Customer feedback analysis without running a survey
The reviews are already public. A fixed taxonomy and a spreadsheet can turn them into a useful operating list.
Lachlan Fea 8 min read
In this article8 sections
Customer feedback analysis turns what customers wrote into a theme, a count and a decision. For a service business, the raw material may already be sitting on public review profiles, whether a customer was prompted to leave a review or wrote one on their own. The aim is a theme with a number beside it and the reviews behind that number.
Verified 6 September 2026 against Google's help page on managing customer reviews, the Google Business Profile API's review data reference, and the Maps user contributed content policy.
Some customer-feedback guides begin with a survey. If a business already has a steady run of public reviews, it can first read that material in a consistent way and see what it says by location.
What customer feedback analysis has to produce
The problem usually appears at the end of the process. Someone reads a handful of reviews, writes "customers are frustrated with our booking process" in a deck, and nobody knows whether that means nine people or ninety, which site they visited, or whether the issue is new. A useful analysis produces four things.
| Output | The test it has to pass | What it looks like |
|---|---|---|
| Theme | Named in the words customers use, not the words in an org chart | "Cannot get through to reschedule" |
| Count | Carries the denominator it came from | 19 of the 71 reviews at that clinic |
| Evidence | Lets a reader open the number and read what it counted | Those 19 reviews, with dates |
| Comparison | Sets the number against another location or an earlier period | 27% this quarter, 7% last quarter |
A theme without a count is an anecdote. A count without a comparison is hard to interpret. A score alone cannot show which issue moved or where to start.
Keep the source row available when you report the number. A manager should be able to move from "19 booking mentions" to the matching reviews without asking someone to rebuild the analysis. That makes a disputed tag easy to inspect, and it prevents a tidy chart from becoming a claim nobody can check.
The feedback you already have
Reviews and surveys answer different questions. Public reviews are continuous and visible to the next customer. Surveys can ask a narrower question of a chosen group. Both are biased samples, so each count in this method stays next to its denominator.
The longer argument is in review-led versus survey-led reputation management. For published figures on how people use reviews, see our online review statistics roundup.
How to analyse customer reviews by hand
This is a spreadsheet method for a review window you can read consistently. It starts with the reviews themselves, then makes the comparison clear enough for someone to act on it.
1. Put the reviews into one sheet
Google's help page on managing customer reviews documents reading and replying. It does not document a review download control, so copying and pasting may be the practical option for one location.
The Google Business Profile API also exposes accounts.locations.reviews.list for one location and accounts.locations.batchGetReviews across several locations. Review software may offer its own export.
Use one row per review. Include date, location, review site, star rating, review text, and a column for each theme.
2. Set the taxonomy before you start reading
A taxonomy is the closed list of themes used for tagging. Write it before the first review. If the list changes with every new comment, the results will not be comparable.
Choose a small set of themes the business can change. For Coastline Physio, a fictional three-clinic group, the list could include booking and reschedule, wait time, treatment result, practitioner, reception and admin, price and billing, parking and access, facilities, and Other. Treat that as a starting point, then revise the list between reporting periods if the Other column is carrying a meaningful share of the comments.
Write a short definition beside each label. "Booking and reschedule" might include a failed online booking, an unanswered reschedule request, or an appointment that could not be changed. "Reception and admin" might cover the front desk only. The definitions do not need to be elaborate. They need to keep two people from using the same label for different problems.
3. Tag every mention
Put a 1 in each theme column a review mentions. A single review can mention more than one theme.
Tag the mention rather than the star rating. A five-star review that says "brilliant physio, shame about the parking" still contains a parking complaint. A separate complaint-or-praise column makes that distinction visible.
Use a tagging guide and keep it with the spreadsheet. A severity score can be useful only when the people applying it share a stable rubric; otherwise it introduces another moving part.
4. Count with the denominator
Sum each theme by location and for the group. Divide each location's theme count by that location's reviews in the chosen window, rather than by total theme mentions. That denominator stays stable when an unrelated theme moves.
Check that every review profile belonging to a location is in the denominator. A duplicate profile can split the review set, which is why a duplicate profile is a different problem from two locations with the same name.
5. Compare the same view twice
Run the table for the previous window as well. You can then compare each location with the group and with its own earlier result. Use the same window length and taxonomy. A January-to-March report and an April-to-June report can sit beside each other. A two-week report beside a quarter cannot tell the same story.
When you share the result, lead with the location and the fraction: "Milton Street: 19 booking mentions in 71 reviews." Add the earlier result beside it, then link the number to the matching reviews. That small reporting pattern keeps the discussion on the evidence instead of on a chart's colour or a broad label such as "poor customer experience". It also lets the manager ask a useful follow-up: which booking step appears in those 19 reviews, and does the same wording appear in last quarter's five?
A worked example
Coastline Physio is a fictional group with three clinics. In one quarter it received 214 reviews: 96 at Harbour Road, 71 at Milton Street and 47 at Westgate. The table records complaint mentions against the taxonomy above.
Before turning a theme into work, open a handful of the reviews behind it. Counts make a good starting point, but the wording tells you whether the same label is hiding different problems. Five booking mentions might all concern the online form. Or three might be about a delayed confirmation and two about a phone queue. Keep those differences in the notes column, then decide whether they need one operational change or two smaller fixes.
