How to analyse Google reviews by branch and find the real problem

To analyse Google reviews across branches, give every review a sentiment from its star rating first and its wording second, tag it against a short fixed list of topics, then compare each topic's rate per 100 reviews at every branch with the rest of the group. A complaint that runs far above the group rate at one site is a local problem; one that runs at the same rate everywhere belongs to head office.

Your customers are already doing a rough version of this. BrightLocal's Local Consumer Review Survey 2026, a panel of 1,002 US adults, ranked "backed up by other reviews with similar sentiment" as the top factor in trusting a review, chosen by 56%. BrightLocal's reading is that people look for consistent themes, such as repeated praise for service or regular complaints about cleanliness. If a reader can spot the pattern on one branch's profile, you should see it first, across all of them.

Get every review into one table

Analysis starts with a single list where each row carries the branch it came from, the star rating, the text and the date. Copying from Google Maps works for a couple of sites. For more, use the Business Profile API, a scraper or a review tool, and keep a review ID so overlapping pulls can be de-duplicated. Our guide on how to export Google reviews to a spreadsheet compares those routes and the columns to keep.

Sentiment: star rating first, wording second

The star rating is the reviewer's own verdict, so let it decide wherever it is clear. A sensible rule is four and five stars positive, one and two negative, and three stars decided by the words. That keeps "Took ages, but honestly worth it" with five stars on the positive side, where the reviewer put it, and stops a scoring tool from reading "ages" as a complaint.

The wording only has to settle the middle. "Lovely staff, but I won't be back because of the parking" at three stars reads negative. "Fine, does the job" at three stars is neutral. Watch for negation when you read the text: "never felt rushed" is praise, even though "rushed" is a complaint word on its own.

Topics: a short list, each with an owner

Pick six to eight topics that match how the business is run, and give each one an owner who can fix it. For a chain of clinics or salons that might be:

• Waiting time, owned by branch managers and the rota.
• Booking and appointments, owned by whoever runs the booking system.
• Cleanliness, owned by facilities.
• Price and value, owned by head office.
• Staff and service, owned by branch managers.
• Parking and access, which is often nobody's, and worth knowing about anyway.

Write down the words that count for each topic before you start, so "queue", "waited" and "late" all land under waiting time, and two people tagging the same review get the same answer. A review can carry more than one topic. Keep the list fixed for at least a quarter, because a list that changes every month makes month-on-month comparisons unreliable.

Compare branches with rates, not counts

Raw counts mislead because big branches collect more of everything. Work in rates per 100 reviews over the same window. Here is a worked example for a group of eight branches over the last 90 days.

The group received 1,200 reviews, of which 132 were negative. Leeds received 150 of those reviews, and 30 were negative, which is 20%. The other seven branches received 1,050 reviews with 102 negative, which is just under 10%. So far you know Leeds is doing worse. Topics tell you why:

• Waiting time came up in 21 negative reviews at Leeds, 14 per 100 reviews. Across the other seven branches it came up in 19 negative reviews out of 1,050, about 1.8 per 100. Leeds runs at nearly eight times the rate of the other branches, so this is a Leeds problem.
• Booking complaints came up in 3 negative reviews at Leeds, 2.0 per 100, and in 22 elsewhere, 2.1 per 100. That is the same everywhere, so it belongs to the booking system, not to any branch.

The two findings go to different people. The Leeds manager gets the waiting-time reviews and a question about Saturday staffing. The booking complaints go to whoever owns the system, with the group-wide rate as evidence that retraining one branch will not fix it.

One caution on small numbers. A branch with 12 reviews in the window swings by 8 percentage points with every single negative review, so rates are only worth comparing once a branch has a reasonable count, say 30 or more in the window. Below that, read the reviews themselves.

Look at the last 90 days, not all time

Google's star score is the average of every rating ever published for the place, according to its help page on review scores, so an old branch with years of good reviews can hide a bad quarter. Customers look at recent reviews instead: BrightLocal found 74% seek reviews written in the last three months. Run the analysis on a rolling 90-day window, and keep the all-time figures only for context.

Read the names customers mention

Praise that names a team member is some of the most useful feedback a branch gets. Keep a roster of staff names per branch and match review text against it, rather than guessing at every capitalised word. Read the names customers volunteer, but do not ask for them: Google's content policy bars merchants from asking staff to solicit reviews that include "content that identifies a staff member", which we cover in Google's rule on staff-name review requests.

Turn the analysis into a monthly note

Each branch manager should get one short note a month: their sentiment split for the last 90 days next to the group's, their top two complaint topics with a few example reviews, the team members customers named, and any topic running well above the group rate. It fits alongside the weekly checks in our guide to managing Google reviews for multiple locations. The negative reviews in that note still need answers, and our guide to responding to negative Google reviews has examples by complaint type.

The praise side of the analysis has a second use. The things customers repeatedly thank your teams for can give an owner real material for LinkedIn. Braivex's Reachwyn turns a Story Bank and your writing samples into a 7-day LinkedIn calendar of full drafts in your voice, and each draft stays a draft until you approve it.

Running the analysis in Loculens

Loculens does this sorting on every branch's Google reviews. Sentiment works the way described above, star rating first and wording second, so a five-star rant and a three-star rave each land where they belong. You choose the topics, and deleting one removes it everywhere for everyone. Names are matched against your team roster and never guessed, and five-star reviews that name someone rise to the top. Counts, averages and filters sit beside each location, with time windows and custom date ranges, and every list exports to CSV.

Pricing is by locations and reviews: $39 a month for up to 5 locations, $99 for up to 25 and $299 for up to 100, in US dollars excluding tax. Start a 14-day Loculens trial with no card, sync your branches and see which topic is pulling your weakest site down.