Mammalian brain
Social Proof Examples: Why 12,000 Reviews Can Still Steer You Wrong
Copying the crowd is a sensible shortcut. It breaks when the crowd is copying each other, or was paid to show up.
Social proof is treating what other people do as evidence of what you should do, especially when you're unsure. Star ratings, “bestseller” badges and long lines all work this way. In a hotel field experiment, a towel card saying most guests reuse their towels got more people to reuse theirs than a standard environmental appeal (Goldstein et al., 2008).
Key takeaways
- Social proof is a shortcut, not a mistake. It goes wrong when the crowd isn't independent or isn't real.
- Early popularity snowballs: one random upvote raised a comment's final score by about a quarter.
- Fake reviews are common enough to plan for, and they cluster on products that need the help.
- Read the 3-star reviews, check the dates, and compare products with the counts covered.
Two air fryers sit side by side in your search results. One is $79.99 with 4.6 stars from 12,408 ratings and an “Amazon's Choice” badge. The other is $64.99 with 4.7 stars from 212 ratings. You click the first one before reading a single spec, because twelve thousand people feels like proof. That's social proof, and most of the time it's a fair bet. The trouble is that you can't see how those twelve thousand ratings got there.
Social proof examples from everyday life
| The situation | What the crowd says | Ask this instead |
|---|---|---|
| An air fryer with 12,408 ratings next to one with 212 | “Twelve thousand people can't be wrong.” | “Do the 3-star reviews mention the problem I'd care about?” |
| A $38 serum with a “#1 Bestseller” badge | “It's the bestseller, so it must work.” | “Bestseller in which category, and since when?” |
| A ramen place with a 40-minute line next to an empty one | “The line proves it's better.” | “Did the line start because of the food, or because there was already a line?” |
| A hotel listing that says “23 people are looking at this right now” | “Everyone wants this one. Book it.” | “Would I pick this hotel from the photos, location and price alone?” |
| A new phone charger with 300 five-star reviews posted in one week | “All these people love it.” | “Are the reviews detailed, spread over months, and from verified buyers?” |
Each row has a real signal inside it. The right-hand question checks whether that signal came from independent, honest people who wanted what you want.
What the research shows
A sign about other guests changes behavior. Noah Goldstein, Robert Cialdini and Vladas Griskevicius tested towel cards in hotel bathrooms. In the first field experiment, a card saying most guests reuse their towels raised reuse from 35.1% to 44.1% compared with the standard “help the environment” card. In the second, the most effective version described guests who had stayed in the same room (Goldstein et al., 2008). The closer the crowd is to your situation, the harder it pulls.
Popularity can be partly luck. Matthew Salganik, Peter Dodds and Duncan Watts built an online music market where 14,341 people could listen to and download 48 songs by unknown bands. Some people saw how often each song had been downloaded, split into eight separate “worlds,” and others saw no counts. Showing the counts made hits bigger and much less predictable. The best songs rarely did badly and the worst rarely did well, but almost anything else could happen (Salganik et al., 2006).
One early vote snowballs. Lev Muchnik, Sinan Aral and Sean Taylor gave more than 100,000 comments on a news site a random first upvote, a random downvote, or nothing. A fake upvote made later positive ratings 32% more likely, and those comments ended with scores about 25% higher on average. Fake downvotes tended to get corrected by later readers (Muchnik et al., 2013). Praise compounds more easily than criticism.
Some of the crowd isn't real. Michael Luca and Georgios Zervas found that Yelp's filter flagged about 16% of Boston restaurant reviews as suspicious. Restaurants were more likely to post fake praise when they had few reviews or a weak reputation, and competition went with more fake negative reviews of rivals (Luca & Zervas, 2016). The filter isn't perfect, so the true share could be higher or lower.
Bought reviews fade, and so do the products. Sherry He, Brett Hollenbeck and Davide Proserpio tracked Amazon sellers who recruited fake reviewers through private Facebook groups. Ratings and sales rank rose for a few weeks. After sellers stopped buying, ratings fell and the share of one-star reviews went up, a sign that the products relying on fake reviews were often low quality (He et al., 2022).
Which brain is doing this
In Decision IQ's six-brain model, social proof belongs to the mammalian brain: social gravity beats spreadsheets. For most of history, doing what the group did was a cheap way to find the safe berry and the good waterhole, so a big number of other people reads as safety. The reptilian brain chips in when the crowd comes with a clock, like “23 people are looking.”
The mammalian brain can't tell a crowd of independent buyers from a crowd that copied the first upvote or was paid in free products. The steps below hide the crowd for a moment and make you check where it came from, so the choice moves to the part of you that reads specs and prices.
How to read the crowd without being herded: a 4-step check
- 1
Sort by newest and read five 3-star reviews. Change the sort to most recent, then filter to 3 stars. Those reviewers liked the product enough to keep it and still tell you what went wrong.
- 2
Cover the count and the badge. Put your thumb over the rating count and any badge. Compare the two products on the two features you wrote down before you started searching.
- 3
Check the review dates. Scroll the five-star reviews and look at when they were posted. If dozens landed in the same week with one-line praise, treat the rating as unknown.
- 4
Text one person who owns it. Send one message to a friend who bought it: “Would you buy it again?” One honest owner beats a thousand strangers you can't check.
If the popular one still wins after all four, buy it with a clear conscience.
Frequently asked questions
What are some examples of social proof?
Star ratings and review counts, “bestseller” and “most popular” badges, “X people bought this today” notices, long lines outside restaurants, and friends recommending a product. Each one uses other people's choices as a hint about what's good.
Is social proof always bad?
No. When many independent people with needs like yours choose something, that's useful information. It misleads when people are copying each other, when the crowd is very different from you, or when the reviews are fake.
What's the difference between social proof and the bandwagon effect?
Social proof is using others' behavior as information when you're unsure what's right. The bandwagon effect is joining in because something is popular and growing, whether or not you're unsure. In practice they overlap, and both get stronger as the numbers climb.
How can I tell if online reviews are fake?
Look for many short five-star reviews posted in a burst, reviewers with no other history, and a big gap between the five-star and three-star stories. Read the most recent reviews, since products that bought reviews often slide afterwards.
Research notes
Sources
- 1A room with a viewpoint: using social norms to motivate environmental conservation in hotels Goldstein, Cialdini & Griskevicius, Journal of Consumer Research 35(3), 472–482 (2008)
- 2Experimental study of inequality and unpredictability in an artificial cultural market Salganik, Dodds & Watts, Science 311(5762), 854–856 (2006)
- 3Social influence bias: a randomized experiment Muchnik, Aral & Taylor, Science 341(6146), 647–651 (2013)
- 4Fake it till you make it: reputation, competition, and Yelp review fraud Luca & Zervas, Management Science 62(12), 3412–3427 (2016)
- 5The market for fake reviews He, Hollenbeck & Proserpio, Marketing Science 41(5), 896–921 (2022)
Read next
One decision idea a week
One useful idea. No daily drip, no pressure.