Berkus Method Weaknesses: Why Five Fixed Categories Miss Deep Tech and Biotech Startups

The Berkus Method values a pre-revenue startup by giving up to $500,000 to each of five risk-reducing milestones: a sound idea, a prototype, a quality team, strategic relationships, and product rollout. The result is capped at roughly $2 million to $2.5 million. The main Berkus Method weaknesses are its fixed dollar caps, equal category weights and software-era categories. Together these understate the scientific, regulatory and capital risk that decides whether a deep tech or biotech startup survives.

How the Berkus Method Works and Why Angels Still Use It

Dave Berkus built the method in the mid-1990s to value early-stage technology companies. It reached a wider audience in 2001, when Harvard's Amis and Stevenson published it in Winning Angels. In his 2016 update on the Angel Capital Association blog, Berkus explained the reasoning. Fewer than one in a thousand startups hit their projected revenues on plan, so forecasts are a poor basis for value. Instead, he gives credit for the progress that reduces risk.

The standard matrix, as MicroVentures and most calculators present it, looks like this:

Sound idea (basic value, product risk): up to $500,000
Prototype (technology risk): up to $500,000
Quality management team (execution risk): up to $500,000
Strategic relationships (market risk): up to $500,000
Product rollout or sales (production risk): up to $500,000

Berkus's 2016 version allows a pre-revenue valuation of up to $2 million, or up to $2.5 million after rollout. He also says that once a company has revenue, the method no longer applies. Angels like it because it is quick. You can score a seed deal in ten minutes on a spreadsheet, and the founder can follow the logic.

Takeaway. Use the standard Berkus matrix as a first screen for capital-light software and consumer deals at the $1 million to $2.5 million pre-money range. For anything that has to get past a lab bench, a regulator or a pilot plant, treat the output as a floor to be adjusted.

Berkus Method Weaknesses: Five Structural Gaps for Deep Tech and Biotech

Berkus himself wrote in 2016 that the matrix "should be a suggestion rather than a rigid form". He gave the example of medical device companies swapping in FDA approval risk for marketing risk. Most practitioners skip that caveat and apply the five categories as written. For science-driven companies, that creates five specific problems.

1. The dollar caps ignore how much capital the company needs

A $2.5 million ceiling makes sense when a $500,000 angel round can carry a team to product-market fit. It stops making sense when the next milestone costs nine figures. According to BCG's March 2021 deep tech report, the average disclosed private investment in deep tech startups and scale-ups rose from $13 million in 2016 to $44 million in 2020. If a company needs $40 million for its Series A, a $2 million seed pre-money sets up a cap table that later investors will have to restructure.

2. Equal weights spread risk evenly when it is concentrated

In software, the five risks are roughly similar in size. In biotech, a single risk dominates: whether the molecule works in humans. The Berkus matrix gives "sound idea" and "strategic relationships" the same $500,000 as the clinical data that decides everything else.

3. "Prototype" means something different in each sector

A SaaS prototype is a clickable demo. A biotech "prototype" might be in vitro data, a mouse efficacy study or an IND-enabling tox package, and each carries a very different probability of success. A fusion "prototype" is a demonstration machine that costs hundreds of millions of dollars. One binary box cannot hold all of these stages.

4. There is no category for regulatory or scientific risk

The five categories cover product, technology, execution, market and production risk. None of them covers regulatory risk (FDA, EMA, NRC licensing) or basic-science risk (does the physics or biology hold at scale). For drug developers, regulatory risk is often the largest single risk before revenue.

5. Time is missing entirely

Berkus scores the progress a company has made. It has no way to discount for how far away value creation is. A biotech ten years from launch and a marketplace app six months from launch can reach the same $2 million score.

Takeaway. Before you score a deep tech deal, write down its dominant risk, how many years it is from revenue and the capital it needs to reach the next value inflection. If any of those looks nothing like a typical software seed deal, use a modified matrix.

The Numbers Behind the Mismatch: Biotech and Fusion Examples

Biotech: falling approval odds

BIO's Clinical Development Success Rates 2011-2020 report, produced with Informa Pharma Intelligence and QLS Advisors, is the standard benchmark for how often drugs move from Phase I to approval. Writing for pharmaphorum in July 2024, Norstella's Daniel Chancellor tracked the likelihood of approval from Phase I across successive updates. It fell from 10.4% in the 2014 study to 9.6% in 2016, 7.9% in 2021 and 6.7% in 2024.

The spread by disease area matters even more for valuation. The same analysis puts haematology at 19.1% and oncology at 4.7%. Under standard Berkus scoring, a haematology asset and an oncology asset at the same stage, with similar teams, get the same value. On a probability-weighted basis, the haematology asset is about four times more likely to reach approval.

Cost and timeline make the gap wider. A peer-reviewed benchmarking study in PMC estimates that going from target validation to launch takes 10 to 15 years, at an average of $1 billion to $2 billion per successful drug. Berkus's $500,000 for a "prototype" is a rounding error against that spend, and a very large number against a 6.7% probability.

