--- title: "How to Analyze Customer Reviews to Find What to Fix" description: "A spreadsheet method for analyzing customer reviews: collect, clean, tag themes and sentiment, compare locations and periods, assign owners and re-measure." url: https://brandprotecto.com/blog/how-to-analyze-customer-reviews updated: 2026-10-28 publisher: BRANDprotecto (info@brandprotecto.com) --- # How to Analyze Customer Reviews to Find What to Fix [Home](https://brandprotecto.com) / [Blog](https://brandprotecto.com/blog) / Multi-location Multi-location # How to Analyze Customer Reviews to Find What to Fix Star ratings tell you how customers feel. Themes tell you why. A step-by-step method to turn reviews into a short list of fixes with owners and measures. By the BRANDprotecto editorial team · Published October 28, 2026 · Last reviewed October 28, 2026 · 7 min read ## In short - Start with a specific question, then collect reviews from all platforms into one table with the same columns. - Tag each review with themes from a short written codebook and record sentiment per theme, not per review. - Separate one-off incidents from patterns by checking dates, spread across locations and persistence over time. - Every recurring negative theme needs an owner, an action, a start date and a measure. - AI can speed up tagging and translation, but a person should check samples, sensitive items and all decisions. To analyze customer reviews, collect them in one table, tag each review with the topics it mentions, such as waiting time, staff, price, quality or cleanliness, and record whether each topic was praised or criticized. Then compare topics by location and period. Recurring negative themes become fixes with a named owner, and you re-measure after the fix to see whether complaints about that theme fall. ## Start with a question Review analysis is quicker and more useful when it answers a specific question. For example: why did ratings at one location fall this quarter, what do customers mention most since a price change, or which problems appear at every location rather than at one. The question decides the time window, the locations and the platforms you include. You do not need special software to begin. A spreadsheet and a written list of themes are enough for a few hundred reviews. Tools and AI help with volume, but the method stays the same. ## Step 1: Collect reviews in one place Put every review into one table with the same columns: platform, location, date, star rating, review text, language, whether you replied and the reply date. - **Google:** for small volumes, read reviews in Business Profile, as described in Google's help page on [managing customer reviews](https://support.google.com/business/answer/3474050?hl=en) . For larger volumes, the Business Profile API can [list all reviews for a location and fetch reviews for several locations in one request](https://developers.google.com/my-business/content/review-data) , including the star rating, comment, creation time and any reply. Google notes that listing reviews this way is useful for auditing them in bulk. Access is not automatic: you register an application with OAuth credentials and submit an access request, and Google's [API prerequisites](https://developers.google.com/my-business/content/prereqs) include managing a Business Profile that has been verified and active for 60 or more days. If you would rather use a tool, our guide on [choosing a brand monitoring tool](https://brandprotecto.com/blog/choosing-brand-monitoring-tool) lists questions to ask about where its data comes from. - **Other platforms:** export options differ by platform and account type. Where there is no export, copy recent reviews manually and follow each platform's terms of use. Record the date you collected the data. Reviews can change, and Google points out that customers can still change their review after reading your reply, so a snapshot is only accurate for the day it was taken. ## Step 2: Clean the data - Remove duplicates, such as the same review collected twice. - Keep rating-only reviews for rating calculations, but set them aside for theme analysis, because there is no text to tag. The Business Profile API even has an option to ignore rating-only reviews when fetching reviews across locations. - Use one spelling for each location name. - Translate reviews written in other languages, or tag them in the original language with the same codebook. - Leave out reviewer names and anything else you do not need. The analysis is about topics, not people. If you combine platforms, keep the platform column. Ratings are not always calculated the same way. Google's page on [review scores](https://support.google.com/business/answer/4801187?hl=en) says the score is the average of all ratings published on Google for that place or business, while Yelp says [only the star ratings of recommended reviews](https://www.yelp-support.com/article/Do-reviews-that-arent-currently-recommended-impact-the-businesss-star-rating?l=en_US) are used to calculate a business's overall rating. Compare like with like. ## Step 3: Build a theme codebook A codebook is a short list of themes with a one-line definition