How to find your target audience when you have no users
Generic ICP templates sound precise and change nothing. This is how to extract named, interviewable segments from the signal your product already emits — before launch, without a panel.
Why generic ICPs fail
Most Ideal Customer Profile templates are a sentence of demographics and a list of verbs. They read as precise and act as a blindfold. They never tell you which message to write, which channel to pick, or why the person who sees your landing page will close the tab.
The useful unit is a segment: a named group you could interview tomorrow, defined by a situation and a motivation rather than a category. “Series-A founders” is an ICP. “Solo founders who just shut down their first landing page because it converted at 0.4%” is a segment.
Core / adjacent / edge / non-target: the only four labels that matter
For every segment you name, force yourself to label which of the four it is. This is the one discipline that keeps roadmap decisions honest.
- Core. This segment is the reason your product exists. Every feature must make sense to them.
- Adjacent. They benefit if you nail core. Don’t design for them first; design so you can reach them later.
- Edge. Interesting cases that stretch the product without defining it. Useful for stress-testing positioning.
- Non-target. Explicitly not your audience. Naming them out loud prevents half your roadmap going sideways.
How to extract segments from signal you already have
- Your one-sentence pitch. What it does, who it’s for, what becomes possible. If the sentence is vague, your segments will be vague.
- Your landing page or pre-launch copy. Every headline, subhead and feature block is a claim about who benefits. Segments fall out of these claims.
- Your category’s public complaints. App reviews, subreddits, and GitHub issues are a fossil record of frustration. Real segments reuse real language.
- Analogues that inspired you. Someone already sold into an overlapping market. Who loved them, who tolerated them, who bounced?
From these four inputs you can usually enumerate 4–8 candidate segments, then label them core / adjacent / edge / non-target. Tools like synthetic audience generation automate the extraction and make the labeling explicit.
A concrete example
A founder building a daily-journaling app for new parents listed their ICP as “parents, 28–45.” When they forced themselves into four labels, the core segment turned out to be first-time parents in the first 12 weeks, who haven’t slept more than four hours in a row and feel guilty for not remembering details. Everything downstream sharpened: the onboarding asks for the baby’s date of birth, not the parent’s age. Retention went up.
When to narrow vs. broaden
Narrow when
- Messaging works for every segment but converts well for none.
- Your core persona does not have a worse month without you.
- Two segments are fighting over the same roadmap.
Broaden when
- Core is nailed, retention is stable, and adjacent segments are already self-serving onto you.
- You have the evidence that the wedge product works end-to-end.
How to validate a segment before launch
- Generate the segment and a persona inside it.
- Run a 1-on-1 interview focused on last-problem stories, not opinions.
- Log their top three objections verbatim.
- Search for those objections in public reviews of analogous products. Matches mean the segment is real. No matches mean you are imagining them.
FAQ
Do I need demographics to define a target audience?▾
Not at first. Behavioral and motivational segments — what the person is trying to accomplish, what they tried before, and what they protect against — are far more predictive of adoption than age or income. Add demographics only when they change the behavior.
Does this work for B2B?▾
Yes, with one adjustment: your segments are roles and buying committees, not individuals. A B2B segment is typically (role, team size, trigger event, alternative-in-use). Interview the role, then walk up the committee.
How many segments should I start with?▾
Four to eight. Fewer than four and you are probably collapsing real differences; more than eight and you are fragmenting on trivia. Rank them core / adjacent / edge / non-target so your roadmap has a clear axis.
What if my audience is 'everyone'?▾
Then you don't have an audience yet, you have a wish. 'Everyone' never maps to a distribution channel, a message, or a price. Force yourself to name the one segment whose month is worse without your product.
Can I validate a segment without real users?▾
You can pressure-test it: generate the segment, interview a persona 1-on-1, log the objections, and check whether they match what real customers in analogous products complain about publicly. Real validation still requires real users — but synthetic research narrows the list.