define-icplisted

# Define ICP Build an ICP from evidence and falsifiable hypotheses, not a flattering description of everyone who could buy. ## Start Read `.agents/gtm-context.md` if present. Gather available customer, pipeline, loss, usage, and retention evidence. State the sample size and data limits. Choose the task: - **Create** an initial ICP from qualitative evidence. - **Refine** an existing ICP using wins, losses, retention, or expansion. - **Score** accounts against explicit fit criteria. - **Compare** candidate segments. ## Method 1. Define the job and painful current state the buyer is trying to change. 2. Identify company attributes that affect ability to realize value: size, stage, industry, geography, stack, process maturity, and constraints. 3. Identify trigger events that make the problem timely. 4. Identify disqualifiers and anti-patterns that predict poor retention or expensive delivery. 5. Separate company fit from contact role, intent, and timing. 6. Assess scoreability before scoring. Use only observable criteria, and do not score vague traits such as "innovative." If sample, outcome, or calibration evidence cannot support numerical distinctions, mark the segment or account `Not scoreable` and use qualitative importance and validation priority instead of invented weights, ranges, points, or tier cutoffs. 7. Compare the proposed ICP with known wins and losses. Name counterexamples. 8. Produce testable hypotheses for missing evidence. Use [the scoring guide](referen