each, so that two people tagging the same review pick the same theme. Start with a list like the one below and adapt it to your business. Theme What counts Usual operational owner Speed and waiting Queues, delays, slow replies, late delivery Operations or location manager Staff attitude Friendliness, rudeness, helpfulness Location manager, training Product or service quality The core thing the customer paid for Product, kitchen or service lead Price and value Price level, unexpected charges, value for money Commercial or finance Cleanliness and facilities Hygiene, restrooms, parking, accessibility Facilities Booking and ordering Online booking, phone lines, order errors Digital or customer service Billing and refunds Wrong charges, refund delays Finance or customer service Communication and follow-up Promises not kept, no call back Customer service Keep the list short enough to remember, allow several themes per review, and add an Other theme. Read the Other bucket every month; if a new topic keeps landing there, give it its own theme and a definition. ## Step 4: Tag sentiment by theme, not by review A star rating summarizes a whole visit. Themes show what drove it. Tag each theme in a review as positive, negative or mixed. **Example (illustrative):** A four-star review says: Great coffee, but we waited a long time to order. Tag it as Product quality: positive and Speed and waiting: negative. Read only by its stars, the review looks like good news. Read by theme, it is an early warning about waiting. Add one more column for severity. A safety, hygiene or discrimination complaint should be flagged even if it appears only once, because it needs a different response from a comment about the music being too loud. ## Step 5: Count, compare and separate one-offs from patterns With every review tagged, count three things for each theme, location and month: how many reviews mention the theme, how many of those mentions are negative, and how both numbers compare with the previous period. **Worked example (illustrative arithmetic):** a location receives 120 reviews in a quarter. Thirty of them mention waiting, and 24 of those 30 mentions are negative. Waiting therefore appears in one review in four, and four out of five mentions of it are negative. If the next quarter shows 12 negative mentions out of 20, the theme is shrinking and the negative share has dropped from four in five to three in five. Before you call something a pattern, check: - **Dates:** several complaints on the same day may describe a single incident, such as an outage. - **Spread:** a theme at one location points to local causes; the same theme at many locations points to a process, product or policy. - **Persistence:** a theme that returns month after month is a pattern even when the counts are small. - **Suspicious reviews:** reviews that look like spam or a conflict of interest belong on a separate list. Google's page on [reporting inappropriate reviews](https://support.google.com/business/answer/4596773?hl=en) explains how to flag them and says not to report a review just because you disagree with it. You can also set review trends next to Google's [Business Profile performance data](https://support.google.com/business/answer/9918094?hl=en) , which covers views, searches and interactions such as calls, direction requests and website clicks. Treat any link between the two as a hypothesis to test, not as proof. ## Step 6: Turn findings into owned actions Analysis pays off only when someone changes something. For each recurring negative theme, write down: - The finding in plain words, with two or three example reviews. - The owner who can change it, usually the operational owner from your codebook. - The action and the date it starts. - The measure you will check, for example negative mentions of the theme per month. Positive themes deserve a line too. If customers repeatedly praise a particular service, that tells you what to protect when you change processes. This [sample monthly reputation report with demo data](https://brandprotecto.com/sample-report) shows one way to present themes next to ratings for management. ## Step 7: Re-measure Use the same codebook and the same period length after the fix. Look at the theme first and the rating second, because an average moves slowly when a location already has many reviews. Allow for delays as well: Google's page on review scores says that after someone leaves a new review, it may take up to 2 weeks for the score to update. If the theme does not shrink, revisit the cause rather than the wording of your replies. For a wider check that also covers profiles, replies and search results, use our [online reputation audit checklist](https://brandprotecto.com/blog/online-reputation-audit-checklist) . If you manage many sites, our [multi-location review management playbook](https://brandprotecto.com/blog/multi-location-review-management) shows how to turn this analysis into a monthly routine. ## Where AI helps and where people must check AI tools can speed up the slow parts: a first pass of theme tagging, translation, grouping similar comments and drafting summaries. They can also make confident