Review Mining and Voice-of-Customer Synthesis

Paste a batch of reviews, survey verbatims or support tickets. You get the recurring pain themes with counts and quotes, the exact words customers use for the problem, what they compare you against, and the copy-ready language bank drawn from their sentences rather than yours.

A Markdown voice-of-customer synthesis containing: (1) a count of texts analysed and any truncation warning; (2) pain themes ordered by support, each with a count, two or three verbatim quotes and the violated expectation; (3) praise themes with the same treatment; (4) a language bank of customer vocabulary grouped for copy use, including which marketing terms never appear; (5) alternatives and workarounds mentioned; (6) churn signals; (7) a 'What this sample is biased toward' section. No percentages, no sentiment scores.

# Voice-of-customer synthesis — expense management (12 reviews)

Twelve texts, all appear complete. Small sample: treat every count below as a count, not a rate.

## Pain themes

### 1. Approval routing is broken, and it is the theme that loses customers (4 of 12: R1, R6, R8, R11)
> R1: "Approvals are a nightmare though, my manager never gets the notification."
> R8: "Submitted my expenses, they sat for two weeks, turned out my approver had left the company and nothing reassigned."
> R6: "I never know whose desk something is on."

**The expectation being violated:** that a submitted expense is now someone else's problem. Instead the submitter remains responsible for chasing it, which is the exact job they bought the product to remove. R8 describes a specific failure — no reassignment when an approver leaves — that is a concrete, fixable defect rather than a design opinion.

This is the only theme that appears in the churn signal below, which makes it the most expensive item in this synthesis.

### 2. The product feels unfinished around a strong core (3 of 12: R3, R10, R12)
> R10: "The receipt OCR is the best I have used. Everything around it feels half finished."
> R3: "The mobile app crashes when I try to submit more than about 10 receipts at once."
> R12: "Good for a small team. I would not run 200 people on this."

**The expectation being violated:** that the quality of the part they touch first predicts the quality of the rest. R10 states the pattern explicitly and R3 gives it a reproducible bug.

### 3. Onboarding took longer than promised (1 of 12: R2)
> R2: "Setup took three weeks longer than they said."

Single review, kept separate from the supported themes. Worth noting only because it is a promise-versus-delivery gap, which tends to be under-reported in reviews rather than over-reported.

### 4. Support responsiveness (1 of 12: R4)
> R4: "Support took 4 days to answer a question about a duplicate charge. Product is fine, support is not."

Single review. Listed because a duplicate charge is a money question, and four days on a money question generates a chargeback rather than a follow-up ticket.

## Praise themes

### 1. Receipt capture and OCR quality (3 of 12: R1, R5, R10)
> R10: "The receipt OCR is the best I have used."
> R1: "It reads crumpled receipts my old tool could not."
> R5: "Love that I can just photograph a receipt in the taxi and forget about it."

This is your positioning. It is the only capability anyone describes in superlatives, and R5 describes the emotional outcome — photograph and forget — better than any marketing copy would.

### 2. Accounting integration works (1 of 12: R7)
> R7: "Integrates with our accounting system which is why we bought it. That part works."

Notable because R7 names it as the purchase reason. One review, but it identifies a buying trigger.

### 3. Price relative to alternatives (2 of 12: R9, R11)
> R9: "Cheaper than the alternatives we looked at and does 90% of what they do."

Read R9 and R11 together. R9 is the value argument; R11 is the same customer position after the approval problems — "the approval chaos was not worth the savings". Price advantage is real and it is conditional on the approvals theme being fixed.

## Language bank

**For the problem:** "crumpled receipts", "in the taxi", "whose desk something is on", "sat for two weeks", "chasing", "back from spreadsheets".

**For the product:** "photograph a receipt and forget about it", "reads receipts my old tool could not", "the OCR". Note that customers say *receipt scanning* and *OCR*; nobody says *expense capture*.

**For the outcome:** "forget about it", "that part works", "does 90% of what they do".

**Words customers never use:** *automation*, *streamlined*, *seamless*, *spend management*, *visibility*, *compliance*. If those appear in your marketing, they are your vocabulary and not theirs. "Photograph it in the taxi and forget about it" is a better headline than anything on that list.

## Alternatives and workarounds mentioned
- **Spreadsheets** (R2) — the real prior state, and the comparison your value argument should be built against.
- **A previous tool with weaker OCR** (R1).
- **An old system being returned to** (R11) — the competitor here is the incumbent they left, which is the most dangerous kind.
- No competitor is named anywhere in twelve reviews. That is itself a finding: buyers in this sample are comparing you to what they were doing, not to another vendor.

## Churn signals
> R11: "We are moving back to our old system next quarter. The approval chaos was not worth the savings."
> R12: "I would not run 200 people on this." — not churn, but a stated ceiling on expansion.

R11 is an explicit departure with a named cause and a timeline, tied directly to the largest pain theme. One review, and it is the most actionable sentence in this entire set.

## What this sample is biased toward
- **Reviews are written by the unusually delighted and the unusually angry.** The quiet middle is absent by construction and it is probably most of your customers.
- **Submitters, not approvers or finance teams.** Almost every text is written from the perspective of someone filing an expense. The approver's experience — the side that appears broken — is described only second-hand, and the finance administrator who actually buys the product is not represented at all.
- **Small teams.** R12 explicitly caveats scale and nothing here comes from a large deployment, so the approval-routing failures may be considerably worse at size than these reviews suggest.
- **Twelve texts.** Every count in this document is a count out of twelve. Do not turn any of them into a percentage.