How Do Ratings Work?

Since August 19 I have been recording Amazon UK sales and rating figures for my book The Black Orchestra. That’s numbers for 65 days. It struck me that the ratings on Amazon UK are not as responsive as the equivalent. The ratings in the UK seem to be updated less regularly — maybe every couple of hours. I have seen sudden small late afternoon spikes in sales (I’m talking 2 or 3 books, here) that aren’t reflected in the ratings until close to midnight.

I was puzzled by the way the ratings seem to react to small fluctuations in sales, which suggested that the relationship between sales and ratings is a short-term one. My best guess was that this relationship was based on no more than 2 days’ sales.

I decided to check this out, using MS-Excel’s CORREL function, that produces a correlation coefficient  for sets of numbers taken in pairs. Here’s what I found:

1 day coefficient: -0.54

2 day coefficient -0.72

3 day coefficient -0.69

4 day coefficient -0.67

5 day coefficient -0.65

6 day coefficient -0.68

7 day coefficient -0.68

10 day coefficient -0.61

14 day coefficient -0.43

30 day coefficient +0.12

These figures tell me that, as I suspected, the ratings are most strongly correlated with the 2-day sales figures, and bear no relationship at all to the 30-day sales figures.

Notes: The sales figures used in these calculations exclude KOLL sales. However, I don’t think they would have made much difference to the result if I had recorded and included them.

Total sales for the period of the study: 229 (KOLL figures for the period: 6)

Highest : lowest rating achieved during the period: 4,592 : 27,061

Leave a Reply

Your email address will not be published. Required fields are marked *