mistakes. Sarcasm, mixed reviews, local slang and reviews that describe two different visits are easy to mis-tag. - Have a person check a random sample of AI tags every month, and refine the codebook when the same error repeats. - Have a person review every severe or sensitive item, such as health, safety, discrimination or a named employee. - Do not paste customer records or other personal data into a tool unless your data policy allows it. - Keep decisions about actions, and about any public reply, with people. Our guide to [AI review responses](https://brandprotecto.com/blog/ai-review-responses) covers drafting replies safely. ## Frequently asked questions ### What is the simplest way to analyze customer reviews? Use a spreadsheet. Put every review in a row with the platform, location, date, rating and text. Add columns for themes such as waiting, staff, quality and price, and mark each theme as positive, negative or mixed. Count negative mentions per theme each month, compare locations and periods, and assign an owner to every theme that keeps coming back. ### How many themes should a review codebook have? There is no fixed number. Keep the list short enough that people remember the definitions without looking them up, allow several themes per review, and include an Other theme. Read the Other bucket every month and promote a topic to its own theme when it keeps appearing. Consistent definitions matter more than the exact number of themes. ### Can I export all my Google reviews for analysis? Google's Business Profile API can list all reviews for a location and fetch reviews for several locations in one request, including the rating, text, date and reply. Access needs OAuth credentials, an approved access request and, per Google's prerequisites, a Business Profile verified and active for 60 or more days, so you will need a developer or a tool built on the API. For small volumes, copying reviews from Business Profile into a sheet also works. ### Should I analyze the star rating or the review text? Both, for different purposes. The star rating shows the overall direction and is what customers see first. The text explains why. A four-star review can still contain a clear complaint, and a one-star review can praise part of the visit. Tagging themes in the text is what turns reviews into specific fixes. ## Sources - [Manage customer reviews — Google Business Profile Help](https://support.google.com/business/answer/3474050?hl=en) (checked October 28, 2026) - [Work with review data — Google Business Profile APIs, Google for Developers](https://developers.google.com/my-business/content/review-data) (checked October 28, 2026) - [Prerequisites — Google Business Profile APIs, Google for Developers](https://developers.google.com/my-business/content/prereqs) (checked October 28, 2026) - [Understand review scores for local places & businesses — Google Business Profile Help](https://support.google.com/business/answer/4801187?hl=en) (checked October 28, 2026) - [Do reviews that aren't currently recommended impact the business's star rating? — Yelp Support Center](https://www.yelp-support.com/article/Do-reviews-that-arent-currently-recommended-impact-the-businesss-star-rating?l=en_US) (checked October 28, 2026) - [Report inappropriate reviews on your Business Profile — Google Business Profile Help](https://support.google.com/business/answer/4596773?hl=en) (checked October 28, 2026) - [Understand your Business Profile performance & insights — Google Business Profile Help](https://support.google.com/business/answer/9918094?hl=en) (checked October 28, 2026) Prepared by the BRANDprotecto editorial team with AI assistance and checked against the sources above. Platform rules and laws change: always confirm the current wording at the source. This article is general information, not legal advice. ## See how your business looks online Request a free reputation audit. We look at your Google Business Profile and your Google reviews and email you the summary with your Brand Health Score and the main problem areas within one business day. [Start free audit](https://brandprotecto.com/audit/start) ## Related articles [Multi-location A Review Response Policy for Location Managers: Who Replies, Who Escalates A written template with three tiers that tells location managers which reviews they can answer, which go to the central team, and which escalate immediately. November 5, 2026 · 8 min read](https://brandprotecto.com/blog/employee-social-media-review-policy) [Multi-location Multi-Location Review Management: A Playbook for Central and Local Teams A step-by-step playbook for brands with many locations: who owns which profile, who replies, what gets escalated, and which numbers to review every month. October 28, 2026 · 7 min read](https://brandprotecto.com/blog/multi-location-review-management) [Multi-location Franchise Reputation Management: Who Owns the Profiles and the Reviews? Nobody owns the reviews, but someone has to own the profiles, the replies and the fixes. A practical split of responsibilities between franchisor and franchisee. October 28, 2026 · 8 min read](https://brandprotecto.com/blog/franchise-reputation-management)