"No market need" is an explanation you apply after a company has failed. It is useless on the day you need to decide, because it does not tell you which commitment was visible beforehand. Here are 5 companies whose stories are finished, the one-liner each could have been described by at founding, and the property 3 of them shared — one you can see on day one.
The five, as they would have looked on day one
| Company | How it ended | The shape, as we read it |
|---|---|---|
| Quibi | shut down ~6 months after launch, ~$1.75B raised | Money out before the answer was in |
| Juicero | shut down in 2017 (packs were squeezable by hand) | Money out before the answer was in |
| Google Glass (consumer) | withdrawn as a consumer product ~2015 (privacy backlash, no use-case) | A capability looking for a complaint |
| Webvan | bankrupt in 2001, ~$800M+ raised | Money out before the answer was in |
| single-platform-API research tools | platform-risk wipeout when the platform's API access/pricing changed (2023–2025) | A business made of someone else's API |
Outcomes and sources are data, imported from the benchmark. The third column is OURS — a human reading of the five one-liners, not a verdict any model produced. That distinction is the point of the last section.
The shape: the money went out before the answer came in
3 of the 5 share one property, and it is visible in the founding one-liner rather than in the post-mortem.
Studio-produced episodes had to be commissioned before anyone could discover whether people wanted seven-minute premium video on a phone. A countertop press and its proprietary produce packs had to be manufactured before anyone found out the packs squeezed fine by hand. Automated warehouses and a delivery fleet had to be built before the grocery demand at that price was known. In each case the expensive commitment came first and the answer came second, and by the time the answer arrived it could not change anything.
That is a different failure from "no market need". Plenty of businesses discover there is no market need and stop, having spent a landing page and three weeks. What killed these was the ordering: the test could not run until after the money was gone.
The fourth has the opposite shape. Always-on glasses with a camera were a capability in search of a complaint — nobody had written down the problem it solved, and the objection it eventually met (other people did not want to be filmed) was not a market-size question at all.
The one nobody killed
A SaaS that mines one large social platform's data through its API to surface customer pain points and demand signals for founders.
The row our own gate did not kill on either profile. It stays in the table for the same reason the misses stay in the benchmark: a set of cases with the failures removed is a testimonial.
Both lenses held it for evidence and both surfaced platform dependency as the thing to go and validate — which is the correct instinct and not a kill. Then the platform changed its API pricing and the category was wiped out.
Reasonable people can read that two ways. Either the gate was right to hold and the world moved, or a business whose entire supply is one company's API should never have cleared a check. We lean toward the second, which is why platform risk is the one filter we later softened rather than removed — and why the softening is narrow enough that this idea would still not have survived it: platform dependency plus anything else stays dead.
Why we will not tell you which check would have caught each one
This is where a page like this usually overreaches, so here is the line.
Our gate caught 4 of the 5 — that is a fact about verdicts, and it is in the benchmark with every row visible. What we will not do is claim the reason it gave for each kill was the reason the company actually died. An adversarial audit of our own output found the cited reason is sometimes adjacent to the real one. Catch is a claim we can support; diagnosis is not.
So the third column of that first table is labelled as our reading, and the honest use of this page is not "the checklist would have saved these companies". It is narrower and more useful: before you commit money that a test could have spent instead, notice whether your answer can arrive first.
What to take from it
- Do not spend your way to the question. If the expensive thing has to exist before anyone can tell you whether they want it, the ordering is the risk, whatever the market size says.
- A capability is not a complaint. If you cannot find someone describing the problem in their own words, you have a technology, not a business.
- One company's API is not a supply chain. It is a business decision made by people who do not know you exist.
- 4 of 5 caught, with the miss published. Treat any source that only shows you its hits — including a page about failures — as marketing.