ICP That Sharpens Itself

What this task does

Most ICPs were written once, from intuition. This routine compares what your ICP claims against what your win/loss data proves — which segments actually close fast, which supposedly great fits keep dying at security review — and proposes evidence-backed edits to the file every other task relies on.

Set up once

1. Confirm product-marketing.md holds the current ICP and scoring logic.
2. Connect HubSpot read-only with access to closed won/lost history.
3. Run monthly; expect meaningful edits quarterly.

The task brief

Paste this into Claude Cowork:

Compare our stated ICP in product-marketing.md against actual outcomes: pull closed won and lost from the last [FILL IN: 12] months, segment by industry, size, persona, and source, and compute win rate, cycle length, and deal size per segment. Where the data contradicts the ICP, say so plainly with numbers. Propose edits to the ICP and scoring rules, show the diff, and wait for my approval before saving.

What you get back

• A data-vs-belief gap report on your own ICP
• Proposed ICP edits with the evidence attached
• Scoring rules that route reps toward deals that actually close

Guardrails

• Small segments produce loud noise — require minimum deal counts before believing a pattern.
• ICP changes cascade into every downstream task; approve diffs deliberately.