomes at scale.

What the 2026 Pharmaceutical Innovation + Invention Index Reveals

This year marks the 15th anniversary of the Pharmaceutical Innovation + Invention Index - the longest-running benchmark of its kind. I invented/ created the index after asking myself (and anyone I could bore…) the same provocative question: if you gave the same molecule to two different companies in early phase, who would be more successful and why? No-one then, and I assume no-one now, believed the answer was ‘yes’ but pharma still, 15 years later, acts as though it definitely is.

The 2026 results are clear.

Innovation and invention converge at the top. Eli Lilly and Company is #1 on both the Innovation Index (commercial return from invention) and the Invention Index (pipeline novelty) for the first time. Its incretin platform - anchored in Mounjaro/Zepbound and extended through retatrutide, orforglipron, and more - shows exactly what real innovation looks like: platform science paired with disciplined late-stage execution and commercial delivery.

Sanofi climbed +19 places to #3 on the Innovation Index - the biggest jump in the Top 10 - by becoming a launch- and readout-driven organization. GSK rose +10 places by translating a strategic shift into oncology, respiratory, and vaccine wins. These companies aren’t just inventing; they’re refreshing their portfolios with meaningful revenue from recent launches (the Freshness Index is revealing, alongside the Fragility Index - which shows who’s exposed to gaps).

By contrast, Bristol Myers Squibb dropped 14 places on the Innovation Index despite continued research output. The reason? A concentration of Phase 3 setbacks and loss-of-exclusivity pressure. Scientific activity without sustained late-stage delivery gets penalised. That’s the index calling out innovation theatre in real time.

Why Innovation Theatre Persists

Because it’s easier.

Real innovation is messy, risky, and slow. It requires killing sacred cows, reallocating real budget, accepting public failure, and having hard conversations with regulators, payers, and internal stakeholders. The fact that revenue (or the inverse, not losing invested dollars that don’t repay) shows up 10-15 years after those decisions - a very delayed response - makes it hard to show that you just made the right decision.

The effects show up in charts like this (downloadable from IDEA’s website). Imagine still believing or using ‘average’ costs per launch in your eNPV…

Innovation theatre lets you signal clever stuff, scientific stuff, to investors, recruits, and the board without making those tough choices.

What Real Innovation Actually Requires

Real innovation looks less glamorous and more like hard work:

• Building repeatable systems around inventions (the “garage” that actually commercializes the playground ideas).

• Ruthless prioritization - saying no to 99% of things so the 1% can succeed.

• Measuring what matters: not activity metrics, but Freshness (percentage of revenue from recent launches) and real health-economic impact.

• Accepting asymmetry - some bets will be huge winners, most will fail, and that’s the only way the maths works. Placing lots of bets, and learning from small bets quickly, helps here.

The companies that top the indices consistently do more of the latter and less of the former. They treat innovation as a discipline, not an event or a department.

Here’s the Freshness/ Fragility index from IDEA’s report - it’s fair that it focused on US approvals, as many companies are now questioning whether global (mostly EU) launches make sense; no-one questions that they’ll launch in the US first.

So… Have I Failed?

Fifteen years of banging this drum and the Princess Bride/ William Goldman meme is still painfully relevant. Maybe I have, a little.

But I remain optimistic. The 2026 Index shows that the gap can be closed. The tools are better. The pressure from investors and patients is higher. And a handful of companies are quietly proving it can be done differently.

The question for all of us in 2026 isn’t whether we can invent more stuff. It’s whether we’re willing to stop performing and start innovating - turning those inventions into medicines that actually reach patients and create lasting value.

Because if we keep using that word the way we have been… well, you know the rest.

Small, subconscious signals decide who survives under pressure. The best agents - and the best drug developers - train to spot and eliminate their own tells.

So, you’ll understand why I was so drawn to the Tarantino nuances here. In the middle of a crowded French tavern in Inglourious Basterds, a British officer orders “drei Whiskys” perfectly. Then he signals “three” with his index, middle, and ring fingers. The German major notices instantly. Game over. The movie is a rich seam of similar invisible clues that I only learned later.

That single cultural tell - not counting from the thumb, as Germans do - unravels the entire cover. It’s cinematic, but it’s also brutally accurate.

During WWII, the British Special Operations Executive (SOE) sent hundreds of agents - many of them women - into occupied Europe on missions that seem almost impossible in hindsight. These were not elite soldiers but often ordinary people with the right languages, nerves, and courage: a French-speaking Jewish refugee, a wireless operator from the Home Counties, a multilingual secretary. They parachuted or landed by Lysander at night, carrying false identities, suicide pills, and the weight of knowing that capture usually meant torture and execution. Some lasted only days. Others survived for years, building resistance networks, sabotaging infrastructure, and feeding critical intelligence back to London - all while maintaining an utterly convincing local persona under constant, lethal scrutiny. Their journeys were extraordinary acts of personal transformation under the most extreme pressure imaginable.

They weren’t faking fluency from scratch. They were embodying a new identity under mortal risk. And the parallels to pharmaceutical and biotech innovation are sharper than they first appear.

The Premise of a Good Cover

Successful SOEs started with a strong Premise: genuine bilingual roots, family ties, or time spent abroad. Fabricated covers collapsed the moment real scrutiny arrived.

The same discipline applies in drug development. A weak scientific premise - thin biology, me-too mechanism, or hope-based target product profile - leads to late, expensive failure when real-world proof hits. Winners build on genuine asymmetric advantage from the outset.

Proof Under Pressure: Training the Tells Away

SOE training was obsessive. Agents drilled regional dialects, local slang, mannerisms, and cover stories until they became muscle memory. They role-played Gestapo interrogations and learned to suppress British habits while remaining calm enough to improvise.

In pharma and biotech, teams too often design studies that protect the narrative rather than maximise Value of Information. They avoid the questions that could kill a project early because bad news carries career risk.

The espionage lesson is clear: you must want to know your tells. The best agents actively sought feedback from locals and trainers. They treated every near-miss as gold. High-performing teams do the same through Real-Time Clinical Trials, adaptive designs, and cultures that celebrate good kills.

Biotech Operating Like Pharma - and Vice Versa

The spy analogy becomes especially sharp when we examine how some biotechs adopt big-pharma habits (polished but defensive covers) while others maintain ruthless adaptability.

Alnylam Pharmaceuticals stands as a masterclass in asymmetric persistence. Founded on RNAi technology - a high-risk premise that many abandoned after early delivery setbacks - Alnylam iterated relentlessly on lipid nanoparticles and clinical learnings. Rather than papering over failures, they confronted them head-on. The result: multiple approvals (Onpattro, Givlaari, Amvuttra), successful expansion into cardiovascular and CNS indications, and a valuation that reached tens of billions, driven largely by organic platform strength rather than mega-M&A. They refined their “accent” under fire instead of switching stories.

Contrast this with biotechs that begin to “operate like pharma” too early: heavy process, risk-averse study design, and narrative protection. They construct elaborate PowerPoint covers that impress boards and investors but crumble when real-world tells emerge - marginal differentiation, inconvenient regimens, or unexpected biomarker data in Phase III. Many such programmes end in quiet wind-downs or fire-sale acquisitions once the market spots the off-accent.

Parabilis Medicines (formerly FogPharma) provides a compelling current example of staying true to a bold premise. Their Helicon peptide platform targets historically undruggable intracellular proteins such as transcription factors. Instead of compromising early for a safer but crowded therapeutic space, the company has raised hundreds of millions while remaining private longer than typical, pressure-testing the platform with rigorous data. Recent milestones - a major Regeneron collaboration on antibody-Helicon conjugates and plans for later-stage development - reflect an earned Promise: platform validation through honest proof, not premature scaling. Parabilis has retained biotech agility (rapid iteration on a novel modality) while forging smart pharma-style partnerships. The tell they have avoided? Diluting the premise to chase quick exits.

These cases illustrate the trap: biotechs that mimic big-pharma bureaucracy lose the learning velocity that made them dangerous in the first place. Meanwhile, pharmas that adopt a biotech mindset - bold premises, fast failure loops, and platform obsession - gain asymmetric edge (witness Lilly’s GLP-1 journey or Regeneron’s repeated platform successes).

The Promise: Delivery Only After Honest Proof

No agent delivered real value - intelligence, sabotage, or networks - until their cover had been thoroughly stress-tested. The Promise was earned, never declared upfront.

This is why Target Product Profiles that survive genuine scrutiny create durable differentiation, while Compromise Product Profiles quietly fade in crowded markets. The market eventually detects the tells - lack of differentiation, marginal efficacy data, or inconvenient dosing - just as surely as Major Hellström spotted the fingers.

Asymmetric Edge in Chaotic Systems

Occupied Europe was noisy and chaotic: refugees, displaced workers, regional dialects, and a population focused on survival. A near-native blend often sufficed if you had local support and the ability to adapt rapidly.

Today’s pharma and biotech landscape is equally noisy - evolving regulators, real-world evidence, payer pressure, and competitor noise. The winners are not necessarily those with the most resources, but those with the fastest, most honest learning loops. They treat small signals (biomarker drift, patient feedback, competitor moves) as early tells. They update beliefs without ego. They build cultures where wanting the uncomfortable answer is the ultimate competitive weapon.

Final Thought

War stories romanticise the glamour (and I’m in no way attempting to minimise the reality), but the real lesson is epistemological: under extreme uncertainty, victory belongs to those who can update fastest on reality’s own terms.

Whether you are dropping behind enemy lines or steering a development programme, the playbook remains the same:

• Start with a premise worth defending.

• Ruthlessly eliminate your tells through pressure-testing.

• Only promise what you have honestly proved.

The market (or the enemy) has an excellent nose for theatre.

But every once in a while, the system shows cracks where genuine truth-seeking can break through - and the last month delivered a cluster of them.

1. FDA’s Real-Time Clinical Trials (RTCT) Pilot: Ending the “Don’t Look Too Closely” Era

On 28 April 2026, the FDA announced proof-of-concept real-time clinical trials with AstraZeneca and Amgen. Data - endpoints, safety signals - flows continuously to regulators via platforms such as Paradigm Health, rather than in dusty end-of-phase dumps.

• AstraZeneca’s Phase 2 TRAVERSE trial (acalabrutinib + venetoclax + rituximab in first-line mantle cell lymphoma, involving MD Anderson and Penn).

• Amgen’s Phase 1b STREAM-SCLC (tarlatamab in limited-stage small cell lung cancer).

An RFI is out for a broader AI-enabled pilot launching this summer.

This is high-VoI infrastructure. (Value of Information…) Traditional development wastes around 45% of time in “no active trial” limbo, with data siloed and incentives warped toward late revelation. RTCT forces faster, more honest loops: see the signal (good, bad, or neutral), decide, iterate. It aligns incentives with wanting to know - the core of asymmetric learning.

