Turn closed deals into a change, not a post-mortem deck.
Who to interview, the question guide, how to code themes across deals, and how to turn the pattern into a messaging or product change.
Inside: The interview slate, the question guide, and pattern-to-change with an owner.
Install it in one line, or paste it in.
~/.claude/skills/ and runs automatically when it is relevant.Connect your context. Set it to your motion.
it reads closed-won and closed-lost with stage history and loss reasons so the sample is not cherry-picked.
it mines what buyers actually said, not just the field a rep typed.
it folds structured buyer responses into the same theme coding.
This was built for a B2B SaaS org running competitive, committee-driven deals. Set these to your stack:
Run any theme taxonomy you like. The skill codes across deals and requires a pattern before it recommends a change, so point it at your buckets, not anyone else's.
| Set this | What it is | Default / Example |
|---|---|---|
| SCOPE | which deals qualify | closed in the last 1 to 2 quarters |
| MIX | won-to-lost interview split | roughly half and halfplus no-decisions |
| SAMPLE | how many interviews per round | 10 to 20 to see a pattern |
| SELECTION | who to prioritize | by segmentdeal sizecompetitorreason |
| INTERVIEWER | who runs them | a neutral third partynever the deal rep |
| THEMES | the coding buckets | product gapmessagepricetrusttiming |
| PATTERN_BAR | when a theme becomes a finding | a theme across 3+ independent deals |
| OWNER | who acts on each finding | PMM for messageproduct for gaps |
Everything the skill does, in full.
Stands up a repeatable win/loss program instead of a one-off post-mortem. It chooses which won and lost deals to interview and why, hands you the question guide that gets past "price," codes the answers into themes across many deals so you act on a pattern and not the loudest anecdote, and turns that pattern into a named change to the message or the product.
- 1Who to interview
Do not only interview losses, and do not let reps pick. Build a slate across MIX so you learn from wins, losses, and no-decisions, because a no-decision is a loss to the status quo and it teaches the most. Prioritize by SELECTION so the sample spans segments and competitors, not just the three deals someone remembers. Use a neutral INTERVIEWER, because buyers do not tell the rep who lost them the real reason.
- 2The question guide
Get past the stated reason. "Price" is almost never the reason, it is the reason that is polite to say. Ask what they were solving, who else was in the room, what nearly changed their mind, and what the winner did that you did not. Ask the same core questions every time so answers are comparable across deals. Open questions first, then probe, never lead.
- 3Coding themes across deals
Tag every interview into THEMES, then count across the sample. One buyer saying the onboarding scared them is an anecdote. Five saying it is a finding. A theme is only a finding when it clears PATTERN_BAR across independent deals, so a single loud loss cannot set the roadmap. Show the theme counts and the deals behind each.
- 4From pattern to change
Every finding gets an owner and a specific change, not a discussion. A message theme goes to PMM as a positioning or objection-handling change. A product theme goes to product as a gap with the deals it cost. A trust or proof theme becomes a case study or a security page. The program has failed if it produces a deck nobody acts on, so each finding ends as a named change with an owner and a date.
- Interviews are run by a neutral party, never the rep who owned the deal.
- Wins and no-decisions are in the sample, not only losses.
- No theme becomes a finding until it clears the pattern bar across independent deals.
- "Price" is probed, never accepted at face value as the real reason.
- Every finding ends with a named owner and a specific change, not a summary slide.
WIN-LOSS PROGRAM · 14 interviews · illustrative Sample: 6 lost · 5 won · 3 no-decision · mix of segments and 2 competitors Themes (count across independent deals): Product gap: integration missing 5 deals FINDING Message: value story unclear early 4 deals FINDING Trust: no proof at our size 3 deals FINDING Price stated as reason 4 deals probe showed it was fit, not price Changes: 1. Integration gap -> Product. Costed 5 deals this quarter. Roadmap review. 2. Value story -> PMM. Rewrite the first-call narrative, ship objection guide. 3. Proof gap -> PMM. Publish 2 case studies in the losing segment.
SAMPLE (10 to 20), PATTERN_BAR (3+ deals), and the MIX split are defaults, not laws. They suited a mid-market competitive motion. If your deal volume is lower, run smaller rounds more often and let patterns build over time. The coding logic does not change. The thresholds are yours.
Where an operator takes this next.
The round is coded. Here is where an operator takes it once the manual version proves out.
The interviews find the pattern; the automation is what makes sure it turns into a shipped change.
A scheduled Claude task pulls closed-won and closed-lost from Salesforce every quarter and proposes the interview slate automatically, mixed across segments and competitors.
Run past discovery and closing calls through Gong to pre-code themes before a single interview is booked, so live conversations go deeper instead of covering ground you already have.
Push each finding with its owner and deadline straight into a task tool or Slack channel for PMM and product, so "the value story is unclear" does not die in a deck.
One skill is the on-ramp.
A single skill does one job. Chained into a playbook, or run as a full build, it becomes a system. Here is where this one plugs in.