How to Turn an Ideal Customer Profile Into a Research Brief
An ICP describes who you want. A research brief tells a researcher exactly which companies qualify and which do not.
5 min read · Updated 2026-03-11
Separate belief from criteria
Most ICPs mix evidence with assumption: a real observation about deal size sits next to a guess about industry fit. Split the two before briefing anyone, because only the first should constrain the research universe.
Convert each attribute into a testable rule
"Mid-market technology companies" is not researchable. "Software and IT services companies with 100 to 999 employees, headquartered in the UK or Ireland, excluding resellers" is.
Every attribute needs a rule a researcher can apply from public evidence, plus a decision for what happens when the evidence is ambiguous.
Write the exclusions down
Exclusions do more work than inclusions. Competitors, existing customers, dormant entities, holding companies with no trading activity and regions you cannot service should all be named in the brief.
Define the decision-maker layer
Name the functions and seniority you need, and the acceptable substitutes when the primary role does not exist at smaller companies. Without substitutes, a researcher either guesses or returns a gap.
Agree the evidence standard upfront
State which sources count as authoritative, how recent they must be, and what verification status a record needs to be counted as delivered.
Key takeaways
- Split evidence-backed ICP attributes from assumptions; only the first should constrain the universe.
- Every criterion needs a rule a researcher can apply plus a ruling for ambiguous evidence.
- Name your exclusions and acceptable role substitutes, or you will get guesses instead of gaps.
The brief we ask clients to approve
A workable brief runs to about a page: market boundary, inclusion rules, exclusion rules, decision-maker functions with substitutes, required fields, evidence standard, and the handling rule for records that cannot be confirmed.
Everything in it is a decision someone on the client side has to make. When those decisions are deferred to the researcher, the output reflects a researcher's guess about the business rather than the business.
Running a calibration batch first
Before a full build, we research a small batch of 50 to 100 companies and walk through the borderline cases together. Almost every brief changes at this point, and it is far cheaper to change it here than after 5,000 records.
The calibration batch also produces the working definition of a marginal account, which is the thing most briefs leave undefined.
Practitioner note: if two people in your team would classify the same borderline company differently, the brief is not finished yet.