An old friend of mine – I’ll call her Gloria to save her face – admitted to me, “I’ve written a review of your book on Amazon, but I could only give it three stars. I just don’t like violence.”
Gloria’s conscience was bothering her, naturally enough, because she feared that knocking off a star or two for reasons of her purely personal likes and dislikes would handicap my new novel and mislead would–be buyers into thinking the book was a worse read than it actually is
Gloria needn’t have worried, because, without realising it, she had done me a big favour. And she would have done me an even bigger favour if she had given me only one star. Let me explain.
Take a look at Amazon’s book pages at the all–time best sellers: books like J. K. Rowling’s Harry Potter and the Philosopher’s Stone or Dan Brown’s The Da Vinci Code or Stephen King’s It. Considering these are some of the most popular and exciting books ever published, they not only have a mass of approving five–star rating, they also – rather paradoxically – have much less approving one-stars from some readers. Why should this be?
The phenomenon was noticed and investigated in 2009 by three academics, Nan Hu of Singapore Management University, Paul A. Pavlou of University of California, and Jennifer Zhang of University of Toledo.*
Why should a book that is immensely popular have some raspberries as well as plaudits? There are several possible explanations. The buyer could have been expecting something different and is disappointed. They may have been led to expect a very high literary standard and felt (ahem) let down by the prose of (ahem) Dan Brown or Joanne Rowling. They may have found it harder going than they expected, or too different from the film version they had watched and enjoyed.
Whatever the reason, every best–seller without exception has a smattering of brickbats from critics as well as the plentiful bouquets from admiring fans. This natural phenomenon and the characteristic “J–shaped” distribution of review stars it gives rise to is described as “an asymmetric bimodal distribution” by the academics.
There are a lot of 5–stars across the top, fewer four–stars next, very few, almost no, 3–star and 2–star rankings, but then there is a substantial number of 1–stars, making up the foot of the J.
Why is this J–shape of interest? In any random sampling process you get a very different result. If a thousand people sampled sandwiches or boiled eggs or coach journeys or flavoured crisps or hair styles, there would be roughly as many people in favour of each as against – a normal distribution or “binomially symmetrical distribution” with its familiar bell curve.
With books (and some other products on Amazon) this natural process is being skewed by some form of bias and it was this question that intrigued the academics.
They concluded, “This implies two biases: (1) purchasing bias – only consumers with favorable disposition towards a product purchase the product and have the opportunity to write a product review, and (2) under-reporting bias – consumers with polarized (either positive or negative) reviews are more likely to report their reviews than consumers with moderate reviews.”
In plain English, It’s mainly people who are already fans who buy and it’s only people who feel strongly for or against who bother to write a review.
This apparently arcane academic finding is of great interest to book buyers, writers and to Amazon for a number of reasons. Firstly, authors, or their publishers, sometimes try to game with Amazon by posting fake reviews. The more positive reviews a book gets, the better it sells and the higher it is automatically ranked in the best-seller list by Amazon’s algorithm. But if anyone trying to game the system in this way forgets to include poor reviews, the distribution will immediately show them up as fake.
Amazon’s software engineers have gone to great lengths to prevent people abusing their review system and are very hot both on identifying abusers and deleting their reviews. Lots of reviews emanating from the same IP address, for example, are likely to get the chop.