PRACTICAL GUIDE · 7 MIN READ
How to analyse Google reviews and find recurring complaints
The practical answer
To analyse Google reviews, choose a consistent date range, tag each review by topic and sentiment, count how often each issue appears, then prioritise problems your team can actually fix. Keep the original wording nearby so a summary never replaces the customer’s meaning.
Start with a question your team can act on
An average star rating tells you whether reviewers are broadly satisfied. It rarely tells you what to change on Monday. A useful analysis starts with a narrower question: “What repeatedly makes arrival difficult?” or “Which part of our service do customers praise?”
Begin with the most recent four weeks. If there are too few reviews to identify a pattern, expand to three months and state the period clearly. Compare periods of similar length; ten complaints across 200 reviews means something different from ten across 20.
Separate the overall review from the individual topics
One review can contain several signals. “The cleaner was excellent, but the team arrived late and nobody called” includes positive feedback about service quality and negative feedback about arrival time and communication. Calling the entire review “negative” loses the strength your team should preserve.
- Sentiment: positive, neutral or negative for a specific statement. Keep mixed or unclear feedback separate when a single label would be misleading.
- Topic: the subject, such as arrival time, communication, pricing clarity or workmanship.
- Request: a concrete change the reviewer asks for, such as an arrival window or advance notice.
- Evidence: a short excerpt that explains the tag, with a link to the original review.
Use a small, consistent topic list. Combine “late arrival” and “poor punctuality” when they describe the same problem. Keep “appointment cancelled” separate: the operational fix may be different. Manually check ambiguous wording, sarcasm and reviews written in languages your team does not understand confidently.
A worked example: 40 reviews, one month
The following is fictional and illustrates the method. It is not a Kīrtelo customer result. Count a topic once per review, even when the reviewer repeats it.
| Theme | Reviews mentioning it | Share of all reviews | Interpretation |
|---|---|---|---|
| Service quality praised | 18 | 45% | A strength to preserve |
| Late arrival complained about | 8 | 20% | Recurring operational problem |
| Updates missing | 6 | 15% | Communication gap |
| Price unclear | 3 | 7.5% | Investigate estimate wording |
These categories can overlap. A review mentioning both late arrival and missing updates belongs in both rows, so topic percentages need not add to 100%. The denominator is all 40 reviews, including those with no written text; record how many have text so readers know the limits of the analysis.
Turn mentions into a realistic weekly priority
The most mentioned complaint is a useful starting point, but frequency alone is not enough. Check the seriousness of the issue, whether it has increased, who can fix it and whether the evidence supports a specific action. A single safety concern can deserve faster attention than many comments about a minor inconvenience.
In this sample, late arrival and missing updates suggest a combined first step: send a realistic arrival window and notify customers promptly when it changes. The operations lead could trial that process for one week. The action is specific, has an owner and can be checked without waiting for a rating change.
- Problem: eight reviews mention late arrival; six mention missing updates.
- Evidence check: read the relevant reviews and confirm that the complaints describe recent appointments.
- Action: introduce an arrival-window message and a delay-notification rule.
- Owner and deadline: operations lead, trial starts this week.
- Check: audit whether messages were sent, then compare related review mentions over the next comparable period.
When a pattern needs a public reply, use our review response approval checklist. For teams comparing branches, follow the multi-location comparison method before deciding where to focus.
Use a simple analysis worksheet
You can start in a spreadsheet with these columns: review date, location, rating, original review link, topic, topic sentiment, evidence excerpt, suggested action, owner and status. For reviews with several themes, use one row per review-topic pair and a review ID to avoid double-counting unique reviews.
Keep access limited to the people who need it. Public reviews may still contain personal details; include only what the analysis needs. Employee mentions also require context: “Maya explained the process clearly” praises Maya, while “Maya booked the appointment, but delivery was late” does not establish that Maya caused the delay.
Watch for misleading conclusions
Reviewers are a self-selected sample, not everyone who bought from you. Low review volume, seasonal demand and changes in who leaves feedback can all distort comparisons. A fall in complaints after a process change is encouraging, but does not by itself prove the change caused it.
Pair review themes with operational evidence such as arrival records, support tickets or repeat-service requests. The aim is a better decision, not a dashboard full of percentages.
Common questions
Is review sentiment the same as a star rating?
No. Stars are the reviewer’s overall rating. Sentiment describes the tone of the written feedback, potentially for each topic. A five-star review can still contain a useful complaint.
How often should I analyse reviews?
For an active business, a weekly check for urgent issues and a monthly comparison of themes is a practical starting point. Lower-volume businesses may need a longer period to find reliable patterns.
Written by the Kīrtelo team. Workflows and examples are editorial suggestions, not Google requirements or verified product outcomes.