This ties directly to Richard Feynman’s approach to uncertainty. As Feynman emphasised, “I would rather have questions that can’t be answered than answers that can’t be questioned.” In drug development we too often prefer the comfort of delayed or curated answers. Real-time data flow challenges that directly: it demands we confront uncertainty early, iterate on weak signals, and resist the temptation to shield fragile premises. Without this cultural willingness to embrace doubt and revise beliefs rapidly, even the best infrastructure risks becoming expensive theatre. Early validated signals in AstraZeneca’s trial already hint at proof.

This could be the structural enabler that makes Premise → Proof → Promise actually promise something asymmetric.

2. Lilly Doubles Down on In Vivo CAR-T (Kelonia Acquisition, 20 April)

Lilly agreed to acquire Kelonia Therapeutics for $3.25 billion upfront plus milestones taking the total up to $7 billion. The prize: a lentiviral in vivo gene delivery platform and lead KLN-1010 (BCMA CAR-T for multiple myeloma), showing strong expansion, MRD negativity, and responses without lymphodepletion or ex vivo manufacturing hassle.

This follows Lilly’s earlier moves in the space and fits a broader wave of in vivo bets. Ex vivo CAR-T works but scales poorly. The premise here - reprogramme T-cells in the body for off-the-shelf-like simplicity - is bold. Early proof looks compelling enough for a massive commitment.

Asymmetric angle: This is not incremental line extension. It is a platform bet on modality innovation to escape the cost/complexity trap. High-conviction M&A like this rewards strong premises tested rigorously, rather than “safe” me-too assets. In a world of portfolio attrition, these are the moves that create real separation - provided the proof phase stays honest.

Lilly’s broader acquisitiveness (and the M&A wave around it) underscores a market voting for learning velocity over stasis.

3. Lilly’s Q1 2026: The Earned Promise of Tirzepatide

$19.8 billion revenue (+56% YoY), driven by Mounjaro/Zepbound volume explosion. Guidance raised. This is not hype - it is the endpoint of a sequenced journey: solid incretin biology premise → rigorous sequential proof across diabetes/obesity/CV → massive real-world promise.

It stands in contrast to programmes that falter because incentives favoured “forward motion” over hard questions. Here, learning loops delivered durable value. Expansions, orals, combinations - the asymmetry compounds.

Tying It Together: The Philosophy Test

These stories are not isolated. They test the Premise → Proof → Promise framework in real time:

• Premise: Strong and differentiated (real-time data flow; in-body reprogramming; multi-benefit incretins).

• Proof: Accelerated or de-risked through better tools/incentives (RTCT signals, early clinical data, sequential trials).

• Promise: Scaled impact when learning wins (revenue, platform potential, patient access).

The enemy remains the Compromise Product Profile and “we don’t actually want to know” study designs. But pockets of the industry are shifting toward asymmetric learning: faster feedback, bolder modality bets, and earned scale.

The market notices. Biotech IPOs rebounded in April (Kailera’s obesity play and others), fuelled by M&A appetite for high-conviction premises.

As always, this is about building better learning cultures in pharma — not cheerleading specific companies. Ideas > headlines.

Sources for easy reference (full URLs):

• FDA RTCT announcement

• Lilly Kelonia acquisition

• Lilly Q1 2026 earnings

• Kailera IPO context

But I’ve come to realise something more precise: the TPP itself is rarely the true villain. What actually does the damage is its evil twin - the Compromise Product Profile, or CPP.

The CPP is what you get when a single TPP meets reality, internal politics, risk aversion, and the irresistible urge to “be realistic.” It is not ambitious. It is not differentiated. It is the lowest common denominator that everyone can live with - and therefore the profile that excites no one, least of all patients or prescribers.

What the CPP Actually Is

The classic TPP is meant to be a North Star: a bold articulation of what the product should be to win in the market. The CPP is what emerges when that star is negotiated into mediocrity.

• Marketing wants best-in-class efficacy and convenience.

• Clinical says the data only supports “good enough.”

• Regulatory insists on conservative safety margins.

• Finance demands something achievable within budget and timeline.

The result? A profile that is safe, defensible, and almost guaranteed to be incremental. It feels professional and collaborative. In truth, it is the slow, quiet death of asymmetric opportunity.

This is not a target. It is a negotiated settlement. And biology, markets, and competitors do not reward settlements.

How the CPP Emerges (and Why It Feels So Reasonable)

It usually starts innocently enough. A team sits down with the best available data and begins “refining” the TPP. One by one, the edges are sanded off:

• “We probably won’t hit that efficacy bar - let’s lower it a bit.”

• “That dosing schedule is aggressive; patients might not adhere.”

• “To get broad labelling we’ll need to compromise on the primary endpoint.”

Each change feels pragmatic. Collectively, they produce a document that no longer aims at delight or differentiation - only at “viability.” The single story is preserved, but at the cost of any real edge.

This is exactly what I warned about in “Who puts the T in the TPP?” — the profile is rarely a true target. More often it becomes a projected or expected profile: a self-referential guess dressed up as strategy.

Worse, once the CPP is written, the entire organisation aligns around defending it, begins to forecast based on it, and uses it as the basis for ends as diverse as market research, advisory meetings and more. Any new data that points somewhere more exciting is viewed as a threat rather than an invitation. Montaigne would have recognised the pattern immediately: we have locked the pistol in a box and thrown away the key, terrified it might shoot somewhere unexpected.

The Damage the CPP Does

The Compromise Product Profile doesn’t just limit upside - it actively manufactures downside:

• It turns potential blockbusters into “me-too” or niche products.

• It rewards defensibility over delight.

• It creates the illusion of progress while the real opportunities gallop off in another direction.

• It symmetrises your risk: everyone ends up chasing the same crowded, mediocre space.

Look at the GLP-1 story I explored earlier. The winners did not compromise their way to success. They kept multiple ends in mind and let the data reveal the real profile.

The CPP would have produced another incremental diabetes drug. Instead, we got obesity medicines that changed medicine.

Killing the CPP: Thenning First, Multiple Ends, and Deliberate Optionality

The antidote is not to abandon profiles altogether. It is to refuse the slide into compromise from the outset.

Start with thenning - plural desired ends - before you begin iffing. Design multiple Target Opportunity Profiles rather than negotiating a single compromised one. Keep the fork garden alive. Run small, cheap experiments that let biology speak before the organisation has time to water everything down.

This is precisely what I’ve been advocating in pieces such as:

• Does the Then come before the If…?

• Designing Multiple TPPs with Multiple Ends in Mind

• The Worst Number of Options in Pharma? Zero. The Second-Worst? One.

• Attacking the TPP — crafting a better alternative

The goal is not a perfect single story. It is a living set of options that can evolve gracefully when the pistol shoots wide or the horse bolts in an unexpected direction.

A Call to Action

Next time you sit in a room where people are “refining” the TPP, ask the uncomfortable question:

“Are we building a true Target Product Profile - or are we quietly constructing a Compromise Product Profile?”

If it feels like the latter, stop. Reopen the forks. Restore the ambition. Keep the curiosity alive.

The CPP is seductive because it feels safe. But in pharma, as in life, the safe path is rarely the one that matters.

Kill the CPP before it kills your programme’s potential.

Imagine standing on a hill at first light. A vast landscape rolls out in front of you - rolling fields, dark forests, distant hills, a river catching the early sun, a plume of smoke from a farmhouse chimney, maybe the flicker of movement in the trees. With your naked eyes, you take it all in. You notice patterns: the way the light shifts across the valley, the shape of clouds hinting at weather coming in from the west, a subtle change at the edge of your peripheral vision that might be a muntjac or a fox. Your field of view is enormous - roughly 120–180 degrees horizontally. Resolution isn’t perfect everywhere, but it’s good enough, fast enough, and wide enough to scan, orient, and spot what deserves a closer look. Your eyes are built for discovery. (Well, your eyes - mine would miss everything further than 10 feet away…)

Now you lift a pair of binoculars. Suddenly that distant farmhouse leaps into razor-sharp focus. You can count the roof tiles, see the John Deere parked beside it, read the expression on the farmer’s face. The detail is astonishing. But the rest of the world vanishes. Your field of view collapses to perhaps 5–10 degrees. Everything else - the river, the treeline, the weather front, the movement you had just noticed - is gone, blacked out by the narrow tubes. To get any value from the binoculars at all, you first had to decide exactly where to point them. You had to rule out 99% of the landscape to inspect that 1% in high resolution.

This is not just optics. It is a near-perfect metaphor for how we observe, think, decide, and innovate - especially in pharmaceutical strategy, R&D, and market positioning. In my “Seeing Further” series on here I have written about the value of looking beyond the obvious horizon. But there is a crucial nuance worth exploring here: seeing further is not always about sharper focus on a single distant point. Sometimes the real advantage comes from seeing more things first, at a usable level of resolution, before you ever reach for the high-powered tools.

Human vision itself is designed this way. The fovea - the small central pit in the retina - delivers high-acuity colour vision (less than many animals/ insects, but good for a mammal), but it covers only about two degrees of arc (roughly the width of your thumbnail at arm’s length). The rest of the retina trades detail for breadth, motion detection, and change awareness. Evolution gave us excellent peripheral vision precisely because, in uncertain environments, knowing what to look at is often more valuable than knowing how to look at it once chosen. Binoculars (or telescopes, microscopes, dashboards, AI summaries, or focused due-diligence decks) amplify one dimension - resolution - at the direct expense of another: context and optionality. They force exclusion before exploration is complete.

In pharma and biotech this tension plays out constantly. The “eyes” mode is broad scanning: reading widely across indications, attending conferences without a rigid agenda, talking to clinicians and patients outside your core therapeutic area, reviewing old trial data that didn’t fit the prevailing narrative, noticing weak signals in adjacent fields. It is hypothesis generation, pattern recognition, and orientation. The “binoculars” mode is exploitation: deep mechanistic studies on a single target, competitive landscaping of three rival assets, detailed positioning work on one lead candidate, or granular market research on a narrow patient segment. Both are essential. The art - and the frequent failure point - is knowing when to switch and refusing to reach for the binoculars before the wide scan has earned you a direction worth magnifying.

Too many teams, companies, and even entire therapeutic areas default to binoculars too early. They stare “dead ahead” at the most crowded, most visible part of the landscape - the mechanism everyone else is chasing, the patient population with the biggest immediate revenue, the regulatory path that looks safest. They rule out the periphery before they have truly seen it. The result is incrementalism, crowded battlegrounds, and missed asymmetric opportunities.

One example that stays with me is the positioning of OCREVUS (ocrelizumab) - one of the best launches ever in our industry. When we supported our wonderful client to position this asset, we didn’t look dead ahead.

Around 2010–2015 the multiple sclerosis field was intensely focused on the obvious centre of the landscape. The dominant paradigm was T-cell-mediated inflammation in relapsing-remitting MS (RRMS). Multiple high-efficacy therapies were already approved or in late development, all competing fiercely on relapse rates, MRI lesions, and convenience. The binoculars were pointed squarely at that crowded patch. Many observers (and some companies) assumed that was the only game worth playing.

