The hidden bias in online ratings

Imagine you’re shopping online for headphones. One pair shows a 3.5-star rating, the other a 3.5 out of 5 numerical rating. You’d think they mean exactly the same thing. But according to new research, our brains interpret these two ratings very differently.

A team of researchers, Deepak Sirwani from the University of British Columbia, Srishti Kumar from the Indian Institute of Management Ahmedabad, and Manoj Thomas from Cornell University, have uncovered a surprising psychological quirk: star ratings and number ratings systematically fool us in opposite directions.

Ratings drive the internet economy. A small bump in product ratings, say, from 3.3 to 3.5, can dramatically boost sales. Studies suggest that even a 0.2 increase in average ratings can increase sales by 30% to 200%. That means whether a platform uses stars or numbers can subtly sway what millions of consumers decide to buy, where they choose to eat, which movie they stream, or even whom they swipe right on. And it’s not only about products. Ratings shape how we view restaurants (Yelp), travel destinations (TripAdvisor), drivers (Uber), hosts (Airbnb), and even people on dating apps. If those ratings are consistently skewed, the playing field isn’t level.

Across 12 carefully designed experiments, the researchers found that star ratings get overestimated. When you see 3.5 stars, your brain tends to treat it as closer to 4 stars. Why? Because of something called the visual-completion effect. Our visual system doesn’t like incomplete images. When you see a half-filled star, your mind instinctively fills it in, nudging your judgment upward. On the other hand, Numerical ratings get underestimated. So, when you read “3.5,” your brain latches onto the “3” on the left (the so-called left-digit effect) and drags the number downward. So, 3.5 feels more like a 3 than a 4.

The result is that neither system gives us a perfectly accurate impression. Stars make things seem better than they are; numbers make them seem worse.

Here’s another strange aspect the researchers discovered: the way stars are filled matters. Platforms use two main methods for fractional stars: proportion-of-area fill, where the star is shaded based on area (25% of the star filled in for a quarter), and proportion-of-width fill, where the star is shaded by slicing it vertically (25% width filled for a quarter). These look different, and people perceive them differently. In some cases, one method makes ratings look higher than they really are, amplifying the overestimation effect.

The team suggests that using “visually complete” stars, symbols that don’t leave half-empty outlines, can reduce the bias. It’s a small design tweak that could make ratings much fairer.

But right now, there’s no rulebook. Amazon uses stars. Uber sticks with numbers. Google Maps shows both. This patchwork of formats means consumers are constantly being nudged, sometimes upward, sometimes downward, without realizing it. Although the consistent use of one system within the same platform evens things out, browsing for ratings on two websites using different systems can lead to inconsistencies. So, the researchers argue it’s time for standardization. Just like food labels or credit scores, rating systems are too important to be left inconsistent. A shared standard could help consumers make better, fairer decisions.

What I love about this study is that it shows how deeply human characteristics shape digital life. Our brains evolved to handle whole numbers and simple images, not fractional decimals or half-filled icons. So when we face those fractions, 3.25, 3.5, 3.75, we reach for shortcuts. Stars push us up. Numbers pull us down. Until platforms fix the issue, here’s some practical advice. If you’re on a site that uses stars, remember the rating may feel a little rosier than it really is. If you’re on a site that uses numbers, be aware that your brain might be shortchanging it. And if you’re comparing across platforms, be extra cautious, because a 3.5 isn’t always the same 3.5.

If you want to learn more, read the original article titled "Overestimating Stars, Underestimating Numbers: The Hidden Impact of Rating Formats" on Journal of Marketing Research at http://dx.doi.org/10.1177/00222437251322425.