The Loss Reason Postmortem: Auditing a Win-Loss Interview Dataset
Objective: Given a coded dataset of 12 loss-interview summaries with matching CRM loss-reason fields, separate CRM guesswork from buyer-verified themes and identify the single priority signal worth acting on this quarter.
You're the product marketing analyst on Grab's enterprise logistics team. Grab Business lost 12 of 20 competitive deals last quarter, and the CRM says 'Price' was the reason in 9 of them. You have buyer-interview transcripts for all 12 losses.
Distrust the CRM field, code the real interview themes, quantify them against the win pool, and segment by competitor before recommending an action.
Before you start
What you'll need
Free path (everything below is enough to finish)
Free, no account friction, pivot tables handle the theme x outcome cross-tab without a dedicated qual-analysis tool
The process
3 steps
Step 01 of 03
Anova Consulting and Klue's 2025 research found 85% of CRM loss-reason fields are inaccurate, reps guess, or record whatever is least embarrassing to log.
The CRM tags 9 of 12 losses as 'Price.' The buyer-interview transcripts for those same 9 deals only mention price directly in 2. What do you trust for the analysis?
Procedure
- Import loss-dataset.csv and freeze the header row
- Add a column flagging every row where crm_reason does not match either interview_theme column
- Count the mismatches: 7 of 9 'Price'-tagged deals show a different real theme in the interview
MISMATCH FLAGGED (7 of 9 'Price' rows) deal-04: CRM=Price | Interview=implementation timeline concern, sales rep unresponsive after demo deal-11: CRM=Price | Interview=unclear ROI framing, no dedicated onboarding contact named
Healthy
The analysis is built entirely from interview_theme columns; CRM fields are used only to spot the gap, never as evidence.
Unhealthy
Building the loss-reason chart straight from the CRM's crm_reason column because it's already there and tagged.
What this means
A CRM field that agrees with the interview only 22% of the time is not data, it's a guess with a dropdown attached.
So what do I do about it?
| Symptom | Action | Effort |
|---|---|---|
| CRM loss-reason report says 'we lose on price' | Re-tag the same 12 deals from interview transcripts before presenting any loss-reason chart to leadership | 30 min |
Step 02 of 03
A theme appearing in 40% of losses and 5% of wins is a priority signal. A theme appearing at similar rates in both wins and losses is noise, not a fix target.
Re-coded, 'implementation timeline concern' appears in 7 of 12 losses (58%) and 1 of 8 wins (12.5%). 'Sales rep responsiveness' appears in 3 of 12 losses (25%) and 2 of 8 wins (25%). Which one is the priority signal?
Procedure
- Build a pivot: rows = theme, columns = win/loss, values = count
- Convert counts to percentages within each outcome column
- Sort by the gap between loss% and win%, largest gap first
THEME LOSS% WIN% GAP Implementation timeline 58% 12.5% 45.5 pts <- priority Sales rep responsiveness 25% 25% 0 pts <- noise Unclear ROI framing 33% 25% 8 pts
Healthy
'Implementation timeline concern' gets flagged to product marketing as the priority theme with a 45.5-point gap.
Unhealthy
Treating 'Sales rep responsiveness' as equally urgent because it also showed up in a quarter of the losses, ignoring that it shows up just as often in wins.
What this means
A theme is only a signal relative to its own base rate in the win pool, not its raw frequency in losses.
So what do I do about it?
| Symptom | Action | Effort |
|---|---|---|
| Multiple themes appear in 25%+ of losses and nobody knows which to prioritize | Compute the win-pool base rate for every theme before ranking priorities | 30 min |
Step 03 of 03
Segmentation reveals insights aggregate data hides. Deal size, vertical, and competitor matchup can flip the story a headline number tells.
The overall win rate against Competitor X is 45%. Segmented by deal size, it's 70% under $50K and 20% above $50K. What does the aggregate number hide?
Procedure
- Bucket deals into under-$50K and over-$50K
- Pivot win rate by competitor and size band
- Flag any competitor where the size-banded win rates diverge by more than 20 points
vs Competitor X Win rate Under $50K (8 deals) 70% Over $50K (4 deals) 20% Blended (12 deals) 45% <- hides both realities
Healthy
The recommendation is scoped: 'we win the SMB segment against Competitor X, we lose the enterprise segment' with different actions for each.
Unhealthy
Reporting a single 45% win rate against Competitor X and recommending one fix for both segments.
What this means
A blended win rate is an average of two different competitive stories, act on the segments, not the blend.
So what do I do about it?
| Symptom | Action | Effort |
|---|---|---|
| A competitor's blended win rate looks mediocre but not alarming | Segment by deal size or vertical before deciding whether it needs an urgent fix | 30 min |
Final deliverable
A one-page loss-reason report that replaces the CRM's guessed reasons with interview-coded themes, ranks them by loss% minus win% gap, and segments the top competitor matchup by deal size.
See a reference example
Nubank Business, Q2 Loss-Reason Report (excerpt) PRIORITY SIGNAL: Implementation timeline concern (58% of losses, 12.5% of wins, 45.5-pt gap) NOISE: Sales rep responsiveness (25% of losses, 25% of wins, 0-pt gap) vs Competitor Y: blended win rate 52% hides a 78%/24% split between SMB and enterprise deals. Recommend: enterprise-specific onboarding proof points, not a blanket fix.
Success criteria
You're done when you can:
- Correctly identifies that the CRM's crm_reason field disagrees with the interview themes on the majority of flagged rows
- Ranks themes by the loss% minus win% gap, not raw loss frequency
- Segments at least one competitor matchup by deal size and flags the divergence