But a wider scan with the eyes revealed something different. Emerging data - some from off-label use of the earlier anti-CD20 molecule rituximab - hinted that B-cells might play a far more central role in MS pathology than the mainstream narrative suggested. At the same time, the scan picked up a much larger unmet need further out on the periphery: primary progressive MS (PPMS). No approved disease-modifying therapy existed for PPMS patients, who faced steady, relentless disability accumulation with almost no options. The landscape contained a quiet, high-consequence opportunity that the narrow view had largely ignored.

OCREVUS, a humanized anti-CD20 monoclonal antibody from Genentech/Roche, was developed and positioned not as another incremental player in the relapsing space, but as the first therapy capable of addressing both relapsing and primary progressive forms by targeting B-cells centrally across the MS spectrum. The positioning emphasized early high-efficacy use, slowing of disability progression (even in PPMS), and compelling MRI outcomes. It wasn’t about staring dead ahead at the conventional T-cell battlefield; it was about scanning the full immunological and clinical terrain first, spotting the broader unmet need, then applying focused development and commercial effort where it mattered most. The result was a true landscape-changer: approval in 2017 as the first and only therapy for both RMS and PPMS, rapid uptake, and blockbuster status that redefined what disease modification could mean in MS.

This wasn’t luck or a single brilliant insight. It was the disciplined outcome of sufficient wide-field scanning before the high-resolution work began. Had the team fixated only on the most visible, competitive slice of the market, the PPMS breakthrough and the B-cell repositioning might have been delayed or missed entirely.

We see the same pattern repeatedly. In oncology, companies that scan broadly across tumour biology and combination approaches sometimes uncover unexpected synergies that narrow “basket trial” thinking never reaches. In portfolio strategy, teams that insist on wide exploration across modalities (small molecules, antibodies, cell/gene therapies, even non-traditional platforms) surface asymmetric bets that the binoculars-only crowd dismisses as “off-strategy.” Conversely, organisations glued to the hot area of the moment - whatever the current vogue mechanism or patient population happens to be - frequently overpay, enter saturated spaces, or solve yesterday’s problem with tomorrow’s capital. We’re in the middle of the obesity focus still, despite being 5 years too late…

The risk is only growing. Modern tools - AI-driven literature summaries, real-time dashboards, competitive intelligence platforms - often function as super-binoculars. They deliver exquisite detail on whatever slice of data you already told them to watch, while making the wider landscape feel increasingly distant or irrelevant. The temptation to skip the eyes phase is stronger than ever.

So how do we get better at this?

First, protect deliberate wide-scanning time. Block time for unfocused exploration: reading outside your therapeutic area, attending sessions at conferences on topics you know nothing about, reviewing “failed” programs with fresh eyes, talking to people whose worldview differs from your own. Treat it as non-negotiable R&D.

Second, earn the right to zoom. Only reach for the high-resolution tools - deep target validation, detailed competitive war-gaming, granular positioning - after the broad scan has given you a defensible sense of direction. The best strategists I know keep asking: “Have we truly mapped the full landscape, or are we already pointing the binoculars at the first shiny object?”

Third, value and integrate cognitive diversity. Some people are natural wide scanners - intuitive, pattern-seeking, comfortable with ambiguity (this is me - zooming in hurts my brain). Others are natural zoomers - precise, analytical, relentless on detail. Neither is superior; both are incomplete. Organizations that design processes allowing scanners to inform where the zoomers point their effort create a genuine advantage. (I explored this in more depth in my earlier piece on cognitive diversity in “Seeing Further.”)

Fourth, in positioning and commercialisation specifically, resist the reflex to narrow immediately to competitive differentiation. Map the entire patient journey, the full scientific terrain, the adjacent unmet needs first. Only then decide which story is worth telling at high resolution. In reality, I have successfully positioned some of the biggest and best blockbusters over the past 30 years, but I probably lucked into it at first - this rational strategy (eyes, then binoculars) is just how I work, and it always passed the gut feel test for me…

The most powerful observers - whether individuals, project teams, or entire companies - master the fluid dance between eyes and binoculars. They use the wide view to see more possibilities than anyone else, then selective, disciplined high-resolution focus to turn the best of those possibilities into reality. In pharma’s asymmetric game of innovation, where the landscape is vast, the unknowns many, and the stakes enormous, seeing further almost always begins with seeing more broadly first.

The hill is still there. The landscape hasn’t changed. The only question is whether we lift the binoculars before we’ve earned the right to point them - or whether we trust our eyes long enough to discover what is actually worth looking at.

But reorder it - slot Proof firmly in the middle - and the whole framework transforms. What was a cautionary tale about bad strategy becomes a practical, honest philosophy for how pharmaceutical development should work. Premise. Proof. Promise. This is the sequence I’ve come to believe in after years watching (and occasionally helping) programmes succeed or stumble. It’s not flashy, but it’s robust, it respects the science, and it gives genuine innovation a fighting chance. Importantly it also leaves the messaging of the Proof until you genuinely know it is a promise rather than a hope - and promises something that a) you know you can deliver and b) that you know your audience wants/ values.

Start with a strong Premise

Everything worthwhile begins with the idea. Not the generic “there’s an unmet need in oncology/cardio/metabolic” slide that appears in every deck, but a sharp, hypothesis-driven premise about what this specific molecule might achieve in the real world.

Why this target? Why this modality at this moment in biological understanding? What precise patient experience or disease dynamic are we aiming to shift, and why do we believe our approach will succeed where others have fallen short?

A strong premise is falsifiable, ambitious yet grounded, and carries a genuine point of view. It draws on deep biology or clinical observation but adds that spark of asymmetric insight - the kind I’ve written about before. Choose the right premise and you create an immediate asymmetric learning advantage: every subsequent experiment, trial, and regulatory interaction becomes an opportunity to pull ahead while competitors are still splashing around in the same data pond.

This is where the metronome of innovation starts ticking properly. Rush or dilute the premise and you’re condemned to incrementalism from day one. Nail it and you’ve set the tempo for everything that follows.

Then build rigorous Proof

Only once the premise is locked down do you commit fully to generating proof that it actually holds water. And by proof I don’t mean the comforting slides that R&D and commercial teams love. I mean mechanistic understanding, translational data, and well-designed clinical evidence that the molecule does what you hoped - in the patients who matter.

This is the stage where most organisations lose their way. Commercial pressure leads them to write an optimistic Promise first, then scramble backwards to manufacture supporting data. The predictable consequences: late-stage surprises, label negotiations that strip away the headline claims, and post-approval commitments that quietly confess the original hype was ahead of the evidence.

When you protect the Proof phase, magic happens. You generate real optionality. You discover not just whether the drug “works,” but where it works best, in whom, and with what trade-offs. You can pivot intelligently, kill early when needed, or double down with conviction. Trial design stays scientific rather than defensive. Regulatory conversations become demonstrations of truth rather than exercises in persuasion.

This aligns closely with the asymmetric learning philosophy I’ve explored: the companies that win aren’t necessarily the ones with the biggest budgets or the loudest voice - they’re the ones that learn faster and more honestly about their molecule’s true nature.

Only then make the Promise

The Promise belongs at the end. Not as a clever slogan dreamed up. (or pieced together) in a workshop, but as a clear, evidence-backed commitment to regulators, payers, physicians, and - most importantly - patients.

“This molecule, grounded in this premise and proven through this body of evidence, delivers these specific clinical benefits, with this safety profile, generating this level of economic and humanistic value for these well-defined patients.”

Because the Promise flows directly from the Premise and the Proof, it feels earned rather than aspirational. Positioning finally earns its seat at the table - no longer papering over scientific gaps, but translating hard-won truth into language the market can understand, trust, and act upon.

Why this order matters - illustrated by real programmes

Run the sequence backwards (or start with the Promise) and the pattern is depressingly familiar: beautiful early decks, wobbly Phase 3 readouts, products that reach the market with modest differentiation and even more modest pricing power.

Contrast that with programmes that - whether consciously or by good fortune - followed something closer to Premise → Proof → Promise.

The GLP-1 receptor agonists are a textbook case. The foundational premise emerged from decades of incretin biology - understanding how these hormones regulate not just glucose but energy balance, appetite, and cardiovascular health. Rigorous proof followed: mechanistic studies, then clinical programmes that started in type 2 diabetes and expanded thoughtfully into obesity and cardiovascular outcomes. Only with that foundation did the powerful promise crystallise - substantial, durable weight loss and risk reduction that payers and patients could embrace as genuine value.

Pembrolizumab (Keytruda) followed a similar path. The premise was rooted in deep immunology: PD-1 blockade could unleash the immune system against tumours in ways that cut across traditional cancer classifications. Targeted proof in biomarker-defined populations (MSI-high, high TMB) came first. The resulting promise - including the landmark tumour-agnostic approval - felt biologically credible and has genuinely shifted treatment paradigms.

You see echoes of this disciplined sequence in other differentiated medicines that didn’t merely hit an endpoint but changed how clinicians think about a disease. They succeeded because the Promise was the conclusion of serious work, not the starting gun.

Practical implications for how we work

Adopting Premise → Proof → Promise as your guiding philosophy changes behaviours across the organisation:

• Portfolio decisions become premise-led first, with financial modelling as a supporting voice rather than the loudest one.

• Clinical development teams design trials around the scientific question, not the hoped-for label.

• Regulatory strategy focuses on demonstrating the premise with integrity, reducing the need for last-minute rescue negotiations.

• Early commercial and medical affairs involvement shifts from “what can we claim?” to the far more useful question: “what proof do we genuinely need, and how do we generate it efficiently?”

It also creates space for the kind of honest kill decisions that protect capital and reputation. In an industry that talks constantly about innovation but often delivers incremental me-too products, this sequence is one practical way to tilt the odds back toward meaningful progress.

Final thought

Pharmaceutical development is not marketing with some science attached. It is a scientific and medical endeavour whose commercial success should be the natural outcome of doing the hard parts properly.

The industry loves to celebrate disruption and breakthrough thinking. Fine. Here’s one concrete way to move beyond the slogans: start with a strong, asymmetric Premise, generate rigorous Proof that honours the biology, and only then articulate a Promise you can actually stand behind.

Get this sequence right and you don’t just launch products - you build durable competitive advantage and, more importantly, deliver medicines that genuinely matter.

Siloed metrics across Discovery, Development, and Commercial keep teams choosing “easy” deliverables over truth – rewarding people for “successful” phase II leads, inevitably, to doing phase IIs that succeed on their own merits, but do nothing to set up the phase IIIs that matter for market performance.

Winners align everyone around learning what actually matters for the label and the market.

In the world of pharma R&D, we’ve spent years talking about faster cycles, better decision frameworks, parallel teams, and shifting from rigid Target Product Profiles (TPPs) to more flexible Target Opportunity Profiles (TOPs). All of that matters.

But there’s a deeper layer that determines whether any of it actually works: Do you really want to know the answer? Or are you secretly hoping the data stays comfortably positive – or at least neutral enough not to rock the boat?

