talk to users, always
There is no stage at which you are done talking to users. The founders who lose the thread are always the ones who stopped.
Every accelerator tells you to talk to users. It’s true and it’s not enough, because talking is the easy part and people lie to you for free. Everyone wants to encourage a founder. Ask someone if your thing is cool and they’ll say yes. Ask the same person to pay and the yes disappears. That gap is the only real signal in the conversation.
So talk to users constantly — there’s no stage where you graduate from it — but talk about their problem, never your solution. The factory owner doesn’t know he has a coordination problem. In his head his system is as efficient as any he’s seen. Tell him you’re solving fragmentation in manufacturing and he’ll smile, say nice things, and not pay you. Tell him his accountant fills the e-way bill by hand on a Friday night and you can automate it, and he reaches for his wallet. Same business, same owner. One conversation found the bottleneck; the other performed one.
Go to the right person, too. Talk to the workers, not the managers. Map the pipeline on a board — raw material in, two hours to unload, two weeks sitting, machine one, machine two — and somewhere there’s one step that’s the real bottleneck. The people doing the work know where it is. The person in the nice office has a story about it.
And watch what they do, not what they say. Someone on a 10-megapixel camera won’t switch to your 11-megapixel one no matter how warmly they describe it. Retention is the honest version of this: 500 users and 100 who come back next week without being paid have told you something 100,000 signups never will. What people repeat is what they mean.
The way I coach early teams is ping pong. Go DM a hundred users yourself this week. Come back Friday. Can those hundred do one repeatable action? Next week, again. The channel to users is never closed, at any scale. The founders who lose the thread are the ones who decided they were done listening.