Star rating vs review sentiment: why the average lags

Two signals, one review

Every review carries two signals: a star rating and the text. Rating vs sentiment is not a contest; they answer different questions. The rating says how satisfied the customer was overall. The text says what they noticed and how they felt about each part. Most brands track the first closely and the second barely at all.

Why the average moves slowly

Google defines a place's review score as the average of all ratings published for it, and says the score can take up to two weeks to update after a new review. Averages are heavy. A store with 1,000 ratings averaging 4.5 that suddenly receives 50 one-star ratings in a month drops to about 4.33. That is a serious problem for the store, and a change of less than a fifth of a star.

• The more reviews a store has, the less each new one moves the average.
• Old reviews keep counting, long after the experience they describe has changed.
• A problem has to affect many customers before it shows clearly in the rating.

When stars and text disagree

• "Five stars for the staff, but please fix the parking." Positive rating, negative mention of parking.
• "Lovely designs, rude cashier, three stars." A middle rating hiding one strong positive and one strong negative.
• "Never coming back!!" with four stars, which happens: people tap the wrong number.
• A rating with no text, which tells you nothing about why.

Counting the rating alone means treating the first review as pure good news and the second as average, when both contain a specific, fixable complaint.

Why the text shows problems sooner

A new problem first appears as a handful of mentions in reviews that are otherwise positive. Customers still like the store, so their ratings stay high, but they mention the new queue or the out-of-stock sizes. Measured as negative share for that topic, the change is visible as soon as enough mentions accumulate, often weeks before it would move the average.

Using both well

• Report the rating as the headline, with the review count and the recent-window average beside it.
• Report topic negative share as the diagnosis: which parts of the experience are getting better or worse.
• When the rating moves, look at which topics moved with it.
• When a topic moves and the rating has not, act anyway; the rating will follow.

An example

A restaurant's rating holds at 4.4 for three months. In the same period, the share of mentions about delivery that are negative rises from 18% to 34%, mostly in four- and five-star reviews that praise the food and complain about late delivery. By the time the rating falls to 4.3, the delivery problem has been visible in the text for two months. Watching topic sentiment would have given the team those two months to fix it.

Pitfalls

• Comparing ratings across stores with very different review counts.
• Using all-time ratings to judge recent changes.
• Treating rating-only reviews as having no information; they still count in the average.
• Using one sentiment score per review, which hides mixed reviews just as the rating does.

Common questions

Checklist

• Report the all-time and the 90-day rating, with review counts.
• Track negative share for each topic monthly.
• When the rating moves, find which topics moved.
• When a topic moves, act before the rating does.
• Keep quotes beside every number.

Doing this with Falcon

Falcon measures sentiment for every theme a review mentions, not one score per review. Each mention keeps the exact words that support its label, negative share is reported by topic and month with small samples flagged, and unclear texts are kept visible rather than forced into a category. Every number opens the reviews behind it.