This is the motivational asymmetry at the heart of asymmetric learning. Some teams treat every study as a high-value information quest, where even a clear negative result is a win because it de-risks the portfolio, frees capital, and creates proprietary insight. Others design experiments that are almost guaranteed not to deliver bad news. They optimise for career safety, publication potential, or short-term portfolio optics instead of truth.

Value of Information (VoI) thinking makes this concrete.

VoI quantifies the expected benefit of reducing uncertainty before committing more resources. It asks: how much better could our portfolio decisions be with the answer to this specific question? A high-VoI study might reveal that an asset has no realistic path to a competitive label – freeing hundreds of millions for better bets. A low-VoI study simply confirms what we already suspect, delivering symmetric knowledge that competitors can replicate at little cost. The best organisations explicitly debate VoI upfront, prioritising experiments where even disappointing outcomes unlock disproportionate optionality.

The cost? Bloated pipelines full of zombie assets, late-stage failures that could have been spotted early, and billions wasted on programmes that were never truly tested against reality.

The Ostrich Effect in the Boardroom

Psychologists call it the ostrich effect: the tendency to avoid information that might be unpleasant, even when that information is decision-critical. In pharma, this shows up as teams quietly favouring surrogate endpoints, underpowered studies, or head-to-head designs that avoid the hardest comparisons. Why? Because a clean negative or neutral result can kill budgets, reputations, bonuses, and entire project teams. No one wants to be the person who delivered the data that sank the golden molecule.

Publication bias and regulatory realities amplify this. Positive data gets celebrated (and published). Neutral or negative data often gets buried or spun.

Contrast that with teams that treat negative data as an asset. When organisations move decisively on early signals – shelving, pivoting, or resurrecting based on truth – they preserve resources for higher-upside bets. The winners don’t just fail fast – they learn asymmetrically from the failure.

Why “Easy” Studies Feel So Productive (But Deliver Symmetric, Low-Edge Knowledge)

Most companies default to studies that are easy for structural and psychological reasons. And nowhere is this more visible than in the siloed metrics that govern different parts of the organisation.

Discovery teams are typically measured on volume: number of viable candidates pushed into the pipeline, hit-to-lead rates, IND filings, and preclinical proof-of-concept milestones. The incentive is clear – fill the funnel. Easier deliverables (surrogate signals, well-trodden mechanisms, or modest improvements) get assets moving forward faster and keep the numbers looking strong.

Early-to-mid Development (translational and clinical teams) often track pipeline progression rates, Phase 1/2 entry numbers, and “on-time” advancement to the next gate. The reward is momentum, not resolution of the toughest uncertainties.

Late-stage R&D is ultimately judged on approvals (NDAs/BLAs), time-to-market, and first-pass regulatory success rates. But the pressure to hit quarterly portfolio reviews or annual bonus thresholds means teams are incentivised to keep assets alive long enough to “show progress” – even when deeper questions remain unanswered.

Meanwhile, Commercial, Medical Affairs, and Market Access live downstream: they care about label strength, reimbursement evidence, payer value propositions, and real-world differentiation. But those requirements – clinically meaningful endpoints, comparative data, subgroup analyses that actually support pricing and access – often aren’t baked into upstream metrics. Discovery and early Development only need to learn “just enough” to clear the next internal hurdle, not what will actually make the product successful in the market.

Siloed metrics across Discovery, Development, and Commercial keep teams choosing “easy” deliverables over truth – rewarding people for “successful” phase II leads, inevitably, to doing phase IIs that succeed on their own merits, but do nothing to set up the phase IIIs that matter for market performance.

The result is perfectly rational behaviour inside each silo: design studies that deliver easy, positive-ish signals for your scorecard. Symmetric learning that everyone else can replicate. Low Value of Information (VoI). And a quiet avoidance of the hard experiments that would reveal whether the asset truly has label-enabling potential – or whether it should be killed early to free resources for better bets.

This isn’t laziness or stupidity. It’s human nature meeting perfectly misaligned corporate incentives. The ostrich effect isn’t abstract psychology – it’s why “fast-fail” remains more slogan than widespread practice in many organisations.

When Wanting to Know Creates Asymmetric Advantage

The teams that pull ahead treat curiosity as a competitive weapon. They set up experiments precisely to answer the uncomfortable questions early – and they align metrics across functions so that everyone is rewarded for truth, not just forward motion.

Real patterns in the market today show this shift: companies embracing parallel bets, modular development, and rapid iteration are effectively voting for asymmetric learning. They run more hypotheses per cycle, fail cheaper, and pivot faster because they want the data – good, bad, or surprising. And crucially, they redesign incentives so Discovery isn’t just pushing volume, but generating candidates with the depth of insight Commercial will actually need.

A real-world illustration: Pfizer’s torcetrapib decision

In 2006, Pfizer faced a stark test of the “want to know” mindset. After investing around $800 million over 15 years, the CETP inhibitor torcetrapib – widely expected to succeed Lipitor as a blockbuster – showed in a large Phase III trial that it raised “good” HDL cholesterol as hoped but also increased blood pressure, heart failure risk, and mortality. Rather than spin the surrogate benefits or push forward in hope of a narrower label, Pfizer immediately halted the programme on the independent data monitoring board’s recommendation. The decision was painful and costly in the short term, triggering an 11% drop in market value and forcing a strategic rethink. Yet it exemplified high-VoI action: the negative result provided clear, actionable information that prevented far larger downstream losses and redirected capital toward other opportunities (including the eventual pivot to biologics via the Wyeth acquisition). Contrast this with programmes that linger on weak or neutral data – the torcetrapib case shows how embracing the uncomfortable truth can protect portfolio health even when it stings.

One pattern stands out: organisations that publicly celebrate “good kills.” When a project is terminated early based on rigorous, pre-set criteria (and those criteria include label-readiness signals), the team responsible gets recognised, not sidelined. This flips the incentive structure. It turns truth-seeking into a rewarded behaviour rather than a career risk.

Other operational signals of a “want to know” culture:

• Pre-defined kill criteria and truth-seeking rituals (triage sessions that force uncomfortable questions before post-mortems).

• Parallel teams where one subgroup is explicitly tasked with challenging core assumptions and flagging label/commercial gaps early.

• Leadership language that frames every study as “What will we learn – for the label and the patient?” rather than “How do we make this advance?”

• Cross-functional metric reforms: track “uncertainties resolved,” “insights per dollar,” “good kills that improved portfolio health,” and early label-enabling data generation – not just pipeline count or approval volume.

In sparse-reward environments like drug discovery, where most bets fail, the asymmetric upside of genuine curiosity (and aligned incentives) is enormous. A small investment in the right hard question can yield outsized knowledge that no competitor possesses.

From Ostrich to Eagle: Making “Want to Know” Operational

Shifting culture isn’t about slogans. It’s about structure and incentives.

• Embed VoI thinking upfront: For every proposed study, explicitly debate the value of all possible outcomes – especially the negative one. What portfolio options does it unlock? What capital does it free? Does it address what Commercial will need for the label?

• Design for asymmetry: Favour experiments that can deliver uneven knowledge gains and label-critical insights. Use adaptive designs, biomarker-driven cohorts, or modular protocols that allow quick pivots.

• Normalise the good kill – and align metrics: Track and celebrate early terminations that demonstrably improved overall portfolio health. Reform scorecards so Discovery is rewarded not just for volume, but for candidates with de-risked, label-relevant data. Make “insights generated” a shared KPI across R&D and Commercial.

• Build curiosity infrastructure: Curiosity isn’t soft – it’s the fuel for asymmetric learning. Encourage “Feynman-style” ignorance: reward people who ask the basic questions others assume are settled, including the commercial ones.

• Measure what matters: Move beyond enrolment metrics or p-values. Track “uncertainties resolved,” “label-readiness score,” or “portfolio health improvement from kills” as key leading indicators.

The companies winning today aren’t necessarily smarter or luckier. They’re the ones whose incentives make wanting the truth more rewarding than protecting the narrative.

The Bottom Line: Learning Is the Moat

Asymmetric learning was never just about mechanics – parallel execution, flexible frameworks, or better tools. At its core, it’s about epistemology in a high-stakes environment: Do you prioritise comfort and siloed milestones, or do you prioritise truth, optionality, and cross-functional alignment on what success actually looks like?

The market is already voting. Investors reward agility and decisive portfolio management. Regulators and payers demand clearer evidence of value. The old prediction paradigm – lock in assumptions early, execute linearly, hope variance doesn’t bite – continues to deliver disappointing productivity.

The edge belongs to teams that run straight at uncertainty because they know the learning itself is the durable advantage. Negative data isn’t failure; it’s raw material for the next asymmetric bet – especially when every function is measured on whether that bet will actually succeed in the real world.

If your organisation still designs studies to avoid bad news (or only learns what the next silo needs), you’re not just slowing learning – you’re ceding the future to those who want to know more than you do.

April, fortunately for my prediction value’s sake, has doubled down on that signal with all the subtlety of a double-decker bus.

On 1 April – no, not an elaborate prank – the FDA approved Eli Lilly’s Foundayo (orforglipron), the first oral small-molecule GLP-1 receptor agonist for weight management that can be taken any time of day, with or without food or water. In the ATTAIN-1 trial, patients on the top dose shed an average of 27 pounds (roughly 12 per cent body weight) over 72 weeks when they stuck with it. No refrigeration required, no injection-site hesitation, no “did I remember the pen?” drama. It is, in the nicest possible way, a game-changer for patient convenience and global scalability. Lilly didn’t build this one entirely in-house; they licensed the asset years ago from Chugai as a modest, low-cost modality probe running in parallel with their injectable franchise. That early, uneven bet – cheap optionality rather than an all-or-nothing wager – is now delivering asymmetric returns in an increasingly crowded metabolic arena.

The approval arrived under the FDA’s shiny new Commissioner’s National Priority Voucher programme – just 50 days after filing, making it one of the fastest new molecular entity approvals since 2002. This isn’t the regulatory wild west some commentators suggested; it is tempo done properly. A metronome keeping the rhythm while still allowing the orchestra to improvise. It imposes order on the chaos without strangling the creativity – precisely the sort of regulated flexibility I’ve been discussing in recent posts.

At the same time, real-world evidence on the broader GLP-1 class continues to throw up delightfully uneven, high-upside signals that no original Target Product Profile could possibly have foreseen. A large VA study of more than 600,000 veterans (published in the BMJ in early March, with follow-on ripples through April) found that GLP-1 users enjoyed significantly lower risks of new substance-use disorders compared with those on SGLT2 inhibitors: alcohol (hazard ratio 0.82), opioids (0.75), cocaine, nicotine, cannabis – the lot. Among those already wrestling with pre-existing SUDs, the benefits were even more striking: fewer emergency-department visits, hospital admissions, overdoses, and even mortality. This was never in the original TPP. It is serendipity writ large – biology’s way of whispering, “You thought this was just for diabetes and obesity? Hold my beer.” Rather like spotting the cleaner, blockbuster-friendly profile of fexofenadine lurking in the supposedly “dangerous” metabolite of terfenadine. The very reward-circuit dampening that drives weight loss appears to have broader, delightfully asymmetric applications in addiction.

