Book sales and Amazon reviews

According to recent surveys, online shoppers love to read reviews.

In 2007 a PowerReviews/e-tailing group survey ¹ found that 68% of online shoppers read at least four product reviews before purchasing. Only 2% of online shoppers claimed not to read reviews.

Two other 2007 surveys emphasize the importance of user reviews. Forrester Research ² found that online shoppers want to see user ratings and reviews more than they want special offers or coupons, videos, personalization, or games. Avenue A|Razorfish ³ found that more online shoppers used user reviews than used comparison charts or expert reviews.

Do reviews affect book sales?

In terms of total book sales, reviews may increase sales to infrequent readers, but probably not to bookworms like me. My book purchases are a relatively fixed volume: I buy as many as I can read.

Do reviews shift which sites I buy from? Not much. I read newspaper and blog reviews, so I'm as likely to research books in an RSS newsreader as on a bookseller's site. I imagine many shoppers like to research and purchase all at one site, but the Avenue A|Razorfish study showed a growing population using RSS feeds, so I'm probably not alone.

So how much do user reviews shift purchases from book to book?

In 2003-04 Chevalier and Mayzlin ⁴ studied Amazon.com and Barnesandnoble.com, focusing on whether specific books' sales were affected by:

• The availability or lack of reviews
  
• Whether the reviews were positive or negative.

Chevalier and Mayzlin compared book-by-book sales data from Amazon and B&N to test several arguments against hosting user reviews. The analysis is a little out of date as Amazon’s review system keeps changing; I’ll interpose some links and thoughts along with their findings.

1. Motivation: Why review for free?

Or, as Steven Levitt of Freakonomics puts it: What motivates the 1,000th reviewer to contribute? Hasn’t everything already been said?

That depends on what the reviewer stands to benefit. Many Amazon reviews are probably written simply as an outlet to gush or rant over a book. I enjoy thinking critically about books, and discussing books with others. In 2003, reviewer Francis McInerney’s goal was to be mentioned in acknowledgments or quoted on a book jacket.

In terms of sales, even that 1,000th review may have some effect. Chevalier and Mayzlin found that over time,

an increase in the number of reviews at Amazon.com relative to bn.com continues to improve sales at Amazon.com relative to bn.com.

In other words, Amazon's approach pays off. Once an active reviewing community is established, those reviewers' interests can drive sales to the public at large. (It's interesting that this effect was documented on Amazon, where the review community can be highly competitive. It may be that a contentious atmosphere within the review community is not visible or important to non-reviewing book-buyers.)

2. Free-riding: Research here, buy there

I admit, I research books on Amazon far more often than I buy from Amazon. I would guess that free-riding happens mostly in that direction: most users know to go to Amazon for reviews, but fewer would leave Amazon for B&N reviews. After all, Chevalier and Mayzlin show that most books have fewer reviews on BN.com than on Amazon and "BN.com’s total sales equal about 15% of Amazon.com’s North American sales" ⁴.

On the other hand, I'm not sure free-riding should be a disincentive to host reviews. Sure, I sometimes read Amazon reviews but purchase elsewhere. But that means Amazon's reviews have lured me away from another site for at least part of the buying process.

Because of free-riding, Chevalier and Mayzlin's analysis

potentially greatly under-estimates the effect of word of mouth on sales…. Barnes & Noble.com customers could read Amazon reviews, or, similarly, Amazon reviews could affect offline sales. In fact, the success of a recently released best-seller “DaVinci Code” was attributed partly to an endorsement by... Francis McInerney.

That's right: a lone reviewer like Harriet Klausner can be influential in book sales.

3. Five-star (in)credibility

Chevalier and Mayzlin speculated that authors probably review their own books, so positive reviews wouldn't be as credible. This issue hit the news in 2004 (during Chevalier and Mayzlin’s study), when Amazon.ca briefly outed a number of authors who had given rave reviews to their own works and one-star reviews to rivals.

Since then Amazon has instituted Real Names, refined reviewers' “reputations”, and built up the Amazon "community" in hopes that it would self-police. These remedies have cut down on proliferating identities and self-reviewing, but of course abuses are impossible to prevent. In 2006 an Amazon UK Marketplace shop bribed a customer to change his review. The current flap involves an author berating and google-stalking reviewers.

Not all review trickery is driven by sales. Some Amazon users are motivated by Top 100 Reviewer status. Positive reviews are more often tagged as “Helpful”, which in turn increases the reviewer’s status in the Amazon community.

Can reviewers conspire to drive sales? As far as I know, I rarely run across coordinated campaigns of positive reviews on Amazon--though when I do, they're usually pretty obvious. However, author (or fan) trickery isn't the only reason to be leery of positive reviews. Many five-star reviews simply gush; there's often more information in mixed reviews. Chevalier and Mayzlin's analysis confirms that perception to some extent; they found that three-star reviews were, on average, longer than one- and five-star reviews.

At the same time I wonder if consumers tend to discount both positive and negative reviews. There's a large number of crank and bogus reviews on Amazon along with the five-star reviews.

4. Negative reviews may depress sales

It surprises me that consumers expect Amazon to act as an unbiased review site. On a website without user reviews, cherry-picking reviews would be business as usual. I don't really suspect Amazon of gross chicanery; it's not in the company's interests to deceive customers. Nonetheless, it wouldn't surprise me if Amazon removed negative reviews more readily than positive reviews.

Chevalier and Mayzlin provide some additional motivation for that idea. They found that one-star reviews (which are relatively rare) have more effect on sales than do five-star reviews. They piggyback this onto the credibility problem:

Although the author can post a large number of meaningless five-star reviews cheaply, he or she cannot prevent others from posting one-star reviews.

Mayzlin reiterated this finding in a February '08 NPR story, but didn’t mention an important wrinkle: their data were collected back before the Real Name program cut down on users with many accounts, and when Amazon displayed the most recent review first. It’s hard to say what these findings mean now that Amazon allows the user to sort reviews by star ratings.

Furthermore, Chevalier and Mayzlin's main analysis was of uncensored reviews. When they analyzed a smaller set of books for which Amazon had “pruned” the reviews, they found that new one-star reviews had no more effect than new five-star reviews. Was that "pruning" the same process that Amazon follows today?

On a positive note, Chevalier and Mayzlin also found that:

• Reviews are generally positive (more so at BN.com than Amazon.com)
  
• Consumers don’t just rely on a product’s average number of stars, but actually read the text of each other’s reviews

References

¹ PowerReviews/the e-tailing group. Social Shopping Study 2007. Nov. 2007.
1,200 people spending at least $500 per year in at least four online transactions.

² Forrester Research, North American Technographics Customer Experience, Marketing and Consumer Technology Online Survey. Q3, 2007.
A non-random survey of 5,366 US and Canadian consumers. Includes purchases of consumer electronics, travel and banking sites. Partial summary in Online Media Daily, Feb. 2008.

³ Avenue A | Razorfish. Digital Consumer Behavior Study. July 2007.
Only 475 respondents from a pretty high-tech group: 60% of respondents write or comment on blogs regularly.

⁴ Judith A. Chevalier and Dina Mayzlin. The Effect of Word of Mouth on Sales: Online Book Reviews. Yale School of Management Working Paper No's. ES-28 & MK-15 . First published online via SSRN, 2003. Final publication in Journal of Marketing Research 43(8): 345-354.