Why 59% of Startup Ideas Should Be Killed

What a KILL verdict actually means

A KILL is not a judgment on the founder. It is a judgment on the specific configuration of problem, customer, solution, and business model described in the submission brief. Several founders who received KILL verdicts have returned with restructured ideas — different segments, different monetization models — and received CONDITIONAL GO or GO verdicts.

What makes a KILL different from a PIVOT is the nature of the failure. PIVOT means the core problem is real but the solution path does not work. KILL means the problem itself — as defined — is not painful enough, frequent enough, or monetizable enough to support a business.

How PRISM verdicts distribute across 214 ideas

Across all PRISM submissions to date, verdicts break down as follows. The numbers are not aspirational targets — they reflect the actual distribution of ideas as founders describe them before validation, without survivorship bias from investor filtering.

Only 14% of ideas earn a GO or CONDITIONAL GO verdict. That figure is not evidence that PRISM is too strict — it reflects the actual quality of unvalidated ideas entering the market. CB Insights data shows 42% of funded startups fail due to product-market fit failure. Those ideas had already passed investor diligence. Before diligence, the real failure rate for raw startup ideas is expected to be higher.

For a full breakdown of verdict distribution by sector and validation lens, see the PRISM verdict distribution dataset.

Which validation lenses trigger KILL verdicts most often

PRISM evaluates each idea across 12 validation lenses. The six lenses below account for the majority of KILL triggers. Failure on any single lens does not automatically produce a KILL — but failure on the top two (Market Demand and Unit Economics) together produces a KILL in over 90% of cases.

Market Demand (82%) is the dominant kill trigger. The evidence does not support that the described customer experiences the described problem with sufficient pain, frequency, or willingness to pay. The most common failure mode here is founders citing broad market statistics ("the market is $80bn") without evidence of specific customer behaviour — search volume, support tickets, waitlists, or structured interviews.

Unit Economics (73%) is the second most common trigger. The numbers don't work: customer acquisition cost is too high relative to lifetime value, or the pricing model cannot reach the revenue required to cover costs at any realistic scale.

KILL vs PIVOT: what separates them

The most common question founders have after receiving a KILL verdict is whether it should have been a PIVOT. The distinction matters because a PIVOT verdict means the problem is salvageable — a KILL verdict means the configuration cannot be fixed without starting over on a different problem.

When founders challenge a KILL verdict in the formal rebuttal round (available on Pro and Ultra tiers), approximately 7% of contested verdicts are upgraded to PIVOT or CONDITIONAL GO on first rebuttal. This means 93% of KILL verdicts are confirmed after the founder's strongest counter-evidence is reviewed.

The four KILL patterns

Across 126 KILL verdicts, four structural patterns account for the majority of failures. Recognising which pattern your idea falls into before submitting can help you restructure it — or save you the cost of building it.

A five-point check before you submit

Based on the most common failure modes in our KILL verdict dataset, these are the five questions worth answering before submitting your idea to PRISM — or before building anything at all. Founders who can answer all five with evidence (not assumptions) are significantly less likely to receive a KILL verdict.

1. Do you have evidence of purchase behaviour, not just stated interest? Interviews where people say they would pay are not evidence. Waitlist sign-ups with a deposit, pre-orders, or Letters of Intent are. If you only have interview quotes, the Market Demand lens will likely fail.
2. Can your unit economics work at the price your target customer has already paid for adjacent solutions? Not at a theoretical price. At the price they have actually paid, to a competitor or substitute, in the last 12 months. If you don't know this number, find it before submitting.
3. Can you identify one distribution channel that reaches your first 100 customers without paid advertising? If your go-to-market plan starts with "we'll run Facebook ads," the Distribution lens will likely fail. Identify the specific community, partner, or channel where your target customer already congregates.
4. Is there at least one incumbent in your space that is growing slowly or not at all? A market with no incumbents is usually not a market — it's an assumption that the problem exists. A market where incumbents are stagnant or losing revenue signals that the problem is real and the current solutions are inadequate.
5. Can you describe your first customer — specifically, not as a persona? Not "SMBs in the UK" but "operations managers at UK logistics companies with 20–80 staff who currently use spreadsheets for fleet scheduling." If you cannot name a company type, role, and current tool, the submission is too abstract for PRISM to evaluate accurately.

Founders who complete structured pre-validation before submitting to PRISM fail at approximately 31% — compared to the 59% baseline for unvalidated submissions. That gap represents the difference between building with evidence and building with assumptions.

Frequently asked questions