1. What Is Founder Risk?
Every deal memo has a slide for market size, a slide for the product, and a slide for unit economics. Founder risk is what's missing from that memo: the probability that the venture fails for reasons that live inside the founding team, not outside it — a capability gap nobody flagged, a co-founder split nobody saw coming, a conflict style that seizes up the moment the roadmap changes. It's a distinct category of risk, separate from "is this market real" or "does this product work," and it deserves its own line of diligence rather than a gut check during the reference call.
In practice, founder risk breaks into a handful of measurable dimensions: individual capability and resilience, leadership and decision-making style, and — for teams of two or more — how compatible those individual profiles are with each other under pressure. The rest of this guide walks through why that risk is large enough to matter, how it's assessed today, and what a structured, evidence-based alternative looks like.
2. Why It Matters
The scale of the problem is well documented, just quietly. Research by Harvard Business School's Shikhar Ghosh, based on more than 2,000 venture-backed companies, found that roughly 75% of VC-backed startups never return investor capital — a figure widely reported after Ghosh told the Wall Street Journal that VCs "bury their dead very quietly." Independent data from Correlation Ventures' analysis of tens of thousands of VC financings lands in the same range: about 65% fail to return even 1× capital.
Diligence on the market and the product is standard practice across the industry — TAM decks, cap-table modeling, technical architecture reviews. Diligence on the founder and the team is comparatively under-instrumented, despite being implicated in a large share of the outcomes above. That asymmetry is the gap this guide is about.
3. Team & Co-Founder Conflict
Co-founder and team issues sit among the top causes of startup failure. Noam Wasserman's research in The Founder's Dilemmas — drawn from data on nearly ten thousand founders — finds that roughly 65% of high-potential startups experience serious co-founder conflict, and frames "people problems" as a leading cause of startup failure.
That's not the same as calling it the single leading cause. CB Insights' analysis of VC-backed shutdowns puts "ran out of cash" at the top of the list — but by its own account, that's almost always the final, proximate cause, not the root one. Team conflict is frequently what leads there: a founding team that can't resolve disagreement burns runway on the disagreement instead of the business. Both readings are true at once, which is why founder-team risk earns its own line item instead of a footnote under "team slide looked fine."
4. The Traditional Approach
Most early-stage venture diligence on founders is unstructured: partner interviews, reference calls, and pattern-matching against founders a fund has backed before. It's fast, requires no new tooling, and every partner already knows how to do it — which is exactly why it persists. The cost shows up elsewhere: it's hard to compare consistently across deals, hard to defend to an LP asking "why this founder," and vulnerable to affinity bias — the well-documented tendency to rate founders who resemble the interviewer more favorably, independent of merit.
Executive search and leadership-consulting firms solved a version of this problem decades ago for C-suite hiring, using structured psychometric assessment instead of interview intuition alone. That category of instrument is common at the executive-search level and uncommon in early-stage venture diligence, where speed has historically outweighed structure — largely because nobody had built a version fast enough for a VC's timeline.
5. The Psychometric Assessment
The first layer of an evidence-based approach is a structured psychometric instrument — a short, standardized questionnaire each founder completes, not an interview and not a clinical evaluation. 1% Better runs this as a focused, non-invasive scan of founder psychology, calibrated using the IPM AG assessment methodology, a licensed personnel-diagnostics instrument, interpreted through a psychometric research lens the founding team — trained at ETH Zurich and the University of St. Gallen (HSG) — built specifically for founders rather than corporate hires.
Two things separate it from a corporate hiring test. First, what it measures: achievement motivation, leadership orientation, conflict style, and stress response — the traits that predict how someone behaves when a pivot is forced or a co-founder pushes back, not whether they'd be compliant in a performance review. Second, for teams of two or more, it doesn't stop at individual profiles. It builds what we call a compatibility topology: a map of how each founder's conflict style, decision velocity, and stress-response pattern interacts with their co-founders', pinpointing where task conflict turns into relationship conflict before it erodes team efficacy.
A questionnaire, on its own, is only half the job — it produces clean, comparable data, but data still needs interpretation. That's where the second layer comes in.
6. The AI Assessment
1% Better's AI layer — internally called DIALECTIC — is a pipeline of 24 specialist AI agents, not one model asked for an opinion. They run across three phases:
- Research (11 agents): gather and classify public evidence on the founders, the team, and the market context — the raw material everything downstream reasons over.
