At elchie.eu (in few weeks a relaunch with more functionality will be move over to bookstats.com) I track the sales ranks of my own books with sales ranks I get from Amazon. Every hour I check the current data and store it within a database to get historical values.

There are some changes within the ranks, some are unmistakable are a sale - but sometimes Amazon is doing some little corrections, your sales rank will get slightly better, but it’s not a sale.

Today a friend of mine and me analyzed all my data I collected for over 18 months - we tried to detect levels, so we want to get some tresholds which indicates a sale for sure. Depending on the previous sales rank (one hour ago) and the current one we got some results depending on the previous rank:

• old rank: 50,000+, new one at least -16,7 %
• old rank: 20,000+, new one at least: -34,3 %
• old rank: 5,000+, new one at least: -17,1 %
• old rank: 2000+, new one at least: -14,3 %
• old rank: 1,500+, new one at least: -3,4 %
• old rank: 1,000+, new one at least: -4,4 %
• old rank: 0+, not enough data collected

I’m tracking right now only my own books, so sometimes they reached a rank in the top 1000, but for a good statistics not often enough.

What does this data mean? For example the Amazon sales rank on 1 pm was 5,541, at 2 pm it must be 4,593 or better (means: less) - if this is true, you can say, a book was sold on Amazon between 1 pm and 2 pm. if the new sales rank just changed to 5,023 for example, this was no sale, probably just other books were degraded.