Meanwhile, the M&A wave I flagged in March has kept rolling along like a well-mannered queue at my local post office. Gilead extended its Arcellx tender offer (partnership-as-probe quietly turning into full integration for CAR-T optionality) after securing all regulatory approvals. Lilly, never one to rest on its metabolic laurels, pounced on Centessa Pharmaceuticals at the end of March for its orexin/sleep-neuro assets in a deal potentially worth up to $7.8 billion. Merck, Biogen, and others added targeted platforms and late-stage gems amid the gathering 2026 patent-cliff clouds. These are not desperate all-or-nothing bets on perfect foresight. They are staged, deliberately uneven knowledge gains: learn cheaply or in parallel, kill the duds swiftly, scale the winners, and always keep multiple shots on goal (or multiple shots on different goals, as I prefer…). As any decent poker player will tell you, one option remains the second-worst choice in this environment – right after folding entirely.

Why This Matters for Asymmetric Learners

The GLP-1 story – from early incretin science through today’s oral convenience play and these emergent addiction signals – illustrates the core asymmetry beautifully: biology reveals value unevenly, often at the most inconvenient moments. Symmetric prediction (lock in one TPP early, march linearly, commit heavily upfront) gets punished when the unexpected arrives. Asymmetric learning wins by design, and April’s cluster shows it compounding in real time.

Consider the elements:

• Optionality early and cheap: Lilly’s licensing of orforglipron was never a make-or-break affair. It was a sensible probe running alongside the peptide assets, quietly generating differential learning on patient preferences, manufacturability, and real-world behaviour. The sort of low-stakes bet that symmetric planners often dismiss as “distraction”.

• Real-world evidence as the ultimate asymmetric probe: RWE is no longer a polite afterthought; it surfaces non-obvious edges (addiction risk reduction) that can reshape pipelines and indications faster than any planned Phase 3 could dream. It is learning-by-doing at population scale – messy, serendipitous, and gloriously uneven.

• Tempo without the rigidity: The fast-tracked approval plus post-marketing commitments (Lilly is already advancing diabetes filings and additional safety data) shows regulatory flexibility that rewards execution speed while still demanding continued curiosity. Metronome as a starting point, not metronomic tyranny.

• Lagom teams and disciplined execution: The winners balance curiosity with discipline – enough structure to keep moving at a decent clip, enough openness to pivot on serendipity without descending into chaos. Not too rigid, not too loose; just right.

In a sparse-reward environment where most R&D bets still fail and patent expirations loom, the market is rewarding organisations that treat development as uneven exploration rather than symmetric forecasting. Platforms, staged deals, modality diversity, and a healthy respect for emergent data are creating moats that rigid planners cannot replicate.

This is not theory. April’s little cluster – Foundayo’s swift approval, the ongoing Gilead/Arcellx momentum, GLP-1’s broadening real-world profile, and the broader M&A tempo – shows the philosophy quietly compounding. The vote is not changing. The only question is whether more organisations will have the good sense to shift their operating models to match it.

Four years ago, I wrote about a quiet killer in pharma pipelines: the refusal to properly account for opportunity cost. The roads not taken. The options we pretend don’t exist until it’s too late.

Nothing has changed about the definition. Everything has changed about the cost of ignoring it.

Between 2025 and 2030, more than $300 billion in pharmaceutical revenue is set to lose patent protection - roughly one-sixth of the industry’s annual sales. R&D costs per approved asset still hover around or above $2 billion (which is a terrible average, running from $1bn in some companies, up to north of $10bn in some…). Meanwhile, AI has exploded the number of viable scientific paths we can realistically explore. The old “follow the science” excuse for narrow focus now looks less like prudence and more like strategic blindness. (Science isn’t an answer in itself - it is a way to answer better questions…)

In 2022, I used a screenshot of the CGRP migraine landscape as an example of herd mentality. It showed where CGRP could go - but you know that CGRP is mostly explored as a migraine asset.

Today, picture the obesity/ GLP-1 space in 2026: oral candidates launching, combination therapies advancing, next-generation mechanisms in hot pursuit. Late movers or single-bet teams are watching value evaporate, not because their molecule failed, but because they never properly priced the alternatives they ignored.

The Matthew Hennessy quote I loved then still rings true: “Opportunity cost is the cost of the next best alternative foregone.” In pharma terms, it’s the value you sacrifice by committing resources to one path while the clock ticks on dozens of others.

Four Years Later: What I’ve Seen

When I hit “publish” in 2022, I was frustrated by arthritic portfolio processes that treated uncertainty as an inconvenience rather than the central feature of drug development. Since then, I’ve watched teams - some I’ve worked with directly, others from a distance - lose billions in forgone value. Not because they picked the wrong molecule, but because they never rigorously valued the ones they didn’t pick.

The best performers have quietly shifted towards something closer to a venture capital mindset: treating early programmes as real options with asymmetric upside, maintaining pluripotency longer, and reallocating resources ruthlessly once data clarifies the landscape.

AI didn’t invent this problem. It just made the opportunity cost brutally visible - and the tools to manage it suddenly affordable. What used to take months of manual literature review and gut feel now happens in days with multimodal models and generative platforms. The “3-second answers” I called for are no longer a fantasy.

Yet many organisations still default to the simplify trap: one target, one TPP, one narrow path forward. “We’ll decide when the science is clearer.” In an AI-accelerated world, that delay is even more expensive.

Simplify vs Complexify: The Choice Has Got Sharper

Simplify feels safe. One asset, one clear story for the board, easier resource allocation. But it carries hidden costs:

• Herd behaviour in hot spaces (remember the rush into certain oncology mechanisms or today’s GLP-1 follow-ons).

• Premature pruning of promising avenues.

• Zero visibility of the true opportunity cost when resources are locked in.

Complexify (or sophisticate, if you prefer) means deliberately holding multiple plausible paths longer. Mapping the landscape early. Assigning rough high/medium/low values to alternatives. Treating exploratory work as an explicit option with quantifiable value.

Four years on, complexify is no longer a nice-to-have. With AI platforms evaluating hundreds or thousands of hypotheses quickly, it has become mandatory. Teams that stay “pluripotent” - modular resourcing, a sensible split between lead optimisation and exploration - are better positioned to pivot when new data emerges or competitors stumble.

Time itself has become an even more expensive dimension. Being third (or fourth) to a differentiated mechanism in a crowded field hurts more when discovery cycles have compressed from years to months.

Real Options Thinking in Practice

The smartest operators I’ve seen treat early-stage work as real options. They ask:

• What is the value of the information this experiment will generate?

• How does keeping this path open affect our overall portfolio asymmetry?

• What resources are we truly forgoing elsewhere?

They kill fast when the data disappoints - not because of sunk cost, but because they can see the opportunity cost on the dashboard. Zombie programmes still exist, but the best portfolios are cleaner precisely because opportunity cost is made explicit.

AI augments this beautifully. It doesn’t replace judgement; it makes the trade-offs clearer and faster. Virtual experiments, predictive models, and dynamic landscape maps turn static PowerPoint decks into living decision tools.

Questions to Ask Your Team Today

Next time you’re in a portfolio review, try these (updated for 2026):

• What are the top three alternative paths we could be pursuing in this space, and what rough value do they carry?

• If we delay this decision by six months, what real options are we explicitly killing?

• How is AI helping us visualise and quantify the opportunity cost landscape — or is it just speeding up the same old narrow thinking?

• Are we maintaining enough pluripotency in the team and resource allocation, or have we already committed too early?

One option is still the second-worst number in pharma. Zero remains the worst.

The 2026 Imperative

In 2022, complexifying felt ambitious. In 2026, with a $300 billion patent cliff bearing down and AI making sophisticated exploration practical, the question has flipped.

It is no longer “Can we afford to complexify?”It is “Can we afford not to?”

The roads not taken have never been more visible. Or more expensive to ignore.

But what if the real differentiator isn’t sharper tactics or flawless operational delivery? What if it is the quality of the strategy itself?

This is the heart of true launch excellence: building an excellent strategy - one that is bold, differentiated, patient-centred, and grounded in deep insight - and then executing it with discipline and agility. In practice, an excellent strategy executed with greatness consistently outperforms a merely great (or solid) strategy executed perfectly.

What “Launch Excellence” Really Means in Pharma

Launch excellence is not a buzzword for slick campaigns or on-time deliverables. It is the disciplined alignment of a high-quality commercial strategy with the organisational capability to bring it to life across global, regional, and local levels.

Core elements typically include:

• A sharply defined, insight-driven value proposition rooted in unmet needs, patient journeys, and evolving stakeholder priorities.

• Clear indication sequencing, positioning, and competitive differentiation.

• Integrated evidence, market access, and medical strategies that support the commercial ambition.

• Cross-functional alignment and organisational readiness.

• Mechanisms for ongoing measurement and adaptive optimisation.

The critical insight from my years of launch analyses is that the ceiling on performance is largely set by the strategy. A mediocre or “safe” strategy - well-positioned but not truly differentiating - has a built-in performance cap, no matter how flawlessly the tactics are delivered. Communicating a vague notion of your product perfectly can only leave people with a vague idea of where it fits.

The Strategy-First Perspective

As Sun Tzu famously observed in The Art of War: “Strategy without tactics is the slowest route to victory. Tactics without strategy is the noise before defeat.”In pharmaceutical commercialisation, this ancient wisdom rings especially true. Strategy is where the hard choices are made: which patient segments to prioritise first, how to sequence indications, what value story will resonate with payers and physicians in a crowded market, and where to accept trade-offs.

An excellent strategy does several things exceptionally well:

• It identifies and owns a distinctive space in the treatment paradigm (biomarker-driven, paradigm-shifting, or addressing a genuine unmet need others have overlooked).

• It anticipates competitive responses, regulatory shifts, and payer evolution.

• It sets an ambitious but achievable aspiration that energises the organisation and guides resource allocation.

A merely great strategy might be sound, evidence-based, and internally consistent (probably with a high degree of buy-in) - but it often lacks that extra edge of differentiation or foresight. It plays in a contested space without reshaping the conversation.

Execution (the tactics) then becomes a simple multiplier. Great execution - agile, aligned, detail-oriented, and adaptive - amplifies the strengths of an excellent strategy. Perfect execution of a merely great strategy, by contrast, can deliver respectable results - but rarely the breakout performance that creates category leaders or multi-billion-dollar franchises.

Case study 1: Excellent strategy – Merck’s Keytruda (pembrolizumab) in immuno-oncology
Keytruda entered a competitive PD-1 space as a second-to-market product behind Bristol Myers Squibb’s Opdivo. Rather than competing head-on in broad populations, Merck pursued a bold, biomarker-driven approach. It focused initially on patients with high PD-L1 expression (a clear predictor of response), secured accelerated approval in melanoma ahead of its rival, and invested aggressively in a vast clinical programme (over 1,600 trials) to expand indications rapidly - eventually reaching dozens of tumour types and tumour-agnostic approvals (e.g., MSI-high). This precise positioning, combined with foresight on indication sequencing and tissue-agnostic potential, allowed Keytruda to own a differentiated narrative, reshape treatment paradigms, and become one of the best-selling drugs in history, far outpacing Opdivo in long-term sales despite starting later.

