Sales prompt of the day

Rebuild the ICP from your own last four quarters, not the deck from 2024

Prospecting feels random and the official ICP was written before the product changed.

You are a revenue analyst rebuilding an ideal customer profile from observed outcomes. I will paste closed-won and closed-lost records. Produce:

A) A WIN-RATE TABLE across every attribute I supplied (size, industry, region, tech stack, entry product, source, deal size band) — with n shown for each cell and any cell below n=5 marked as insufficient rather than ranked.
B) THE DIFFERENTIATORS — the three attributes that most separate won from lost, plus the two attributes commonly believed to matter that the data does NOT support, each with the number that shows it.
C) A SCORING RULE — a simple, human-checkable rule (not a model) that an SDR can apply to a new account in under a minute, with the disqualifiers stated as hard stops.
D) A COVERAGE WARNING — where the data cannot answer the question because we never tried that segment, so a low win rate is not confused with a low-quality segment.

Data: [PASTE CLOSED-WON AND CLOSED-LOST WITH ATTRIBUTES, DATES, DEAL SIZES, LOSS REASONS]

Rules: Use only what I give you — do not invent a company fact, a metric, a quote or a name. If something is missing, write "not provided" and ask for it. Mark every claim I would repeat to a customer as needing verification before it leaves my outbox. Never include another customer's confidential information. Verify every attribute and win-rate against the CRM record before this ICP changes who reps call.

Why it works — An ICP written from opinion sends the whole team into the same wrong accounts. Ranking by observed win rate with n shown — and explicitly separating 'we lose here' from 'we never tried here' — is what keeps a data-driven ICP from strangling a segment on three data points.