The short answer
Find ten people who already have the problem. Ask each person to try one narrow task. Watch for where they stop, then change the product and tell them what changed. That is enough to start. An audience can come later.
First users are a learning group, not a launch number
The first ten people do not need to look like growth. Their job is to expose the gap between what you think the product does and what a new person can actually do with it.
Choose for relevance, not follower count. Someone who has felt the problem this week is more useful than a large account giving you a polite opinion.
A five-step playbook
1. Ask for one small action
“Can I get feedback?” gives the other person a job to invent. Ask them to try one thing instead: import a file, publish a page, compare two options or complete the first useful task. A smaller request gets a clearer answer.
2. Start where the problem already comes up
Look for people discussing the problem your product handles: a focused community, a support thread, replies to a relevant post or products serving the same audience. Join the conversation with context. Do not arrive with a drive-by link.
3. Show the exact thing you want tested
Send one page, screen or short path. Say who it is for, what it should help them do and where you are uncertain. Do not make someone study the whole product before they can help.
4. Make the reply easier than leaving
Put the question beside the work. Offer a comment box, a private note or a direct reply. Ask what happened, not whether they liked it. “Where did you expect to click?” will teach you more than “What do you think?”
5. Close the loop in public
Tell early users what changed because they spoke up. A short changelog entry or build note proves the conversation mattered. It also gives the next person a reason to contribute.
A message worth sending
Hi Sam, I saw you were trying to organise customer interview notes.
I’m building a small tool for that. Would you be willing to import one transcript and tell me where the process stops making sense?
Here’s the page: [link]. A one-line reply is genuinely enough.
The useful parts are the shared problem, the single task and the low-cost reply. Remove any sentence you cannot make true.
What to measure
Keep a simple record: who you asked, whether they tried the task, where they stopped and what you changed. Replies are useful. Completed tasks are stronger. A return visit without another prompt is stronger still.
Do not turn ten conversations into a fake conversion dashboard. At this stage, the question is whether the same problem and the same sticking point keep appearing.
Put it into practice
On Built In Public, you can mark a product as looking for early users, ask one specific question and give people a public comment or private feedback route. Add your product or find products already asking for early users.