Deep tech: Commonwealth Fusion Systems

Commonwealth Fusion Systems (CFS) shows the capital problem clearly. In 2025 the company announced an $863 million Series B2 to finish SPARC, its fusion demonstration machine, and advance its first ARC power plant in Virginia. Investors included NVentures, Nvidia's venture arm. CFS said it was the largest deep tech and energy raise since its own $1.8 billion Series B in 2021. It brought total funding to close to $3 billion, about one-third of all private capital invested in fusion worldwide.

A Berkus scorer looking at CFS in its early days would have given full marks for team (MIT Plasma Science and Fusion Center pedigree) and partial marks for idea and relationships. It would have given zero for prototype and rollout, for a total around $1.5 million. The metric that actually drove value was whether the company's high-temperature superconducting magnet would reach its target field strength, and that single technical milestone has no line in the five-category matrix.

Takeaway. For every deep tech or biotech deal, find the sector base rate (BIO/Norstella likelihood of approval for drugs, technology readiness level for hardware) and the full capital needed to reach revenue. Put both on the first tab of your valuation file before any Berkus score.

How to Adapt the Berkus Method Step by Step for Deep Tech and Biotech

You do not have to discard the method. Berkus himself allowed for category substitution. The fix is to rebuild the matrix around the risks that actually apply. Here is the process we use in a spreadsheet model.

Set the cap from the round. Replace the fixed $2.5 million ceiling with a sector-appropriate maximum pre-money. Use recent comparable seed rounds in the same modality or hardware category. If comparable preclinical seed rounds price at $8 million to $15 million, use that as your ceiling.
Swap the categories. For biotech, use scientific validation (in vivo efficacy), IP position (composition-of-matter patents, freedom to operate), regulatory path (orphan designation, IND readiness), team, and strategic partner (a pharma option or co-development deal). For hardware deep tech, swap regulatory path for technology readiness level and add manufacturing scale-up.
Weight by dominant risk. Give 30% to 40% of the cap to the category that decides survival. In a preclinical biotech that is usually scientific validation. Spread what is left across the other four.
Score each category on a graded scale. Use 0%, 25%, 50%, 75% or 100% of the category cap, and tie each level to a documented milestone. For example, 50% for scientific validation means reproducible efficacy in one animal model, and 100% means two models plus a clean tox signal.
Apply a probability and time check. Multiply the expected exit value by the stage-appropriate likelihood of approval (for example, the 6.7% Phase I-to-approval figure, or the indication-specific rate). Discount back over the years to exit at a venture target return. If the adjusted Berkus score is more than 1.5 times this cross-check, explain the gap in writing.
Stress-test the cap table. Model dilution through the funding rounds needed to reach the first value inflection. If the founders end up below 10% before Phase II or a pilot plant, the pre-money is too low to keep the team motivated, whatever the score says.

Here is an illustrative example. A preclinical rare-disease company with a $12 million cap, 35% weight on scientific validation scored at 75%, 20% on IP at 100%, 15% on regulatory path at 50%, 20% on team at 75% and 10% on partnerships at 0% comes out at about $8.55 million pre-money. That figure would then be checked against the probability-weighted exit model in step 5.

Takeaway. Write the milestone definitions for each score level before you meet the founder. Graded definitions agreed in advance stop the scoring from drifting toward the founder's pitch.

Fixing Berkus Method Weaknesses With a Hybrid Excel Template

Few experienced deep tech investors rely on a single method. Allied Venture Partners' comparison of pre-revenue models sets Berkus alongside the Scorecard method, the VC method and risk-factor summation, and most angel groups triangulate across them. For science-heavy companies, a working hybrid has three tabs:

Modified Berkus tab: sector-specific categories, a cap-driven ceiling and weighted scoring, as set out in the step-by-step section above.
Risk-adjusted NPV tab: phase-by-phase probabilities from the BIO benchmark, development costs and a discount rate, which gives a probability-weighted value.
VC method and dilution tab: the exit value divided by the target multiple, adjusted for projected dilution across every round needed before revenue.

The summary sheet shows all three values side by side, with a weighted blend. In practice, deep tech and biotech deals should put the most weight on risk-adjusted NPV and the least on the Berkus score, because the Berkus categories carry the least information about those companies.

Takeaway. If your current valuation file has only one tab, add the risk-adjusted NPV tab first. It covers the time and probability gaps that the five fixed categories leave open.

Conclusion: Use Berkus as a Floor for Science Startups

The Berkus Method still works for what it was built for. It suits capital-light technology companies raising small rounds. Its weaknesses show up when a company's value depends on a clinical readout, a regulatory decision or a machine like SPARC that costs hundreds of millions to build. Fixed $500,000 caps, equal weights and missing regulatory and time factors will systematically misprice those deals, and oncology assets with a 4.7% approval rate will get the same score as haematology assets at 19.1%.

The fix is a documented process. It uses sector caps, substituted categories, weighted scoring, a probability cross-check and a dilution test. Building that from scratch takes a full day and is easy to get wrong. A ready-made Excel template with the modified Berkus matrix, a risk-adjusted NPV tab and a VC-method dilution tab already linked lets you score a deal in under an hour and defend the number to your investment committee or your co-investors. Start with the template, enter your sector's base rates and capital needs, and adjust the weights to the dominant risk.