Risk-adjusted NPV: how to value pre-revenue biotech

September 11, 2026 · 5 min · Free, with full transcript

How do you put a defensible number on a drug candidate that has never sold a dollar and may never reach a patient? The answer is risk-adjusted NPV, which haircuts projected peak sales by the cumulative probability of success at each clinical stage. This episode walks through the mechanics and argues the math is sound while the inputs are where it quietly fails. You come away knowing the BIO likelihood-of-approval benchmarks, roughly one in ten from phase one and closer to 5% in oncology, why 15 to 20% discount rates beat the standard 10 to 12%, how Alzheimer's programs fell 80% on readout, and why vendor defaults from Alacrita or Evaluate are averages your drug is not.

Listen to the podcast

5 min

Chapters

Key takeaways

  • Never accept a single rNPV figure; demand a two-way sensitivity table of probability of success against discount rate.
  • Use 15 to 20% or higher as the discount rate for pre-revenue biotech, not the textbook 10 to 12%.
  • Check the probability of success against BIO likelihood-of-approval data by indication, around 10% from phase one and near 5% in oncology.
  • Treat vendor benchmarks from tools like Alacrita or Evaluate as industry averages, and adjust down for first-in-class mechanisms nobody has validated.
  • Challenge any peak sales line that assumes large market share against established competitors, since the model will accept it without complaint.
Read the full transcript

Host:Welcome to the Leaders Insights Podcast. Today's episode, Risk-Adjusted NPV for Pre-Revenue Biotech Pipelines. How the math actually works. A biotech with zero revenue, zero approved products, and a market cap north of a billion dollars. Explain to me how anyone puts a number on that without laughing.

Expert:They use Risk-Adjusted NPV. Reppriced present value, where you haircut every future cash flow by the probability the drug actually survives each clinical stage. And the reason it's on everyone's mind right now is that biotech valuations got repriced hard this year. When money isn't free anymore, the model's assumptions get brutal.

Host:Start with the mechanics. What is the model actually doing when you feed it a drug that might never reach a patient?

Expert:It projects peak sales. Say the drug launches in 2032 and hits a billion dollars a year, then discounts those cash flows back to today. But before discounting, you multiply by the cumulative probability of success. A candidate entering phase one has maybe a 10% chance of ever reaching market, industry-wide. So that billion-dollar drug is worth a fraction of that on paper.

Host:10%? So 90% of the time you're valuing a corpse?

Expert:Exactly. And that's the honesty of the model. The bio-industry data—that's the Biotechnology Innovation Organization, the trade group—pegs the overall likelihood of approval from phase one at roughly one in ten across all diseases. Oncology is worse, closer to 5%. You're pricing a lottery ticket, and RNPV forces you to say the odds out loud.

Host:Fine, but every input you just named—peak sales, probability, launch date—is a guess. Isn't this just a spreadsheet dressed up as science?

Expert:That's the sharp part. The math is fine. The inputs are where it quietly fails. Move the probability of success from 10% to 15%, and you can double the valuation. Move peak sales from a billion to a billion-five, and you move it again. Two analysts using the same model can be off by 300% and both be right.

Host:So where does it break worst?

Expert:The discount rate. Critics tell you to discount at 10-12%. For pre-revenue biotech, that's a fantasy. These companies are burning cash with no floor under them. Serious investors use 15-20% or more. Every point you add crushes. The value of cash flows a decade out. I've seen a program worth $400 million at a 12% rate drop to under $150 million at 18%.

Host:Give me a real one. A company where the market and the model disagreed and someone got burned.

Expert:Think about any Alzheimer's play over the last few years. The probability of success in neurology is dismal, well under 10% historically, hehe, yet companies kept getting valued as if approval was likely because the market is enormous. When trials read out negative, the stock didn't fall 20%. It fell 80% because the RNPV was built on a probability nobody believed but everybody used.

Host:If the model didn't fail, the people using it lied to themselves.

Expert:They anchored on the upside and soft-pedaled the probability. That's human, not mathematical. The other quiet failure is the egosate and then a miracle occurs peak sales line. People model a drug capturing 40% of a market with three competitors already there. Our NPV will happily let you.

Host:What about the software everyone sells to run these models? Can I trust the outputs?

Expert:Tools like Alacridos or the Evaluate franchise give you clean benchmarks for probability of success by stage and indication, genuinely useful. But remember, those defaults are vendor-supplied averages. Your drug isn't average. Plug in the industry number for a first-in-class mechanism nobody's validated. And you're being optimistic, not conservative.

Host:So the number that comes out the other end is basically a mood.

Expert:It's a disciplined mood. The value isn't the single figure. It's the sensitivity table. Show me the valuation across a range of probabilities and discount rates, and I'll tell you whether the story holds up or whether it lives and dies on one heroic assumption.

Host:Give me the one thing a listener does tomorrow.

Expert:Never accept a single RNPV number. Demand the two-way sensitivity table. Gui Ti Yutong, probability of success against discount rate. If the whole thesis only works at 12% and a 15% probability, you're not looking at evaluation. You're looking at someone's hope with decimal points.

Host:Sources for today's episode. Accounting Today, Financial Times, CFO Dive, Accountancy Age. Done. Want to know where you actually stand? Take the CFO self-assessment at MBA-training.com.

Read and go deeper