Refine your ICP with win/loss data
Puts your ICP on trial against actual deal outcomes: every criterion gets a confirmed/challenged/untested verdict, hidden win and loss pockets get surfaced, and loss reasons get sorted into targeting versus execution problems. You leave with an evidence-backed revised ICP and one clean hypothesis to test next quarter.
You are a win/loss analyst who has reviewed thousands of B2B deals, and you hold an uncomfortable truth: most ICPs are written once from early wins and never confronted with the losses that followed. The result is teams confidently targeting a profile that made sense two years ago while their actual win patterns have quietly moved. Your method is simple and ruthless: put the ICP on trial, with the win/loss record as evidence. I'll give you my current ICP and my recent deal outcomes. Produce: 1. THE VERDICT TABLE: each ICP criterion tested against my data — does it actually correlate with winning? Label each: CONFIRMED (wins support it), CHALLENGED (losses cluster inside it), or UNTESTED (not enough data). 2. HIDDEN PATTERNS: segments where I win more than my ICP predicts, and segments inside my ICP where I lose disproportionately — with your hypothesis for why. 3. LOSS AUTOPSY THEMES: the top 3 recurring loss reasons and whether each is a targeting problem (wrong accounts) or an execution problem (right accounts, lost anyway). Do not let me blame targeting for execution failures. 4. THE REVISED ICP: my ICP rewritten with the changes the evidence supports — additions, tightenings, and removals, each tagged with its supporting data. 5. NEXT QUARTER'S TEST: one targeting hypothesis from this analysis worth testing deliberately, and how to measure it. Before you write anything, interview me. Ask me these questions ONE AT A TIME, waiting for my answer each time: 1. Paste your current ICP, or describe it as your team actually uses it. 2. List your last 15-30 closed deals — won or lost, with industry, company size, deal size, and the loss reason where known. 3. For losses marked 'no decision' or ghosted: what do you suspect really happened? 4. Has anything about your product, pricing, or market shifted since the ICP was written? Once you have my answers, run the trial. If my data is too thin to support a conclusion, say 'insufficient evidence' rather than manufacturing a pattern.
How to use it
- 1
Copy the prompt into Claude, ChatGPT, or any LLM.
- 2
Export your last 15-30 closed deals from the CRM with outcome, size, industry, and loss reason before starting.
- 3
Be candid on question 3 — 'no decision' losses usually hide the most important pattern in the dataset.
- 4
Act on the verdict table: challenged criteria get revised or explicitly retested, not quietly ignored.
- 5
Rerun this quarterly with fresh deals; ICP refinement is a rhythm, not a rescue mission.
Best practices
Respect the targeting-versus-execution split — tightening your ICP won't fix losses caused by slow follow-up or weak demos.
Weight recent deals over old ones; your market position a year ago is not evidence about today.
Treat 'insufficient evidence' answers as homework: start capturing loss reasons properly so next quarter's trial has a real record.
When a hidden win pocket appears outside your ICP, test it deliberately with a small campaign before rewriting strategy around it.
Example: what this looks like in practice
A RevOps lead at a marketing analytics company pastes their ICP (B2B SaaS, 50-500 employees, has a marketing team of 3+) and 24 recent deals. The verdict table confirms the marketing-team criterion but challenges the SaaS focus: a third of wins came from e-commerce brands the ICP technically excludes, while SaaS deals under 100 employees lost disproportionately, mostly to 'no decision'. The loss autopsy tags the no-decisions as a targeting problem — companies too early to feel the pain — while two competitive losses were execution. The revised ICP raises the SaaS floor to 100 employees and adds e-commerce as a confirmed segment. Next quarter's test: a dedicated 200-account e-commerce campaign, which converts at 1.8x the historical rate.
Best fit
This prompt is one gear in a bigger machine. We orchestrate 20+ tools into outbound systems our clients own — and guarantee the results.
Apply for a Pilot Spot → →Frequently asked questions
Test each ICP criterion against your actual deal record: do wins really cluster inside it, and do losses cluster anywhere specific within it? Look equally hard for wins outside the ICP — they reveal segments you're underestimating. This prompt structures the whole trial and separates targeting problems from execution problems, so you only revise the ICP where evidence supports it.
More icp, targeting & lead scoring prompts
Define your ICP from your 10 best customers
Reverse-engineers a tight, one-page ICP from your actual closed-won history instead of aspirational guesswork. It forces narrow ranges, separates signal from coincidence, and flags when your customer base actually contains two different ICPs — the insight most founders miss until their outbound numbers tell them the hard way.
Design a lead scoring model with firmographic and behavioral points
Produces a complete, implementable lead scoring model: firmographic fit points, behavioral intent points with decay and negative scoring, threshold bands tied to actions, and a monthly validation loop. It caps complexity at 12 criteria and adapts to what you can actually track, so the model ships instead of dying in a spreadsheet.
Build a negative ICP: who to actively avoid and why
Turns your worst-customer war stories into a formal negative ICP: hard disqualifiers, stacking yellow flags, and the false-positive profile that keeps fooling your team. Includes list-building exclusion filters and a polite decline template, so avoiding bad fits becomes operational instead of a lesson everyone relearns quarterly.