Why do some reps close more? A study that sized $1.3M of levers.
85,000 conversations, 240,000 CRM tasks and 420,000 emails turned into ranked experiments with dollar values.
↗Seven levers, $1.3M sized
Main sales pipeline, Apr–Oct 202665,441 threads · every figure script-generated and independently verified
Replies
49,921100%
Engaged
19,60539% of previous
Deals
4,36922% of previous
Closed
3,11671% of previous
15.9%engaged → closed
28.5% vs 10.7%best vs weakest market close rate
$1.41Mdeal value on AI-handled threads
n8n + CRM + Gmail→420K emails→AI vs human split→Attribution→Sized levers
Leadership could see some reps closed far more than others on AI-handled leads, with only anecdotes about why.
Two reports, every number generated by script and checked by an independent verifier. Each lever framed as a controlled test, not a forecast.
What we built
- AI versus human emails separated by a fingerprint with 100% recall.
- Rep attribution through a fallback chain of sources.
- Call-request detection that corrected an inflated count by two-thirds.
Results
- Repeat advertisers' share of 2026 revenue
- 60% ($3.89M)
- Win-back pool
- ~$673K
- On-time material vs never
- 85.7% vs 50.8% approval
- Call within 24h vs not
- 16.1% vs 13.8% close
- Saturday vs Monday follow-up reply
- 18.7% vs 29.4%
“Two of the seven levers became builds. That is what analysis is for.”
Stack
PythonpandasstatsmodelsSalesforceGmail API