4.6 Stars vs 4.4 Stars: Is the Difference Real?

You are choosing between two restaurants, hotels or apps. One shows 4.6 stars, the other 4.4. Most people pick the higher number without a second thought. Businesses care even more, since ratings can move real money. A famous study of Seattle restaurants found that a one-star increase in rating led to a 5-9% rise in revenue (Luca). So a small difference deserves a careful look.

An average with few reviews is shaky

Take Business A with 25 reviews averaging 4.6 stars (standard deviation 0.8), and Business B with 400 reviews averaging 4.4 (standard deviation 0.9). The confidence interval calculator for unknown variance puts A's true average between 4.27 and 4.93, and B's between 4.31 and 4.49. The intervals overlap heavily, and A's is almost four times wider.

Testing the gap

The independent samples t-test calculator compares the two means from their summary figures. A Welch version gives p ≈ 0.24, and a pooled version about 0.28. Either way, a 0.2-star gap is well within what chance could produce with only 25 reviews on one side.

The lesson is that the number of reviews matters as much as the average. The same 4.6 based on 2,500 reviews would be far more convincing.

Ratings are not normally distributed

Star ratings are bounded and often piled at 5 and 1. When the shape is lopsided, the Mann-Whitney test compares the two sets of ratings by ranks and does not rely on normality. With raw data, it is a useful check on the t-test.

Another option treats each review as positive or negative. Evan Miller's well-known article on ranking by ratings recommends using the lower bound of a confidence interval for the share of positive reviews, so items with few reviews are not ranked above items with thousands (Miller). The confidence interval calculator for a single proportion computes such intervals.

What else can mislead

Reviews come from people who chose to write them, so ratings overrepresent the very happy and the very angry. Fake or incentivized reviews, review bombing and rounding on the display all add distortion. A 4.55 rounds to 4.6 on screen while a 4.44 shows as 4.4, so the displayed gap of 0.2 can hide a true gap of 0.1.

Rounding and thresholds

Platforms display rounded averages, and the Yelp study used exactly this to identify the effect of ratings, comparing businesses just above a rounding threshold with those just below. It is a clever design, and it also means that a tiny change in the true average, such as 4.49 to 4.51, can change what customers see and how they act. The study also found that how consumers respond depends on the number of reviews behind the rating.

For a business owner, the practical advice is to collect reviews steadily rather than chase a decimal.

For a shopper, a high rating with few reviews is a promise not yet tested.

Finally, compare like with like. Ratings vary by category and platform, so a 4.4 for a budget hotel and a 4.4 for a luxury one are not the same signal.

Consumers can use the same logic. Prefer a rating with many reviews, read the written comments, and treat gaps under a few tenths of a star as ties. Businesses should avoid reacting to small week-to-week changes.