A decision framework for comparing customer segments by problem frequency, consequence, reachable context, existing alternatives, and ability to learn.
Start with the real decision
An ICP becomes useless when it is a collection of flattering demographics. Early teams need a segment they can recognize, reach responsibly, learn from, and serve with the product they actually have.
Choose the segment with the clearest repeated job, not the largest imaginable market.
A practical sequence
- Define by situation. Combine role, job, trigger, constraint, and current alternative.
- Compare problem frequency. Ask how often the situation occurs and whether it persists.
- Compare consequence. Look for delay, risk, lost revenue, coordination cost, or missed outcome.
- Check reachable evidence. Can you observe the problem and speak with enough people without relying on invasive tactics?
- Test serviceability. Can today’s product deliver a meaningful result for this segment?
Keep the evidence reviewable
For every item, record source and date, customer situation, observed behavior, your interpretation, and the decision it may change. Keep observation separate from inference. Add evidence that would falsify important conclusions so a future reviewer can challenge them.
Deduplicate recurrence. Several posts may quote one announcement, and one story may be reposted across platforms. Mark shared origins before calling something a stable pattern rather than a concentrated reaction or temporary noise.
Common ways the method fails
Demographic portraits. Age and company size do not explain the buying job.
Choosing only by TAM. A huge market can still be unreachable and weakly motivated.
Ignoring rejection evidence. People who choose an alternative reveal the segment boundary.
Separate research, judgment, and outreach
Public discussion can reveal jobs, language, and alternatives, but it does not grant permission for dossiers or unsolicited pitches. Decide on participation separately, follow community rules, disclose interests, and make any answer useful without a product link.
SeeVoid discovers, groups, and preserves source context. It does not decide representativeness or turn signals into automatic public posting.
End with a decision card
Write the decision, three strongest items, one counterexample, unknowns, owner, review date, and reopening trigger. Weak signals justify interviews, medium signals a page or workflow test, and heavier investment needs independent recurrence plus behavior.
Before committing, run the method on a small sample and ask whether two people could reach the same conclusion from the evidence. If not, clarify the rubric, capture the missing context, or lower confidence. Prefer a reversible action that creates new evidence over a large bet that only expresses conviction.
Stop when new material no longer changes the choice. Research should shorten the path to knowing what to do, what not to do yet, and why.
What completion looks like
Pick one primary segment for the next learning cycle, not forever. State why it won, what evidence would dethrone it, and which adjacent segment remains a comparison group.
Further reading:SBA market research guide.