Case study 2: Merely great strategy with strong execution – Bristol Myers Squibb’s Opdivo (nivolumab)
Opdivo launched first in some markets with solid clinical data and broad initial positioning (no strict biomarker restriction in certain indications). Execution was strong: effective field force activity, medical education, and market access efforts followed standard high-quality playbooks. I’d also suggest it ‘won’ launch, being ahead for a crucial few years.

However, the strategy lacked the sharp differentiation and biomarker focus of its rival in key areas such as first-line lung cancer. When broader populations failed to show the expected benefit in later trials, momentum slowed. Despite excellent tactical delivery, Opdivo could not fully reshape the conversation or match Keytruda’s expansive trajectory, illustrating how a sound but less distinctive strategy caps upside even with polished execution.

Why does this hold?

• Markets reward differentiation more than operational polish. Physicians and payers have limited bandwidth; a compelling, ownable narrative cuts through noise where incremental improvements do not.

• Excellent strategy reduces execution risk by providing clear direction, prioritisation, and decision rights. Teams know what matters most and can adapt without losing the plot.

• In today’s environment - accelerated approvals, intense competition, pricing pressure, and complex stakeholder ecosystems - strategy sets the trajectory, and does so with enough of a runway to make a difference. Execution determines how closely you follow it.

Execution drift is a real danger: even the best strategy can erode if tactics, incentives, and behaviours slowly diverge from the original intent. But the root cause is often a strategy that was never sharp enough to maintain alignment in the first place.

Evidence from the Field

Look at launches that have redefined their spaces. Many succeeded not because every tactical element was executed without flaw, but because the underlying strategy was excellent: precise first-indication focus, biomarker-driven positioning, or a value story that reshaped how the disease is managed. Subsequent expansions and optimisations built on that strong foundation.

Case study 3: Excellent strategy – Novo Nordisk’s semaglutide (Ozempic/Wegovy) in diabetes and obesity
Semaglutide was launched first for type 2 diabetes as Ozempic with a strong but conventional positioning. The real strategic masterstroke came with the decision to pursue and heavily invest in the obesity indication (Wegovy), backed by compelling STEP trial data showing significant weight loss. Novo anticipated the paradigm shift in treating obesity as a chronic disease, built a patient-centred value story around cardiometabolic benefits, and sequenced indications while investing in supply and access infrastructure. This bold, foresight-driven approach created a new category leader, drove explosive growth, and turned semaglutide into a multi-billion-dollar franchise that redefined expectations for metabolic therapies - despite later supply challenges and competition.

Case study 4: Merely great strategy – A typical “me-better” oncology or immunology follow-on product
Many second- or third-in-class therapies in crowded fields (e.g., certain later PD-1/L1 inhibitors or JAK inhibitors) enter with robust phase 3 data, solid evidence packages, and well-resourced launches. Execution is often strong - comprehensive digital campaigns, key opinion leader engagement, and payer negotiations follow best-practice templates. Yet without a distinctive biomarker focus, unique patient segment ownership, or a narrative that truly reshapes the treatment paradigm, they achieve respectable but not category-leading uptake. First-year forecasts are frequently missed not due to poor tactics, but because the strategy failed to create meaningful separation in a payer-constrained, noise-filled market.

Conversely, products with solid science and merely competent positioning - cobbled together from team workshops and research decks - have often underperformed when the strategy failed to create meaningful separation from competitors, despite strong field force activity, digital campaigns, or market access efforts. Perfect delivery of incremental tactics rarely overcomes strategic shortcomings in crowded or payer-constrained markets.

Industry observations reinforce this: many launches miss near-launch forecasts not primarily because of poor tactical execution, but because the strategy underestimated competitive dynamics, misjudged value perception, or lacked the boldness needed to break through.

Building the Execution of Excellence

If strategy is the foundation, how do you operationalise this thinking?

1. Invest disproportionately in strategy quality. Spend the time (and external challenge) needed to pressure-test assumptions, explore scenarios, and refine differentiation. Ask: Does this strategy truly own a distinctive space, or is it “me-better” in a noisy field?

2. Design execution to serve the strategy. Remember: a great strategy might need a team to diverge from a template launch plan. Build alignment mechanisms, decision rights, and agility into the launch plan from day one. Focus resources on the few things that will move the strategic needle rather than trying to perfect every activity.

3. Accept “great” execution as sufficient when the strategy is excellent. Empower teams to adapt intelligently within clear guardrails. Speed and judgement often matter more than perfection in dynamic markets.

4. Measure strategic health, not just tactical KPIs. Track whether the core value narrative is landing, whether priority segments are responding, and whether the organisation remains aligned with the original strategic intent.

5. Treat launch capability as strategic. Institutionalise rigorous post-launch reviews focused on what the strategy got right or missed, so future strategies improve.

The Bottom Line

In pharmaceutical commercialisation, strategy sets the ceiling; execution determines how close you get to it.

Teams that master the execution of excellence start with an excellent strategy—bold, differentiated, and insight-led. They then execute it with greatness: disciplined, adaptive, and relentlessly focused on protecting and amplifying that strategic intent.

The result is launches that don’t just meet expectations but redefine what’s possible for patients and for the business.

Perfect execution is admirable. But as Sun Tzu reminds us, tactics without a superior strategy amount to noise before defeat. In an industry where one launch can shape a decade of performance, excellence begins - and largely ends - with strategy.

In music, a metronome does one seemingly simple thing: it clicks. Steadily. Relentlessly. It imposes measure on something as fluid and emotional as time. From the Greek métron (measure) and nómos (law, custom, or regulation), the word itself carries a quiet power. A metronome doesn’t compose the melody or write the symphony. It simply ensures that the orchestra doesn’t descend into beautiful chaos.

Pharmaceutical innovation faces its own version of that tension every day. Discovery is messy, serendipitous, and often asymmetric. Clinical data arrives in unpredictable waves. Regulatory “laws” constrain every beat. Markets shift. Yet we still need reliable progress - a steady tempo that turns raw ideas into medicines that reach patients.

What if we thought of asymmetric learning as a kind of intellectual metronome for the industry?

Measure Without Suffocation

The first root - métron - is about establishing a standard by which things are judged. In pharma, we are drowning in measures: timelines, probability of success (PoS), net present value (NPV), innovation indices, cost per launch, peak sales forecasts (attrition rates by phase is one of my betes noires…). Some of these are helpful. Many become targets, and as Goodhart’s Law reminds us, when a measure becomes a target it ceases to be a good measure.

A metronome doesn’t dictate what you play; it only sets the pulse so the piece can breathe. Similarly, asymmetric learning isn’t about adding more rigid KPIs to an already overburdened R&D machine. It’s about choosing smarter measures early - ones that maximize optionality and reveal high-upside branches before the big money is spent.

Instead of designing a single “if-then” development plan, we design for multiple possible “thens.” We learn faster, cheaper, and more asymmetrically in the early phases: exploring several hypotheses in parallel, failing small and fast, pivoting with agility. The measure shifts from “did we hit the predicted timeline?” to “how many high-value options did we create or eliminate this cycle?” That shift in measurement creates the conditions for true competitive advantage.

Regulation That Enables, Not Just Constrains

The second root - nómos - speaks to law, custom, and assigned order. Pharma lives under some of the heaviest regulatory nómos in any industry: FDA/EMA guidelines, clinical trial requirements, pricing and reimbursement rules, pharmacovigilance obligations. These are necessary. They protect patients. But they can also impose a mechanical, one-size-fits-all rhythm that punishes deviation.

A metronome’s “law” is not oppressive when used well. Great musicians play around the click - they push and pull with rubato, adding expression while staying grounded in the underlying pulse. The best performances feel free precisely because the structure is reliable.

Asymmetric learning applies the same principle to regulated environments. We accept the steady regulatory beat (stage gates, safety requirements, evidence standards) but engineer flexibility within it. We build programs that generate proprietary insights others can’t easily replicate. We treat early data not as a binary go/no-go but as raw material for branching possibilities. Curiosity becomes the improvisational force; structured learning-by-doing becomes the steady click that keeps everything from drifting into expensive noise.

This is why some teams consistently outperform: they don’t fight the metronome of regulation. They use it to enable faster, deeper cycles of learning. They design experiments that are stage-appropriate yet rich in optionality. They prune losers early and double down on winners with asymmetric upside.

The Tempo of Real Innovation

Music without a metronome can be exhilarating for a moment - raw, emotional, free. But sustained creativity at scale needs rhythm. The same holds for pharma pipelines. Pure serendipity is wonderful when it happens. But waiting for lightning strikes is not a strategy.

Asymmetric learning provides the tempo:

• Early exploration sets a quick, light click - cheap, fast, hypothesis-rich.

• Mid-stage commitment shifts to a firmer pulse - focused but still flexible.

• Late-stage execution locks in the steady beat needed for regulatory success and commercial scale.

Recent examples illustrate this beautifully. Gilead’s staged partnership with Arcellx (leading to the $7.8 billion acquisition) wasn’t a blind bet on a preclinical asset. It was a low-cost probe that generated proprietary knowledge edges through co-development. When the data aligned, Gilead internalized the program with conviction - classic asymmetric learning that turned collaboration into decisive upside.

Generate Biomedicines shows the power of platform-level optionality. By integrating machine learning with biological engineering, they generate novel protein therapeutics on demand, exploring vast design spaces far beyond traditional trial-and-error. The platform creates multiple high-potential “thens” from a single generative engine - a faster, more asymmetric rhythm for discovery.

Alltrna’s tRNA platform takes a similar approach, using machine learning to engineer programmable molecules that could address thousands of diseases sharing underlying genetic mutations. (See my interview with Alltrna’s CEO here…) One core technology unlocks broad therapeutic potential - learning that scales asymmetrically across indications rather than being locked into narrow, linear paths.

And then there’s the classic Terfenadine (Seldane) story: a blockbuster non-sedating antihistamine withdrawn due to dangerous interactions, only for its metabolite to become the safer, hugely successful fexofenadine (Allegra). That pivot - from withdrawal risk to a new franchise - is asymmetric learning in its rawest form (if slower than ideal…): biology’s surprise converted into upside by teams willing to listen rather than double down on the original plan.

The beauty is that this rhythm leaves room for genius-level improvisation. It doesn’t eliminate serendipity; it creates the conditions where serendipity is more likely to be recognized and captured. It turns “learning by doing” into a repeatable advantage rather than a happy accident.

A Personal Note on Rhythm and Transition

After many years building IDEA Pharma, you may know that I recently stepped away to start a new chapter. My philosophy of asymmetric learning - the ideas, frameworks, and ways of seeing pharmaceutical innovation - remains unchanged and I’ll continue to develop the ideas within. If anything, this transition has sharpened my sense of tempo. I suppose even experts in pharmaceutical positioning, and failed musicians, like real musicians, sometimes need to reset the metronome: not to slow down or speed up arbitrarily, but to find a healthier, more productive rhythm that aligns structure with creativity.

