Compare stated loss reasons with call evidence across every closed deal
This agent compares the loss reason picked in the CRM with what buyers actually said on calls, across every closed deal. Leaders see why deals are really lost, not just the polite answer.
Paste this into ChatGPT, Claude or your agent tool. See the prompt
- RevOps Reporting
- advanced
- RevOps
- Sales leader
- Category
- RevOps Reporting
- Difficulty
- Advanced
- Est. cost
- Not stated; the source says token cost is low but ownership cost is not
Problem
The CRM loss-reason dropdown usually says price because it is the polite answer. The real reasons sit in call transcripts nobody rereads.
Setup
A recurring Claude Code workflow joins each closed deal's CRM notes with its call transcripts, extracts a structured record (stated reason, evidenced reason, competitors, price sensitivity, missing features, champion strength, attributed quotes) and aggregates themes weighted by deal value. It runs incrementally as deals close, commits state between runs, and publishes two cuts: one for product, one for sales coaching.
Prompt
Starter prompt by The GTM Agent Weekly, written from the setup described on this page. It is not the source's own prompt; see the source link below for their write-up.
# ROLE
You analyze closed deals to compare the reason recorded in the CRM with what the calls
actually show, so product and sales leaders see the real patterns.
# INPUTS PER DEAL
- {{DEAL}}: outcome (won or lost), amount, close date, stated reason and notes from the CRM.
- {{TRANSCRIPTS}}: the deal's call transcripts.
# TASK A: ONE RECORD PER DEAL
Extract:
- stated_reason (from the CRM), evidenced_reason (from the calls), match yes or no
- competitors mentioned, price sensitivity (none, some, high), missing features requested
- champion strength (strong, weak, none)
- two or three attributed quotes supporting the evidenced reason
# TASK B: THEMES (after each batch)
Group reasons and weigh each theme by total deal value. Show how often stated and evidenced
reasons disagree.
# OUTPUT
1. For product: missing features and competitor themes, weighted by deal value, with quotes.
2. For sales coaching: process themes (single-threaded, no champion, late pricing), with quotes.
# GUARDRAILS
- Quotes must be verbatim and attributed. No evidence, say "not evidenced".
- Process only new closed deals each run, and keep a list of deals already done.
- No judgments about individual reps in the product cut. Replace every {{PLACEHOLDER}} before running. Works with any capable LLM and runner.
Tools
Steps
- Match calls to each deal by CRM association or by participant emails and dates, and log every deal where matching fails.
- Extract stated and evidenced reasons per deal, asserting an evidenced reason only with enough confidence.
- Aggregate themes weighted by deal value and flag where transcripts contradict the dropdown.
- Commit state so each run builds on the last, and publish separate product and sales views.
- Hand-read a sample quarterly to calibrate.
Results
Step-by-step guide; no outcome data reported.
What goes wrong
Gotchas
- Silent transcript gaps skew everything downstream; the guide logs every deal where call matching fails.
- Scheduled runs do not carry forward prior conclusions unless state is deliberately committed.
- Taxonomy drift and the temptation to skip the quarterly hand-read calibration erode accuracy over time.
Get 5 like this every week, free.
Real agent setups with sources and failure modes, in one short weekly email.
Free. One email a week. Unsubscribe anytime.
Source and credit
Second Axis · vendor
Source dated 2026-07-06 · last reviewed 2026-10-11. Summarized in our own words; see the source for details.