icp-sharpener-b2b

ICP Sharpener (B2B SaaS)

Turn "we sell to B2B companies that need X" into an ICP a marketer or SDR can act on without asking follow-up questions.

What you need before starting

Gather (ask for whatever is missing):

1. Product: what it does, in the user's own words. One paragraph max.
2. Evidence: 5–15 real accounts, best customers, worst customers, recent churns, recent closed-lost. Names optional; descriptions required.
3. Deal mechanics: typical ACV, sales cycle length, who signed, who blocked.

If the user only has a website URL, fetch it, infer a draft, and label every inference as an assumption to validate.

Workflow

Step 1: Split "customer" into three roles

For each good-fit account, identify separately, these are rarely the same person and conflating them is the #1 ICP error:

• User: who touches the product weekly
• Buyer: who signed / owns the budget line
• Champion: who dragged the deal through internally

Step 2: Find the pattern in the evidence, not the aspiration

Compare best vs. worst accounts across: company size, funding stage, tech stack, team structure, motion (PLG/sales-led), and the situation right before they bought. The ICP is what the best accounts share that the worst lack , not who the founder wishes would buy.

Step 3: Name the buying trigger

For each strong account, answer: what changed that made them start looking? (New leader hired, funding round, tool ripped out, compliance deadline, growth stalled.) An ICP without triggers produces outbound with no reason to reply.

Step 4: Write the disqualifiers

List 3–5 hard disqualifiers (too small, wrong stack, no budget owner, needs on-prem). A segment you can't say no to isn't a segment.

Step 5: Output the ICP one-pager

Produce exactly this structure:

## ICP: <segment name, specific, e.g. "Series A–B PLG SaaS, 20–80 FTE, first marketing hire just landed">

**Firmographics**: size, stage, geo, stack
**User / Buyer / Champion**: title + what each cares about (one line each)
**Buying triggers**: 2–4 observable events
**Disqualifiers**: 3–5 hard nos
**Where they live**: communities, events, newsletters, who they follow
**Message angle**: the one sentence that names their situation better than they can
**Confidence**: which parts are evidence-backed vs. assumed, and the fastest way to test the assumptions

If the evidence supports two distinct segments, write two one-pagers and recommend which to lead with (score on: revenue evidence, reachability, sales cycle speed). Never output more than three, more than three ICPs is zero ICPs.

Quality bar

• Every claim traces to an account in the evidence or is flagged as assumption.
• A stranger could build a 200-account list from the firmographics alone.
• The message angle names a situation, not a benefit ("your first marketing hire just inherited a backlog," not "save time and grow faster").

Related skills

positioning-canvas to turn the ICP into a claim. outbound-sequence-writer to act on it. landing-page-teardown to check the page speaks to it.