The dogma in startup advice is to build things that solve your problem.
The main trouble with that advice is that you don't know how to solve your problem. If you did, you would've solved it. The idea that a product or whatever your building will solve it is just a guess.
The best software I ever built solved problems for people sitting next to me and not directly for me. So much for the dogma.
So why guess when you can solve a problem for the person sitting next to you. Of course, someone sitting next to you is a metaphor for anyone you understand well enough, but their problem isn't necessarily directly your problem anymore. It could even be for yourself six months ago or a year ago.
Let's get concrete; a few years back, I saw the business people sitting next to me struggle with SQL queries against a production database at a startup. So I built a tool, a web app, for them. Then they asked for more features, and before I knew it, half the company was using this tool internally to do their work.
How did I know the solution? Easy! I just saw them struggle, and I was very logical about the solution and not emotional in any way. On the flip side of that, I've built software for myself, many times, to solve my problems, and it's flopped.
The problem with the dogma is that few people can solve problems well for themselves; few people are that introspective without emotions.
Most people are very emotionally attached to their problems, and their solutions are phenomenal in their heads. But that emotion rarely exists when it's someone else's problem. You see their pain, but you can be objective about it.
Still not convinced? It would be crazy for doctors to operate on themselves, I'm sure some could, and in fact, a doctor in the south pole had to do it when his appendix was about to rupture, and no other doctors were around. But the nerves didn't make that operation easy.
Get my weekly email, The Postcard, with the best ideas from each week.