We have about fifteen customers and they are all different. How do I find an ICP out of that?
Do not look at who bought, look at who is happy. Rank the fifteen by usage and by whether they would be annoyed if you disappeared, then take the top three or four and hunt for what they share that the bottom ones do not. It is rarely industry, it is more often a specific situation: they just hired someone into a role, they hit a volume where the spreadsheet broke, they already run a particular system. Write that trigger down as your first ICP hypothesis, go find twenty more companies in the same situation, and see if they convert faster. Fifteen customers is enough to have a hypothesis and not enough to be sure, so keep revising.
Go deeper
4 resources, 1 India-specific, 4 link-checked.
📄 Article
✓ Link checkedFreeIntermediate
Five questions you can answer this afternoon from data you already have, including the underused one about what objections churned customers gave you. Written by a16z's GTM network partner, so it is built for operators rather than theorists.
A real company showing its ICP before and after, including the messy middle where they were guessing. The point that narrow ICPs beat broad ones, and that refinement takes 3 to 12 months, is the honest version nobody puts on a slide.
The story of an Indian company arriving at a very specific ICP, founder-developers at zero-to-one startups facing a build versus buy decision. Shows ICP discovery as a series of uncomfortable questions rather than a workshop.
Finding your ICP starts with interviews you are not allowed to sell in, and this is the clearest guide to running them. The rule about avoiding hypothetical questions alone will save you from a fake ICP.
