── SKILL
closed-won-analysis✓ APPROVED

Extract the repeatable reason you win.

Reads your won deals, finds the signals, personas, and motions that repeat, and ranks active lookalikes to work.

Inside: The pattern extraction, the win-signal ranking, and the lookalike list.

01 / HOW TO USE

Install it in one line, or paste it in.

1
In Claude Code (one command)
Copy the install line, paste it into your terminal, and restart Claude Code. The skill installs itself to ~/.claude/skills/ and runs automatically when it is relevant.
2
In Claude, ChatGPT, or a Project (no terminal)
Open the file, then upload it to your chat or paste its contents in. A skill is just a markdown file of instructions, so any capable AI can follow it.
New to skills? A skill is a plain-text file that teaches your AI a workflow. Point any capable assistant at it and it follows the steps, on your data.
02 / MAKE IT YOURS

Connect your context. Set it to your motion.

CONNECT YOUR CONTEXT · AND WHY IT HELPS
a CRM

it reads all of the above automatically, across every won deal in the window you pick.

a meeting or transcript tool

pulls the actual words buyers used, so the pattern is "why they said yes," not just the fields.

a product-analytics tool

adds which features the winners adopted first, so the profile includes real behavior.

SET IT TO YOUR MOTION

This was built for a B2B SaaS org running a staged pipeline. Set these to your stack:

Run any segmentation you like. The skill reports the patterns that actually repeat in your wins, so point it at your fields, not anyone else's.

Set thisWhat it isDefault / Example
WON filterhow you mark a winStage = Closed Won
WINDOWthe win-date range to analyzetrailing 4 quarters
SEGMENT fieldshow you group accountsindustryemployee countregion
PERSONA fieldthe title that signsprimary contact role or title
MOTION fieldsinbound vs outbound, sourcelead sourcechannel
COMPETITOR fieldwho you displaced or beatcompetitor on the deal
USE_CASE fieldthe problem you solvedprimary use case or product
MIN_SAMPLEfewest wins before a pattern counts5 (raise if you close a lot)
03 / THE FULL SKILL

Everything the skill does, in full.

── WHAT THIS DOES

Reads your closed-won deals and finds what they have in common: the segments you win, the personas who sign, the motions that work, and the competitive spots where you come out ahead. It turns that into one target profile you can hand a rep, then ranks your active pipeline by how closely each open deal matches your winners. The point is not a nice chart. It is a shorter list of accounts that look like the ones you already closed.

── THE METHOD
  1. 1
    Pattern extraction (what repeats)

    Group every won deal across your segment, persona, motion, competitive, and use-case fields. Report only the patterns that clear MIN_SAMPLE, so one lucky whale never becomes a "profile." For each pattern, show the share of wins, the average deal size, and the average cycle length, so a common-but-small pattern is not confused with a rare-but-huge one.

  2. 2
    Win-signal ranking

    Rank the patterns by lift, not raw count. A signal earns its place when won deals show it far more often than the base rate of all your deals. "Half our wins are mid-market" means nothing if half your pipeline is mid-market. The signals that matter are the ones over-represented in wins.

  3. 3
    Target profile (the ICP you actually close)

    Fold the top signals into one plain-language profile: the segment, the persona who signs, the motion that lands them, the use case that resonates, and the competitive setup where you win. This is not the ICP on the website. It is the ICP your win data votes for.

  4. 4
    Lookalike ranking (active pipeline)

    Score every open deal by how many profile signals it hits. Return a ranked list, best matches first, with the signals each one shares with your winners and the ones it is missing. This is the payoff: a shorter list of open accounts that look like deals you already won.

── QUALITY GATES
  • No pattern reported below MIN_SAMPLE. Small samples lie, and one big logo is not a trend.
  • Signals ranked by lift over base rate, never by raw count alone.
  • Every lookalike shows the specific signals it matches, never just a score.
  • Competitive and use-case claims come from a field or a transcript, never invented.
── OUTPUT (EXAMPLE)
EXAMPLE OUTPUT
CLOSED-WON PATTERNS · 34 wins, trailing 4 quarters

Top win signals (by lift over base rate)
Signal                        Share of wins   Avg size   Avg cycle
Mid-market, 200-800 staff     62%             $41K       38 days
Champion = Head of Ops        55%             $44K       35 days
Inbound demo request          48%             $39K       29 days
Displaced a manual/DIY setup  44%             $47K       33 days

Target profile (what you actually close):
  Mid-market ops leader, inbound, replacing a manual process.
  Fast cycle, above-average deal size.

Active lookalikes (open pipeline, best match first)
Account      Signals matched            Missing
Haledon      3 of 4 (seg, persona, mo)  no competitor read
Brightsea    3 of 4 (seg, persona, uc)  outbound, not inbound
Corverin     2 of 4 (seg, use case)     enterprise size

Next move:
  1. Work Haledon and Brightsea first. They match your winners on 3 of 4.
  2. Add a competitor field to your next 10 wins. It is your thinnest signal.
── WHERE THE NUMBERS COME FROM

MIN_SAMPLE (5) and the trailing-4-quarter window are defaults, not laws. They suited a mid-market SaaS cycle with steady volume. If you close a handful of large deals a year, widen the window and lower the floor with your eyes open. If you close hundreds, raise the floor so only strong patterns survive. Lift is always measured against your own base rate, so the profile is yours, not a benchmark.

── EXAMPLE CHAT
EXAMPLE CHAT
You
Pull the pattern out of our Q2 wins, 34 deals, and tell me who in open pipeline looks like them.
CLAUDE
Read's done. Your website ICP and your win data don't agree. Signal Share of wins Lift vs base Mid-market, 200-800 staff 62% 1.4x Champion = Head of Ops 55% 1.8x Displaced a manual/DIY setup 44% 2.1x "Displaced a manual setup" is your strongest signal by lift, not your biggest by share. It's rare in your total pipeline (21%) but shows up in nearly half your wins. That's the profile that actually predicts a close, not the mid-market segment everyone already assumes. One sharpener: add a competitor/incumbent field to more deals, it's your thinnest signal right now.
You
Score my open pipeline against that profile.
CLAUDE
Ranked, best match first. Account Signals matched Missing Haledon 3 of 4 (seg, persona, mo) no incumbent read Brightsea 3 of 4 (seg, persona, uc) outbound not inbound Haledon and Brightsea are your two best bets this quarter, both hit 3 of 4 win signals. One sharpener: connect your CRM so this scoring re-runs automatically as new deals open, instead of a one-time snapshot on today's list.
04 / GO FURTHER

Where an operator takes this next.

The read is step one. Here's where an operator takes it once the manual version proves out.

You built the read once; now it runs itself.

01
Rescore the pipeline weekly, not once

Point a scheduled Claude task at Salesforce nightly and write the lookalike score back to a custom field on the opportunity.

02
Hand the profile to the SDR team

Feed the top win signals into Clay to re-rank the outbound target list so reps prospect the accounts that already look like winners.

03
Alert on a new lookalike the day it appears

DM the account owner in Slack the moment an open deal crosses the match threshold, instead of waiting for the next QBR.

05 / PART OF A BIGGER PLAY

One skill is the on-ramp.

── GET NEW SKILLS IN YOUR INBOX

One operator-built skill a week. Free.