What's the difference between a product rating of 3.5 displayed with stars versus Arabic numerals? It might very well be the difference between a 4 and a 3 rating in the eyes of the consumer, according to new marketing research from the Cornell SC Johnson College of Business.
In six experiments, the researchers found that consumers tend to overestimate fractional star ratings and underestimate fractional Arabic numerals. In either case, the ratings can be misleading, potentially causing a company to unknowingly overpromise and underdeliver - or sell its own product short.
"Overestimating Stars, Underestimating Numbers: The Hidden Impact of Rating Formats," was published May 15 in the Journal of Marketing Research. The team is Deepak Sirwani, M.S. '23, Ph.D. '24; Manoj Thomas, the Demir Sabanci Professor of Management and Marketing; and Srishti Kumar of Indian Institute of Management Ahmedabad.
"When these pictures communicate a fractional number - say 3.5, where there are three full stars and one-half star - our brain automatically completes this half picture," said Sirwani, now an assistant professor in the Marketing and Behavioural Science Division at the University of British Columbia. "But when the same rating is communicated using numbers, we focus on the left digit, which is '3,' and that's why 3.5 feels more like a 3 than a 4. That's why we underestimate it."
This work, Thomas said, is really about the human brain and how it processes images and numerals.
"Our results suggest that the brain representations that are activated when you process stars are completely different from the brain representations that are activated when you process Arabic numerals. That realization was the a-ha! moment," Thomas said. "Most people don't realize that, but it's a huge difference."
The researchers found consistent results across all six experiments they ran. In study 1, for example, they sought to measure the perceived accuracy of ratings using stars versus numerals. A total of 616 participants were randomly assigned to one of three experiment conditions based on the type of rating symbol used: stars; Arabic numerals; or both.
Participants were presented with 17 numeric ratings, from 1 to 5, in increments of 0.25, one at a time and in random order, and were asked to estimate the position of each rating on an unmarked horizontal line, with endpoints 1 and 5.
The participants consistently overestimated the magnitude of fractional star ratings while underestimating the magnitude of Arabic numeral ratings.
The same phenomenon occurred in study 3, which sought to confirm their hypothesis that people round up fractional star ratings and round down Arabic numerals. A total of 321 participants were randomly assigned to either the star or numeral group, shown four half ratings (1.5, 2.5, 3.5 and 4.5), in random order, and asked to indicate how they would approximate these ratings in round numbers.
However, for those in the star group, they were shown a total of five stars, but with stars colored in based on the rating. A rating of 2.5, for example, would show two completely colored-in stars, a third only half-colored in, and then two blank stars.
Participants could choose one of five responses - "around one," "around two," and so on. The results were striking: Participants rounded up 79% of the half-star ratings, but rounded down 71% of the Arabic numerals. The research hypothesizes that the "visual-completion effect" is at play, where when a viewer sees an incomplete image, they will tend to subconsciously fill the image in.
One answer, the researchers said: using visually complete stars - that is, a filled-in half-star without the blank half.
The group said their findings demonstrate that prevalent rating formats are misleading, highlighting the need for new industry standards.
"What makes this research powerful is the importance of ratings in the current marketplace," Sirwani said. "Most of us do not buy anything without checking its rating nowadays, and rating has become as powerful a predictor of purchase as price, or brand or even recommendations from friends and family."
Other research, he said, has indicated that even a rating jump of 0.2 points can increase sales by up to 300%. "We are showing," he said, "that sales could potentially increase by orders of magnitude by just showing stars instead of numbers."