The industry faces enormous challenges - rising R&D costs, compressed timelines, higher bars for differentiation, intense regulatory scrutiny. In that environment, the teams and leaders who thrive will be those who master the metronome: measuring wisely, regulating thoughtfully, and creating space for asymmetric leaps.

They won’t just keep time. They’ll make music that matters.

“For Swedes, cooking is about how we relate to food; how it makes us feel. You’ll no doubt have heard of hygge – the Danish word that embodies a feeling of cosy contentment through everyday pleasures. For Swedes, it’s all about lagom. English offers no perfect translation, but in essence the word means ‘just right’; not too much, not too little. This is a sense you can see reflected time and again in our food. We like balance – flavours that work together in harmony, that never overpower each other. Lagom also manifests in how we think about portions and ingredients – everything in balance. Literally translated, though, the word is a conflation of two others that, in fact, tell you everything you need to know about the way the Swedes think: lag (meaning ‘team’) and om (meaning ‘the whole’). For us, then, when it comes to life and food, it’s all about the whole team.

In that spirit, join me. Let’s dig in.”

That passage from an excellent Niklas Ekstedt cookbook I bought (I’m a huge collector of cookbooks…) recently stopped me. Not because I’m over-obsessed with Nordic cuisine, but because it reframes lagom in a way that feels tailor-made for pharmaceutical innovation. It’s not just “moderation” or “balance” in the bland sense. It’s team + whole – a philosophy where every element contributes in harmony, none overpowering the rest, so the collective functions at its peak.

(Me, with Niklas Ekstedt at Guy Ritchie’s house…)

This is exactly the structure I’ve been exploring for multifunctional (cross-functional)/ multi-disciplinary teams in pharma. Not silos, not star players, not top-down control – but a deliberate “whole team” where science, commercial, regulatory, clinical, and manufacturing voices integrate without friction. Where flavours harmonise. Where the team itself becomes the unit of competitive advantage.

Lagom in Practice

I’ve written before about what this looks like in high-performing pharma organisations.

In “The large and the agile”, I unpacked Eli Lilly’s model: taking a massive portfolio and breaking it into smaller, autonomous project teams that operate like independent biotech companies – each with its own leadership and even a dedicated “board of directors” for multi-year budgeting and go/no-go calls. It’s classic lagom: the stability and resources of Big Pharma, balanced with the ownership, speed, and focus of a lean biotech. No single function or layer overpowers; the whole team owns the outcome.

In “Seeing Further: Cognitive Diversity Needs a Plan in Pharma’s Asymmetric Game” (and its addendum), I argued that raw diversity isn’t enough – you need deliberate structures to make ten minds see further than one. Without a plan, cognitive diversity becomes noise or conflict. With it – through planned integration, clear roles, and psychological safety – it becomes consilience: ideas clicking together in ways no single expert could achieve. That’s the harmony lagom describes.

“Transactive memory – how to trust your team” took this further: shifting from individual knowledge silos to collective team memory (“knowing who knows what”). When the team operates as a true whole, trust scales, friction drops, and distributed expertise becomes a superpower. No one ingredient dominates; the collective memory does the heavy lifting.

And in pieces like “Jumping Together: Consilience as the Engine of Asymmetric Learning” and “Pharma’s worst bet”, I described leaders as gardeners who balance “artists” (creative exploration) and “soldiers” (disciplined execution) without letting either side overpower the other. The result? Teams that nurture diverse shoots, prune without ego, and engineer the collisions that turn uncertainty into edge.

These aren’t separate ideas. They’re all expressions of the same lagom principle applied to team design: everything in balance, the whole team as the unit, flavours that harmonise rather than clash.

The Payoff: Asymmetric Learning

Here’s where it ties directly into the core thesis of this blog.

Asymmetric learning is about designing in competitive advantage through uneven, high-reward knowledge gains. It’s the opposite of the old prediction paradigm (fixed TPPs, linear plans, symmetric risk models that treat every program like a coin flip). Instead, it’s learning by doing: flexibility, continuous information gathering, parallel paths, serendipity, and stage-appropriate decisions that let winners compound unevenly while failures stay cheap.

A lagom-style whole-team structure is the organisational prerequisite for this.

• When functions integrate in harmony (no one overpowering), you get faster, deeper consilience – the cross-functional “war-games” where early clinical data meets payer intelligence or DMPK insights reveal hidden blockbusters.

• When transactive memory and cognitive diversity are planned in, curiosity becomes institutionalised. Teams explore without maps because trust and shared memory lower the cost of dead ends.

• When you balance artists and soldiers at the team level (à la Lilly’s agile sub-teams or the gardener model), you can run multiple TPPs, embrace optionality, and design programs for many possible “thens” rather than one tidy path.

In short: lagom teams don’t just execute better. They learn asymmetrically – faster, cheaper, and with outsized upside when the data aligns. They turn biology’s surprises into franchise opportunities instead of costly pivots or withdrawals.

The Swedish etymology nails it: lag (team) + om (the whole). In pharma, the whole team – balanced, harmonious, multifunctional – is how you design asymmetric advantage into the portfolio from day one.

If…

Add the ellipsis and it opens up. It becomes possibility. It becomes the space where real optionality lives - not a single hoped-for future, but all the futures that might unfold depending on what we learn.

In pharma we live inside conditionals every day:

• If the molecule hits the target…

• If the safety profile holds…

• If the biomarker stratifies patients…

• If the regulator sees what we see…

• If the competitor doesn’t move first…

Most companies treat these “ifs” as risks to be estimated (guess at…) minimised or assumptions to be locked in early. What if we treated them as the raw material of competitive advantage instead?

This is the heart of asymmetric learning: designing programs not for one tidy “then,” but for many possible “thens” - and engineering the conditions that let us reach the best one when the data finally speaks.

Starting with the End(s) in Mind

We’ve all heard the advice: “Begin with the end in mind.” In drug development that usually means writing a single Target Product Profile (TPP) and marching toward it in a straight line.

But what if the smartest move is to begin with multiple ends in mind?

In earlier posts I’ve explored how early-phase optionality is reshaping biotech ROI, and why the market is increasingly voting for asymmetric approaches (see the recent pieces on Gilead, Generate, the FDA, and Merck). The companies pulling ahead aren’t the ones with the strongest single hypothesis - they’re the ones treating Phase I and II as an option-generation engine. Small, cheap bets that probe different indications, subpopulations, combinations, or formulations. They create branching paths so that when biology surprises them (and it always does), they aren’t starting over from zero.

That’s “If…” thinking in practice.

• If the initial indication shows modest efficacy but a clean safety profile… then we have a path to a chronic-use franchise.

• If a biomarker subset lights up… then we pivot to a precision-medicine play with higher pricing power.

• If an unexpected metabolite emerges… then we might have an entirely new asset (Terfenadine → fexofenadine from the last post).

In the Terfenadine story, a dangerous drug’s liability became a blockbuster opportunity precisely because the team kept the “If…” open long enough to listen to the data. Rigorous metabolite work turned a withdrawal risk into optionality. They didn’t double down on the original plan; they let the biology dictate a better “then.”

The Grammar of Uncertainty

Philosophers and logicians have spent centuries dissecting the word “if.” Is it a sufficient condition? A necessary one? A mere supposition?

In pharma the ambiguity is practical, not academic.

We pretend many things are known when they are only “knowable” - or worse, unknowable until we run the experiment. As I’ve written before on knowns, unknowns, knowables, and unknowables, the dangerous illusion is believing we can predict the regulatory pathway, the competitor landscape, or the exact patient population years (with any useful accuracy) in advance.

“If…” thinking rejects that pretence. It keeps more paths alive until the knowables become knowns. It values flexibility over false precision.

This connects directly to why the market increasingly votes for asymmetric approaches. Recent moves show capital flowing toward teams that design for uneven upside rather than symmetrical, mediocre outcomes.

Winners compound. Losers are pruned early and cheaply. That asymmetry only emerges when you’ve built real optionality - which starts with honest “If…” questions rather than rigid plans.

Evaluating Multiple TPPs: From Single Target to Multiple Ends

Traditional TPPs are meant to be the “end in mind” - a clear articulation of the desired label, efficacy thresholds, safety margins, dosing convenience, and commercial positioning. In theory, they align R&D, regulatory, and commercial teams.

In practice, a single TPP often becomes a trap.

As I’ve argued in posts like “Attacking the TPP” and “Who puts the T in the TPP?”, the conventional single TPP tends to:

• Lock in assumptions too early, turning it into a projected profile rather than a true target.

• Encourage linear, symmetrical development - chasing one set of numbers that may no longer reflect the evolving market or biology by the time you reach Phase 3.

• Stifle innovation by focusing inward (what the organisation can deliver) rather than outward (what the market actually needs or what biology might reveal).

• Create shifting goalposts when new data arrives, or worse, force teams to ignore promising signals that don’t fit the original script.

The result? Many programs symmetrise their risk: average probability of success, average returns, and a high chance of becoming “one of the 99%” that fail to differentiate.

Now contrast that with designing multiple TPPs from the outset - the approach I explored in the GLP-1 piece on “Designing Multiple TPPs with Multiple Ends in Mind.”

Here, you explicitly map several plausible “thens” and the enabling “ifs” for each:

• TPP A: Best-in-class for the lead indication (high efficacy bar, premium pricing).

• TPP B: Broader chronic use with superior safety/convenience (lower efficacy acceptable if tolerability wins).

• TPP C: Precision play in a biomarker-defined subset (smaller population, but higher response rate and value).

• TPP D: Platform or combination opportunity (leveraging unexpected mechanistic insights).

Evaluation of the multiple-TPP approach:

Strengths:

• It forces early commercial and scientific imagination. You evaluate not just “will this work?” but “what could this become across different scenarios?”

• It builds genuine optionality. When Phase 2 data arrives, you’re not scrambling - you already have pre-evaluated paths, valuation ranges, and go/no-go criteria for each.

• It aligns with asymmetric learning: cheap exploration in early phases reveals which “If…” branch has the highest upside. Prune the weak ones fast; amplify the strong ones.

• In volatile areas like obesity or oncology, it mirrors how the market rewards platforms and multi-indication assets (see recent GLP-1 ecosystem dynamics and staged collaborations).

• It reduces the Terfenadine-style regret - where a single-TPP mindset might have killed a metabolite opportunity before it was even considered.

Challenges and Realistic Trade-offs:

• Resource intensity: Developing and testing multiple profiles requires more upfront cross-functional work (commercial intelligence, scenario modeling, patient input). Not every small biotech has the bandwidth - certainly something I personally think should be done by an expert…

• Decision paralysis risk: Too many options without clear prioritization frameworks can slow things down. Discipline is essential - use staged gates and real-options valuation to keep momentum.

