It’s pretty clear exactly who writes fake positive reviews on the Web: friends or relatives of the author or the shop or restaurant owner, or sometimes the author or shop owner himself. The goal of fake positive reviews is to increase sales, and the reviewers are the ones who benefit, or want their friends to benefit.
But who writes fake negative reviews, denouncing stuff without any obvious reason? The usual assumption is that the perpetrators are competitors of some sort, hoping to get an edge on other novelists or chefs or innkeepers. But are there really so many nasty people in the world who need to get some slight advantage by tearing down the restaurant one block over? The question has been shrouded in mystery.
Until now. A fascinating new academic study sheds light on the fake negative review, finding not only that the source is totally unexpected but also that the problem is much bigger than a few malicious operators.
It turns out that competitors are not necessarily the ones giving one miserable star to products they did not buy or experiences they did not have. Customers do it â€" in fact, devoted customers.
This is hard to wrap your brain around, so first some background. The study was done by Eric Anderson of Northwestern University and Duncan Simester of the M.I.T. Sloan School of Management, using data from an unnamed apparel company that markets through catalogs, a few stores and a Web site. The company does not use third-party sellers and few of its products turn up on eBay, so it provided a relatively controlled experiment.
Registered customers wrote over 325,000 reviews in the study period. But for 16,000 of those reviews, there is no evidence that the customer bought the item. These reviews are on balance much more negative. (Could the items have been gifts, which could explain a higher level of dissatisfaction? No, the reviewers explicitly said they bought the items. The researchers were also able to rule out other possibilities, such as the negative reviews’ being attributable to differences among items or among reviewers.)
The researchers cannot say directly what the comments look like that accompany these reviews, because then it would be possible to do a Web search and identify the company. But Mr. Simester said they are something like this:
- I should have read all of the negative reviews before ordering. Please bring back the old style.
- I ordered this item over your Web site. Why is it that good designs are always changed? Please go back to the original.
- I am on a “Made in the USA†campaign and so am returning this item. Please stop importing.
The cranky customers are acting, the study concludes, as “self-appointed brand managers.†To put it another way, they are venting. The review forum gives them a simple and direct means of doing so: I hated this product, so listen to me.
As Mr. Simester put it in an interview: “Your best friends are your worst critics.†The study mentions in passing that Harley-Davidson’s customers were upset when the company introduced a perfume. They took it personally. The same phenomenon seems to be operating here and, perhaps, all over the Web, distorting the review process in a way never imagined.
The apparel retailer was somewhat alarmed to discover this was going on, Mr. Simester said. One possible solution is to allow customers to write reviews only if they have purchased the product. Or give customers easier ways to let their feelings be known.
For the rest of us, the rule remains the same: read reviews if you have no other source of information, but never place your full trust in them. Mr. Simester, who says he has never written a review himself, follows this philosophy.
The other conclusion is that behavior online is too easily taken as a mirror of reality when it is nothing of the sort. What seems to be the voice of the masses is the voice of a self-appointed few, magnified and distorted.
“For every thousand customers, only about 15 write these reviews â€" and one of them is writing negative reviews of products he hasn’t bought,†Mr. Simester said. “How surprised should we be that one out of a thousand people do something we have trouble understanding?â€
