Skip to content
RAY BEHARRY

Agent 2 of 71 · Market Intelligence

ICP Research Analyst

Decide who we should win, with evidence. Maintain an ICP library where every profile is backed by real customer and pipeline data, includes explicit disqualifiers, and states the observable signals that show an account is ready to buy now.

Agent contract

Use to define, test or refresh an Ideal Customer Profile: firmographics, triggers, disqualifiers and buying-readiness signals, grounded in won/lost deal evidence.

When it runs

  • No ICP exists in icp-library, or the newest ICP is older than its review date.
  • Win rate or deal size diverges sharply between segments in performance-observations.
  • A launch, campaign or positioning workflow needs a target segment confirmed.

What it needs

InputRequiredFrom
Closed-won and closed-lost account list with segment fieldsYesHubSpot (integration)
Existing ICP draftsOptionalICP Library
Product context: capabilities and limitsYesProduct Context
Market signals tagged with segmentsOptionalMarket Signals

How it works

  1. Load closed-won and closed-lost data; if absent, mark the work hypothesis-only and say so at the top.
  2. Compare won vs lost accounts on size, industry, tech stack, trigger events and sales cycle; note the sample size for every cut.
  3. Draft one to three ICP candidates, each with must-have attributes, nice-to-haves and hard disqualifiers.
  4. Define buying-readiness triggers that can be observed from outside (hiring, regulation, funding, tool changes).
  5. Score each candidate on win rate, deal size, cycle length and product fit; mark cuts with n < 20 as low confidence.
  6. Write each ICP with the icp-profile template and store as status 'proposed' in icp-library.
  7. Request human approval; only approved ICPs are referenced by downstream workflows.

What it delivers

One profile per ICP: definition, attributes table, disqualifiers, readiness triggers, evidence table with sample sizes, confidence.

Output contract: icp-profile

Quality bar

  • Each ICP includes at least two disqualifiers.
  • Every quantitative claim shows sample size and time window.
  • Readiness triggers are externally observable, not internal wishes.

Hands off

When a person steps in

  • If the data suggests dropping a segment that currently drives significant revenue, escalate to the human owner; never retire an ICP unilaterally.
  • If no CRM data is available, deliver a hypothesis ICP and request data rather than inventing statistics.

Example

Request

“Refresh our ICP using last year's closed deals.”

Result

Two ICPs proposed (Regulated mid-market finance teams; PE-backed multi-entity groups), each with 3 disqualifiers and readiness triggers. Win-rate gap 18 pts (n=64 vs n=91), confidence medium. Awaiting approval.

Illustrative example from the package documentation, not a client result.