The Segment-Level Win/Loss Autopsy
Expandi Skills/ Learn from your results

The Segment-Level Win/Loss Autopsy

A segment-level win/loss autopsy of an Expandi campaign: who replied and who didn’t, cut by seniority, function, industry, size, geo and source, with honest sample-size grading, a proposed learnings-ledger diff, and a pre-registered next experiment. Use when reviewing what worked.

DifficultyAdvanced AccessRead-only analysis, optional follow-up actions ask first

One quick form. It also puts you on the update list for this skill.

Try asking

Talk to your agent the way you would brief a teammate.

“What actually worked in the campaign we just wrapped up?”
“Which segment beat our baseline, and is that a real difference or noise?”
“Turn this campaign’s results into our next experiment”

What it does

  • Reconstructs which segments beat the baseline and which copy actually earned the replies.
  • Declares upfront which of three evidence modes it is running in, gold, silver or bronze, and states that mode’s blind spot.
  • Cuts performance by seniority, function, company size, industry, geography, and source, with a visible row for anything unknown.
  • Suppresses any cell too small to call, and gates every claimed difference against a statistical noise threshold.
  • Grades every finding by real evidence as A, B or C, and never lets a small sample drive a default.
  • Proposes a learnings-ledger diff and a pre-registered next experiment, and writes nothing until you approve it.

Install

Pick your platform.

claude.ai

Settings → Capabilities → enable code execution → Skills → upload the zip. Team/Enterprise workspaces may need an org-level toggle enabled first; custom MCP connectors (for live Expandi access) need a paid plan.

Next steps