• Internal resistance: Teams accustomed to “one target, one plan” may see it as inefficient or unfocused. Leadership buy-in and clear governance (e.g., a “possibility review” alongside the usual TPP review) are critical.

• Regulatory nuance: While adaptive designs and master protocols increasingly support this, you still need to lock certain elements before pivotal trials. Multiple TPPs shine most in discovery-to-Phase 2; they evolve into a focused profile later.

Overall evaluation: In today’s environment of high attrition, rapid competitor moves, and biology’s unpredictability, the multiple-TPP mindset delivers far superior risk-adjusted returns compared to the single-TPP default. It doesn’t eliminate failure - it makes failure informative and cheap while preserving (and compounding) upside. As I noted in “The Worst Number of Options in Pharma? Zero. The Second Worst? One,” having just one TPP is often worse than having none, because it creates false confidence and blinds you to better paths.

This is “If… Then What?” in action: start with the desired ends (plural), then back-engineer the conditions and experiments that keep those ends reachable.

From “Gradually, Then Suddenly” to Deliberate Design

We’ve seen how resistance to change often creeps in gradually, then suddenly becomes a crisis. Shifting to “If… Then What?” requires the opposite discipline: deliberate, early experimentation that feels inefficient in the short term but creates massive option value later.

It means treating New Product Planning as New Product Possibilities, using real options thinking as a mindset, and embracing planned serendipity.

The Personal Ellipsis

Readers who have followed the blog know I spent years building IDEA Pharma with a “Big Idea” ethos. Since leaving, I’ve found myself playing with a different framing - the minimalist power of If…

The ellipsis matters. It refuses to close the sentence too early. It leaves room for the data, for serendipity, for the asymmetric upside that no forecast can fully capture. (Yes, I’m old enough to have been trained in the ellipsis - it’s not just dots, always three, and often abused…)

Whether this evolves into a formal venture or remains a thought experiment, the underlying idea is the same: in an industry where most assets fail, the real edge comes from designing with all possible ends in mind, then engineering the conditions (“If…”) that let the best outcome emerge.

Practical Takeaways for Your Portfolio or Program

1. Audit your current TPPs. How many real “If…” branches exist? If there’s only one, you’re probably symmetrizing your risk rather than creating asymmetry.

2. Run a Multiple-TPP workshop. Map 3–5 plausible profiles early. Score them on feasibility, commercial upside, and required “If…” conditions. Include cross-functional voices and external market signals.

3. Measure optionality, not just probability of success. Success rate is symmetrical. Option value is where the asymmetry hides. Use staged investment and adaptive protocols to test branches cheaply. If your company is still rewarding phase transition rates, you’re already in the wrong mindset, and one that has killed innovation…

4. Ask the Terfenadine question. Where in your program is a current “liability” or side signal potentially hiding a better asset or indication? Keep the “If…” open long enough to find out.

5. Start with the desired Then(s), then back into the Ifs. This reverses the usual linear flow and forces more creative, end-first design.

The future of pharmaceutical innovation won’t belong to the companies with the best single prediction. It will belong to those who get comfortable living inside the ellipsis - who treat “If…” not as weakness or indecision, but as the disciplined practice of keeping valuable options alive.

If… we do that well, then what?

Then we learn faster, cheaper, and with far greater upside than the competition.

Then the market continues to vote in our favour.

Then we turn more of the 99% failure rate into asymmetric advantage.

Back in the 1980s, terfenadine (Seldane) launched as a breakthrough: a non-sedating antihistamine that quickly became a blockbuster for seasonal allergies. Patients loved it. Sales soared. Then, in the late 1990s, problems surfaced. When taken with common drugs like erythromycin or ketoconazole - or in patients with liver impairment - terfenadine could cause dangerous QT prolongation and fatal arrhythmias. The drug was withdrawn in many markets.

There is a clear asymmetric learning angle even in how the side effects themselves were uncovered. The dangerous QT prolongation and arrhythmias weren’t flagged by deliberate early DMPK experiments during discovery; they emerged reactively in the late 1980s and early 1990s through post-marketing pharmacovigilance. Case reports of syncope, torsades de pointes, and sudden deaths - especially in patients taking CYP3A4 inhibitors like erythromycin or ketoconazole, or those with liver impairment - triggered FDA signals and eventually a black-box warning. Only then did deeper mechanistic work reveal the truth: the parent terfenadine carried hERG liability that surfaced when metabolism was blocked and the compound accumulated. In classic symmetric fashion, the original single-TPP playbook had treated metabolism as a late checkbox rather than an active source of insight. Yet once the signal hit, the same biology that created the problem provided an escape hatch. Rigorous metabolite profiling turned the liability into optionality, isolating fexofenadine as the cleaner, equally effective active species. The side effects became the cheap (if unintended) experiment that revealed the higher-value branch, even if it took time on market to reveal the science. That pivot - from withdrawal risk to a new franchise - is asymmetric learning in its rawest form: biology’s surprise converted into upside by teams willing to listen rather than double down on the original plan.

But here’s what still fascinates me: the molecule wasn’t really the problem. Or rather, the parent molecule wasn’t delivering the full story/ hadn’t been asked for its secrets.

Terfenadine was being rapidly metabolized in the body - primarily by CYP3A4 - into its active metabolite, fexofenadine. That metabolite was providing allergy relief, without the cardiac liability. Once the industry isolated and developed fexofenadine (Allegra), they had a safer, equally effective drug that went on to its own commercial success.

One thoughtful oxidation of a tert-butyl group turned the liability into a significant commercial franchise. That’s not just serendipity, but asymmetric learning in action.

I’ve been reflecting on this story in light of our recent conversations here about how the market is increasingly voting for companies that learn asymmetrically - faster, cheaper, and with outsized upside when biology cooperates. The terfenadine → fexofenadine saga is a perfect case study from discovery and DMPK, one that echoes the themes we’ve explored around multiple TPPs, early-phase optionality, and why zero options (or just one) is the worst place to be.

The Symmetric Trap That Almost Killed the Franchise

In the old playbook, drug development was largely linear and predictive. You designed a molecule against a fixed target product profile (TPP): non-sedating H1 antagonist, good oral bioavailability, minimal CNS penetration. Metabolism? Often treated as a late-stage checkbox - something to characterize for regulatory filing rather than an active source of insight.

Terfenadine fit that mould beautifully on paper (now in the computer…). It blocked histamine effectively without making patients drowsy. But biology had other plans. The parent compound carried hERG channel liability that only became apparent under certain conditions (CYP3A4 inhibition leading to parent accumulation). The real winner - the metabolite - was always there, generated through routine first-pass metabolism.

This is the symmetric trap I’ve written about before. When you commit to a single TPP early, you symmetrize your learning. You march forward with the herd, optimizing for the same endpoints, the same patient populations, the same commoditized outcomes. If biology throws a curveball - an unexpected metabolite, an off-target effect, a better pathway - you’re left scrambling or, worse, withdrawing the asset.

As I noted in the piece on why zero is the worst number of options and one is only slightly better, a single-minded approach leaves you exposed. The winners aren’t the ones who predict the future perfectly (no one can). They’re the ones who set up small, cheap experiments that reveal asymmetric possibilities before the big Phase 3 bets.

Metabolism as a Classic Asymmetric Learning Engine

This is where DMPK shines as one of the purest forms of asymmetric learning in our industry.

Studying human metabolism early isn’t just de-risking - it’s option generation. Human hepatocyte assays, CYP panels, metabolite identification: these are relatively inexpensive probes that can do two things at once. They can kill a program quickly if the parent is irredeemably problematic. Or they can surface a completely different, higher-value asset - a safer metabolite, an active prodrug, or even a new indication path.

In the terfenadine case, deliberate metabolic profiling revealed that the “problem” molecule was actually a prodrug-like entity. The active species was downstream. Fexofenadine carried the efficacy without the cardiac risk. One CYP3A4-mediated change (oxidation of that tert-butyl group to a carboxylic acid) rewrote the story.

Contrast that with today’s tools. We now have far better predictive models for metabolites, high-throughput assays, and even AI-assisted structure prediction. The cost of this kind of learning has dropped dramatically, while the potential upside has grown. That’s exactly the learning-rate advantage the market is rewarding right now - as we saw with Gilead’s staged approach to Arcellx, Generate’s platform optionality, and the FDA’s moves toward more flexible, insight-driven development.

It reminds me of the GLP-1 story. Companies that designed with multiple ends in mind - not just diabetes, but obesity, cardiovascular benefits, and beyond - captured enormous asymmetric value. Metabolism work is the discovery equivalent: you’re deliberately building possibility trees that include branches you can’t fully predict upfront.

(Quick etymology note, because I can’t resist: “metabolite” comes from the Greek metabolē, meaning “change” or “transformation.” Literally the change that can transform your asset’s future.)

Designing Discovery with Multiple Ends - Including the Metabolite Ones

So what does this mean practically, in 2026?

First, treat metabolism not as a late liability screen but as an early insight engine. In hit-to-lead and lead optimization, design scaffolds with metabolic awareness in mind. Explore intentional prodrug strategies. Map metabolite profiles in parallel with primary activity, not sequentially.

Second, build real possibility trees from the start. Instead of one TPP, sketch 3–5 branching scenarios: What if the parent is the winner? What if a major metabolite drives efficacy? What if an unexpected clearance pathway opens a new dosing or patient-segment opportunity? Each small experiment in DMPK becomes a cheap bet that can compound into major optionality.

Third, reframe “surprises” as signals. An unexpected metabolite or drug-drug interaction isn’t automatically a program killer - it might be the market’s way of telling you where the real value lies. The difference between terfenadine’s withdrawal and Allegra’s success was someone paying close attention to those metabolic branches.

We’ve seen this pattern elsewhere: sildenafil’s pivot from angina to erectile dysfunction, or the way certain kinase inhibitors revealed broader utilities through off-target (or metabolite) profiling. In each case, the asymmetric learners were the ones who treated biology’s wild cards as opportunities rather than deviations from plan.

With today’s regulatory flexibility - single pivotal trials plus confirmatory evidence where the data supports it - the payoff for this mindset is even larger. Companies that generate proprietary, uneven insights early can move faster, pivot smarter, and capture value before competitors symmetrize around the obvious path.

The Next Blockbuster May Already Be Inside a Molecule You’re Tempted to Shelve

Allergy season will come and go, but the lesson from terfenadine and fexofenadine stays because it captures the essence of what I keep coming back to in these posts: pharmaceutical innovation rewards those who design processes for asymmetric learning.

The market is already voting with its capital - favoring teams that embrace optionality, stage learning deliberately, and let winners compound unevenly. Metabolism is one of the oldest, most accessible ways to do exactly that in discovery.

The next blockbuster may already be sitting inside a molecule your team is characterizing right now. It might be the parent. It might be a downstream metabolite. Or it might be something the data hasn’t revealed yet because you haven’t asked the right questions early enough.

The difference between a withdrawal notice and a new franchise is often just one thoughtful oxidation - or one deliberate experiment - away.