Startup ideas do not usually die of the thing you are worried about. On the one Discovery run we publish in full, 36 of 69 ideas were dropped, and the most common cause of death was not a crowded market, a competitor, or the price being wrong. It was that there was no documented problem behind the idea at all — 21 of them, more than twice the next cause.
What the run threw away
The run is a single Discovery sweep, published unedited at a real, unedited run: 69 candidate ideas sourced from 41 sources, 24 of which were read in full. Every rejection is on that page with its reason attached, which is the only reason this article can exist — you cannot itemise deaths you did not record.
Here is the whole census, counted from the run rather than recalled:
Every check that killed at least one idea on the published run, most-cited first. Bars are relative to the largest. Drawn from the same function the run's own page uses, so the chart cannot disagree with the page it describes.
The shape matters more than any single row. Three checks account for most of the damage and the remaining six between them account for very little. Whatever is going wrong is going wrong early, in the same place, over and over.
The rows do not add up, and that is the point
Add the bars together and you get 57 reasons against 36 dead ideas. The census looks like it divides the drops between causes and it does not: 21 of them failed more than one check.
That overshoot is the most useful thing on the chart. A weak idea is rarely weak in one dimension — the thing with no evidence of a problem behind it is usually also the thing that is hard to sell, because both come from the same missing piece of work. If you are hoping that fixing the one objection you have heard will save an idea, the arithmetic here says it probably will not.
The check that kills most ideas
The biggest row, We found no documented problem, is not a judgement about the market. It is a judgement about the evidence: the sweep looks for a documented complaint — a real post, a real thread, a real review — that the idea would answer, and this many candidates had none that could be found.
Two honest readings of that, and only you can tell them apart:
- The problem is real but undocumented. Plenty of expensive problems are discussed on the phone and never written down anywhere a crawler will reach. Our sweep reads what is public; silence is not proof of absence.
- The problem was invented by the idea. This is the common one, and it is the failure mode that costs months, because nothing about it feels wrong from the inside. An idea assembled from a plausible-sounding gap looks exactly like an idea assembled from a real one, right up until launch.
The check cannot distinguish those either, which is why it drops the candidate rather than pronouncing on the market. The instruction it implies is narrow and cheap: before building, go and find the complaint in writing. If you cannot, that is information.
What made it through
One run, three stages. Bar widths are proportional to the counts, which are read from the run.
5 ideas out of 69 is a survival rate that reads as harsh until you notice what the alternative is. A sweep that returned forty promising candidates would have returned the 21 with no evidence behind them, and you would have spent your reading time on them.
What this census is not
It is one run. The counts on the next sweep of a different market will not be these counts, and nothing here says otherwise. Three limits worth stating before you lean on it:
- These are our checks, not the market's verdict. An idea dropped for leaning on one platform was dropped because a solo founder building on someone else's rails is exposed, not because nobody would buy it. The checks are listed on how it works; read them and disagree where you want to.
- The ratings underneath are model output. They sample fresh each run, so a second sweep of the same market would not return an identical list. The arithmetic on top of them is fixed; the judgements are not.
- It counts what the sweep could see. Our sources are public pages. A problem that lives in someone's inbox is invisible to this method, and that is a real ceiling rather than a caveat.
What survives all three limits is the shape: the most common way an idea dies is that nobody wrote down the problem it solves, and most dead ideas die of more than one thing at once. That is worth knowing before you spend a quarter finding out privately.