/ CHAPTER 04 — 6 min read
Analyze what customers actually say
Six ways to slice review data, the five patterns that matter, and how one theme becomes one operational fix per month.
Once your history is imported, resist the urge to stare at the average rating. The value is in the text. Slice the data six ways — each slice answers a different question.
Categorize before you analyze
| Slice | Question it answers |
|---|---|
| By source | Where does my reputation actually live? Where are the gaps? |
| By date | Is this trending up or down? What changed when it turned? |
| By rating | What separates my 5-star and 2-star experiences? |
| By location | Is one site dragging the brand down? |
| By product/service | Which offering generates praise? Which generates complaints? |
| By sentiment/theme | What words do customers repeat — good and bad? |
Sentiment scoring helps here: it separates a polite 3-star from an angry 3-star, something the star count alone can't do.
The five patterns that matter
- Recurring complaints — any negative theme in 3+ reviews is operational, not bad luck. Name it precisely: "callback delays after quotes," not "communication."
- Recurring praise — your marketing copy is hiding here. Customers' exact phrases become headlines.
- Reputation risks — mentions of safety, billing disputes, legal threats, or health issues escalate same-day, regardless of star count.
- Service gaps — requests for things you don't offer ("wish they did weekends") are demand signals.
- Competitive advantage — run the same theme analysis on 2–3 competitors' public reviews. Their recurring complaints are your positioning openings.
Trend detection, monthly
Watch: review velocity (new reviews/month), rolling average rating, sentiment mix, response rate, response time. Falling velocity with a stable rating usually means your request process stalled. A falling rating with stable velocity means an operational problem shipped.
Worked example
A plumbing company imports 84 historical Google reviews. Filtering 1–3 star reviews surfaces "arrival window" in 9 of 14. Fix: the booking confirmation SMS now states a 2-hour window plus a 30-minutes-out text. Sixty days later the theme has disappeared from new reviews and the average rating on new reviews is up. The insight cost nothing — it was sitting in already-public data.
That's the loop this whole course builds toward: one named theme → one operational fix per month → measure whether the theme fades.