- Psychometric interpretation (2 agents): a Founder Profiler and a Team Dynamicist read the IPM AG instrument's output and translate it into risk-relevant findings.
- Debate and synthesis (11 agents): agents argue the bull case and the bear case for each domain, a dedicated Chaos agent stress-tests the optimistic read, and a synthesis stage — including a document-consistency auditor — cross-checks every argument against the evidence before producing one verdict.
The point of the debate stage isn't theater. A single model asked "is this team a good bet" will anchor on whatever framing it's given first. Splitting the reasoning across agents that are explicitly assigned to argue for and against a founder — and a separate agent whose only job is to break the optimistic case — forces the disagreements onto the page instead of hiding inside one model's weights. It's closer to a simulated investment committee than a chatbot answer, and it's documented in the DIALECTIC whitepaper.
It's also not a black box. Every finding in the output traces back to a specific agent, a specific piece of evidence, or a specific psychometric scale — an investor can follow the reasoning chain, not just read a score. The target is a delivered memo within 48 hours of intake; that's a service-level commitment we designed the pipeline around, not yet a measured average across live pilots.
7. Why Pairing Both Works Better
Neither layer does much alone. A psychometric instrument without interpretation is a spreadsheet of scores nobody acts on. An AI system without structured data to reason over is just a model guessing from a pitch deck — articulate, but no more grounded than the gut feel it's replacing. Paired, the instrument supplies the measurement and the AI layer supplies the scrutiny:
| Dimension | Unstructured interviews | Psychometrics alone | Psychometrics + AI debate |
|---|---|---|---|
| Comparable across deals | No — every partner judges differently | Yes — standardized scales | Yes — standardized scales, argued consistently |
| Bias exposure | High — affinity bias, first-impression anchoring | Low on the instrument itself | Low — adversarial agents argue both sides |
| Explainable to an LP | "I have a good feeling about them" | Raw scores, limited narrative | Source-linked findings with a documented reasoning chain |
| Who interprets the data | — | Left to the reader | 11 synthesis-stage agents, cross-checked |
The result is a documented, source-linked assessment instead of a single unstructured judgment call — something you can hand an LP, not just a feeling you can describe to one.
Swiss data residency: AWS Zurich only
(eu-central-2); application and all data at rest in
Switzerland; only model inference leaves Switzerland and it stays
within Europe (EU/EEA).
8. FAQ
What is founder risk?
Founder risk is the probability that a venture fails for reasons rooted in the founder or founding team — capability gaps, co-founder conflict, or team dynamics — as distinct from market, product, or macroeconomic risk.
Is co-founder conflict the number one cause of startup failure?
No. Co-founder and team issues are among the top causes of startup failure, not the single leading one — running out of cash is more commonly cited as the final, proximate cause. Team issues are frequently what leads there.
How do VCs traditionally assess founder risk?
Mostly through unstructured partner interviews and reference calls — fast, but hard to compare across deals and vulnerable to affinity bias. Larger funds and executive search firms sometimes add structured psychometric instruments, but these are uncommon in early-stage venture diligence.
What does a founder psychometric assessment actually measure, and is it invasive?
It's a short, structured questionnaire — not an interview, not a clinical evaluation. It profiles traits like achievement motivation, leadership orientation, conflict style, and stress response, then maps how co-founders' profiles interact under pressure. It measures working style, not mental health, and takes about 15 minutes per founder.
How does the AI assessment work — is it just one model's opinion?
No. It's a pipeline of specialist AI agents split across research, debate, and synthesis. Different agents argue the bull case and the bear case, a dedicated agent stress-tests the optimistic read, and a synthesis stage cross-checks the arguments against the evidence before producing one documented verdict — closer to a simulated investment committee than a single model's output.
Is this an AI black box?
No. Every claim in the output traces back to a specific agent, a specific piece of evidence, or a specific psychometric scale — investors can inspect the reasoning chain, not just the final score.
What does pairing a psychometric instrument with AI debate add over either alone?
A psychometric instrument alone gives you structured, comparable data but no interpretation. An AI system alone can reason over evidence but has nothing objective to reason about. Paired, the instrument supplies the measurement and the AI system supplies the adversarial scrutiny and synthesis — producing a documented, source-linked assessment instead of a single unstructured judgment call.
Request pilot access to see the full assessment on a real deal.