That check is especially useful when a theme rises sharply. A jump may be real, but it can also reflect one busy weekend, a new staff member or several reviews describing the same incident. The count still belongs in the report. The review text simply keeps the next action specific.
| Theme | Harbour Rd (96) | Milton St (71) | Westgate (47) | Group | Share of 214 |
|---|---|---|---|---|---|
| Booking and reschedule | 4 | 19 | 3 | 26 | 12.1% |
| Wait time | 7 | 6 | 2 | 15 | 7.0% |
| Parking and access | 11 | 1 | 0 | 12 | 5.6% |
| Price and billing | 3 | 4 | 2 | 9 | 4.2% |
| Reception and admin | 2 | 5 | 1 | 8 | 3.7% |
| Facilities | 1 | 2 | 3 | 6 | 2.8% |
| Treatment result | 2 | 2 | 1 | 5 | 2.3% |
The group figure says booking and reschedule accounts for 12.1% of the quarter's reviews. The location split changes the conversation: Milton Street has 19 mentions in 71 reviews, or 26.8%. Harbour Road has 4.2%; Westgate has 6.4%.
Milton Street had five booking complaints in 68 reviews in the previous quarter, or 7.4%. The clinic changed its online booking system in July. The change in the count does not prove the system caused the complaints. It gives a manager a specific place to investigate.
Harbour Road's parking issue is different. Eleven of its 96 reviews mention it, compared with one at Milton Street and none at Westgate. The clinic cannot change the street, but it can put clear parking directions in the booking confirmation.
The same example also shows why a lifetime rating can be a poor operating measure. As an illustrative weighted calculation, 552 reviews at 4.83 plus 71 new reviews at 4.52 equals about 4.7947, which rounds to 4.79. That is arithmetic for this example, not a claim about how Google calculates or displays profile ratings.
Where the manual method breaks
The limit is not a universal review count. It is whether someone can tag the chosen window using the same guide, then repeat the work when the next window closes. A three-location group with 214 reviews can use the example above as a planning exercise. A larger review flow may need a different process or software before the routine starts getting skipped.
Two checks matter as volume rises:
- Compare a sample of tags between periods. If "waited 25 minutes past my appointment" moves from wait time to reception and admin because the tagger changed, the trend may be a tagging change.
- Keep the reporting cadence realistic. Missing a period removes the comparison that gives the count context.
What to look for in customer feedback analytics software
Ask what every number represents. A useful screen can tell you how many reviews matched a theme and let you inspect the review set. Ask whether the result is reproducible for the same filters and whether the count is shown per location.
Look for the reviews behind the number. If a screen says 19, a manager should be able to open those 19 reviews and understand why they were included.
Ask how the tool handles one review with two issues. A review that praises a practitioner and complains about parking should be able to contribute to both themes. If the product forces every review into one bucket, the count may hide the issue that prompted the reader to open the report in the first place.
Location comparison matters too. The Coastline Physio example would have sent the wrong team looking for a fix if it had shown only a group total. The useful view puts one location beside another and beside its previous period.
Finally, check what the software turns into an action. A short list of themes, counts and underlying reviews is easier to work through than a screen that leaves the team to decide where to begin. Exporting the data can still be useful, but a manager should not need a fresh spreadsheet each month just to find the reviews behind a reported theme.
Cloutly's Ask answers a question about an account's own reviews with a computed count, shows how many reviews it searched, and links to the reviews behind the number.
Google's Maps user contributed content policy says businesses may not "selectively solicit positive reviews from customers".
Acting on what you find
The public response and the operational response are separate jobs.
Reply to the reviews behind a theme with the next reader in mind. For a parking finding, that may mean naming the nearby car park. For a booking problem, it may mean giving a direct number to call. Our review response examples cover the writing side.
Then choose one theme, name the change, record the starting share, and check the same share in the next reporting window. If Milton Street's booking share stays at 26.8% after a booking-flow change, the team has more work to do.
Reading reviews in a batch is the "Use" part of review management: the four jobs and who does them. Review monitoring covers the separate job of catching reviews as they arrive.
What to distrust
Sentiment scores. A sentiment score may be useful as one signal, but it does not replace a coded-theme count. Check whether it identifies the theme, denominator and location behind a change.
Lifetime star averages. The illustrative Milton Street calculation moves from 4.83 to 4.79 while 26.8% of its current-quarter reviews mention the same booking issue. A lifetime average may still be useful on a profile. For operations, use the chosen window and location.
Word clouds. A word cloud retains word-frequency counts, but it does not by itself tell you whether a word is praise or a complaint, or provide a coded-theme denominator. Use it as a prompt for reading, not as the report.
Frequently asked questions
What is customer feedback analysis?
Customer feedback analysis groups what customers wrote into a fixed set of themes, counts each theme against a stated denominator, and compares those counts by location and over time. The working output is a theme, a count and the reviews behind it.
How do you analyse customer reviews?
Put the reviews in one sheet, set a theme list before reading, tag every theme each review mentions, sum the columns by location, and divide by that location's review count. Run the same view for an earlier period so the result has context.
Is review data biased?
Yes. Public reviews come from customers who chose to write in public, whether prompted or spontaneous. Survey responses come from customers who chose to answer a form. Neither sample is random, so the method keeps the denominator beside the count.
What is the difference between customer feedback analysis and sentiment analysis?
Sentiment analysis classifies text by polarity. Customer feedback analysis asks what the text is about, how often it appears, and which location it concerns. Sentiment can be one field in a wider analysis.
Can you export your Google reviews?
Google's review-management help page does not document a review download control. The Google Business Profile API exposes reviews per location and in batches across locations.