Most product ideas are guesses. These aren't.
Someone Should Build started as a tool for deciding what to build next, and it still works that way — it just happens to be useful to anyone facing the same question.
Why it exists
The usual way to pick a product is to have a hunch and then go looking for evidence that supports it. That is backwards, and it is why so much software gets built for nobody.
People already write down what is wrong with the software they use. They do it constantly, in public, in enormous volume — 68,506 times in this database so far. The information exists. It is just unreadable at that scale without something to group it.
That grouping is the whole product.
How it's run
Independently and deliberately small. There is no sales team to talk to and no demo to book — the free tier shows you real output, and if it isn't useful you will know within a minute.
One price, $29 a month, cancellable from a self-serve portal. No annual lock-in, no enterprise tier, no per-seat maths.
What it refuses to do
- Invent anything
- Every opportunity traces back to posts real people wrote, and every one links to the source. A model summarises; it never authors a complaint.
- Hide the method
- The scoring, the clustering threshold and the source mix are all published. A score you can't interrogate is a horoscope.
- Oversell the data
- The home page says outright that most evidence is app reviews and that this is the wrong tool for greenfield ideas. Better you know now than after paying.