Evidence base

409 startup failures: the study behind Verdika

By the Verdika team · Jul 18, 2026 · 6 min read

In short

We analyzed 409 startup post-mortems documented by CB Insights and turned them into decision patterns Verdika uses as context when weighing alternatives. Three out of four (75.3%) died competing against a market giant. This page documents the numbers, the methodology and (just as important) what this base does not cover.

Verdika promises defensible decision memos. A recommendation can only be defended if you know where its references come from, so we publish the full base: what was analyzed, which patterns emerged, and under which limits it gets used.

The numbers

Each case documents one or more failure causes (categories are not mutually exclusive, which is why they sum past 100%):

Competing against a giant

75.3%

Acquisition stagnation

22.2%

Poor product-market fit

19.8%

Cash mismanagement

18.8%

Niche too small

16.9%

Execution flaws

14.7%

Monetization failure

13.9%

The decision patterns

More useful than the categories are the concrete decisions that repeat across post-mortems: scaling before product-market fit is confirmed, or choosing the largest market over the most defensible one, show up again and again. From the 409 cases we extracted a structured set of failure and success patterns (which decision, in which context, with which signals) that is the proprietary part of Verdika's engine. Here we publish the categories, the numbers and the methodology; the full operational pattern set is what the engine applies to your case when you analyze a decision.

Methodology

The primary source is the startup post-mortem collection compiled and categorized by CB Insights¹ across 6 sectors (finance, health, retail and others). Our work was extracting the decision patterns (not just the declared causes) from those 409 cases, and structuring them so an analysis engine can use them as context when generating and evaluating alternatives.

How Verdika uses it, exactly: the patterns enter as narrative context, never as ground truth. Your evidence always weighs more. On top of that, each memo's declared confidence has a mathematical cap based on the quality of your real evidence: with no objective evidence, it never exceeds 40%. The full methodology is published in the methodology section.

What this base does NOT cover

The 409 cases are startup-level strategic failures. They don't cover decisions about specific people (who to promote or let go) or third-party agreements (signing a partnership, a vendor contract). When Verdika detects a decision from those domains, it excludes this base from the analysis and says so in the memo: the recommendation rests on your evidence alone. We prefer a smaller, honest base over a stretched one. For sector base rates, Verdika complements with live data from the U.S. Census (Business Dynamics Statistics), a verifiable government source.

See the patterns applied to your decision.

Describe your problem and get a memo with ranked options, weighted evidence and margin of error.

Analyze my decision

References

CB Insights — The Top Reasons Startups Fail (analysis of startup post-mortems): cbinsights.com.

You might also find useful: Hire now or stretch the runway · How to write a decision memo