Amazon, Goodreads and Ebook Retail: 8 Ways Amazon Can Improve User Ratings
Published: 29th May 2013 Written by:
At the end of March Amazon announced that it was purchasing the social reading site Goodreads, dealing a serious blow to the rest of the bookselling world. Amazon has dominated book retailing for years but until now their main weakness in the emerging ebook retail sector was that they didn’t have a social offering or a community full of readers. Since discussion of books is so important to a large section of the book-buying audience, this social gap was something competitors could seriously exploit. For the sector as a whole, social has been an unfulfilled promise and the first retailer to combine a successful social network with a retail offering was going to be able to mount a serious attack on Amazon. Even with the acquisition of social site Anobii by a retail consortium led by HMV in 2010, there is still no serious retail/social offering in place. Now, the combination of the biggest retailer with the biggest social site creates a daunting prospect for any rivals.
What was most significant to me from an ecommerce point of view however, was the wealth of user ratings data that Amazon now owns. Collecting product reviews is a major challenge for any retailer. Even with a concerted effort in site design and customer follow-up most sites would be lucky to get greater than 2% of their customers to write a review after purchase, particularly for low cost FMCG products. Therefore large volumes of sales are needed to get a critical mass that delivers significant data, particularly across a long tail of products. Presumably, Amazon are also going to end Goodreads’ agreements with other sites that use their ratings (Google Books most notably) and set them back in the social proof stakes.
As many industry voices have stated, Amazon now has a ludicrously dominant position in the bookselling market and with that in mind, I will make my point. The literary world is familiar with the conceit that with great power comes great responsibility and I’d argue that now Amazon bears this responsibility and sets a new standard for online product reviews. This is particularly relevant to books which while they are products, are also cultural works occupying places in entertainment, art, historical record, education and more. It’s important therefore that a book website as prominent as Amazon recognises how influential the review scores are beyond simple commercial terms. Viewed even as a reference source, Amazon is unlikely to be rivalled by any non-commercial site for the volume of review data it can access.
Amazon led the way in online product reviews for years due to their sizeable user base, an excellent recommendation algorithm and a customer profile design that encouraged users to submit reviews. Recently however, there has been little innovation in this area and if anything, it has attracted negative attention for being easily corrupted by fake reviews from authors, publishers and third parties offering positive reviews for $5 a pop. Amazon have done work behind the scenes, clearing out suspicious reviews and profiles but it is impossible to tell how much irregularity is still present. Even with the dishonest reviews removed, there are numerous low quality and ill-informed opinions that degrade the system.
And this is where the main problem lies with the Goodreads review system. Even with a prioritisation system (driven by ‘likes’ of reviews) and a manual system of removing reviews that contravene their guidelines (abuse, off-topic, spam etc) there is a distinct regression to the mean for many popular book scores. As a result, a “difficult” book that may not be possible to appreciate under a certain reading age will receive a lot of low review scores from people who have heard its reputation but were disappointed when they read it. These scores will be counted in the average and therefore we see many literary classics with mediocre scores on Goodreads. As a result, they often score lower than hugely popular but critically dismissed books.
So what can Amazon do with their wealth of review and user data? What standard should the internet expect now? I offer the following suggestions:
Greater Weight to Top Reviewers
Several of my suggestions relate to providing a weighting to the user profile itself. Amazon already attributes status to top reviewers but does not weight their scores when calculating averages. Interestingly, Goodreads does attribute authority to most ‘liked’ reviewers and use a weighting in their averages. That, for me, is the way forward but I would argue a stronger weighting than Goodreads’ is needed for the most popular reviewers.
Weight Reviewers Based on Authority
Similar to popular reviewers which are judged by a site’s user base, authority can be bestowed by other means. Meta review site Metacritic manually applies a hidden level of authority to every review source they feature, thereby weighting the average in favour of people who are theoretically more knowledgeable about their subjects. Likewise, Amazon could conceivably apply an authority score to profiles in a number of ways such as allowing professional critics, authors, academics etc to register their profile for inclusion; incorporating social influence scores; manually identifying authoritative profiles.
Segment Authority Based on Product Category
Reviews of other products should be treated separately to books when calculating a profile’s authority. A review of an olive pitter, for example should not have a great bearing on a user’s ability to review books. Authority weighting can therefore be category-specific and provide a truer score for products.
Personalise Review Scores to the Visitor
Goodreads and LibraryThing both have a system that personalises book review scores towards an individual user, essentially providing recommendations. The systems involve comparing a user’s reviews of books to other profiles’ scores and predicting how much the user will like any book, using similar profiles. Amazon’s current recommendation system is comparable but simply outputs a recommendation rather than a score saying just how much you should like it. This is an objective system and therefore of interest for ecommerce where the goal is to sell books to individuals. It is a powerful selling tool and perhaps surprising therefore that the leading ecommerce site on the internet has not implemented it yet.
Reduce Weight of Reviewers with Extreme Score History
Some people see everything as superlative and give only minimum or maximum review scores. Also, the nature of online reviews sees people who have had extreme experiences with a product or service incentivised to complain or congratulate the supplier. While these views have validity, their extreme nature should be taken into account and when an individual profile provides consistently extreme reviews, they should be weighted down.
Reduce Weight of Reviewers with Poor Grammar and Spelling
This sounds like elitism but it can simply be applied to the most extreme cases. For book reviews in particular, the ability to write a sentence correctly should correlate with a basic reading age and experience. A review of a car by someone who can barely drive is of limited value.
Reduce Weight for Reviews with Flagged Key Words
The credibility of a review can be questioned should certain phrases like “didn’t finish”, “so boring” or “best book ever” or similar appear. Granted, the logistics of creating an algorithm that accurately finds these (ab)uses might be tricky, but we can dream.
Provide Users with a Separate Place to Review Value and Delivery Success
As alluded to earlier, reviews on retail sites can be about the experience of buying rather than the product itself. Amazon could segregate these reviews by providing a separate stream of reviews and scores for people wishing to comment on price, delivery, condition issues etc.
Amazon owes their mass of user reviews and data to their customers who have freely provided this over years. I would argue that they are far from reciprocating this generosity with the current ratings systems they provide. The userbase deserves a far better system that is worthy of their collective effort. It is time Amazon (or someone) set a new standard for product ratings that brings the technology into the new decade and sets a new benchmark for ecommerce and online experience.