Over the last few days, I've been promoting and advertising the first book in the Jesse McDermitt Caribbean Adventure Series, Fallen Out. Yesterday, it sold 2,692 Kindle copies on Amazon and so far over the whole four day promo, 3,153. This resulted in a peak ranking on Amazon of #18 in the US store and #87 in the UK store.
To achieve that number of sales, I spent a total of $794.95 in various advertising on book promoting websites and newsletters. The revenue generated so far stands at $2,163 for a profit of $1,368.05. The real profit will be on sell through to my other five books, which should result in tens of thousands in income over the next several weeks.
The main ad, the one that generated the most sales, was obviously BookBub. But rather than place several other ads after the BookBub ad to keep the rank higher for longer, I chose to do the opposite. I placed my ads on the two days before the BookBub ad.
We all know that the Amazon algorithm for ranking books, which is updated hourly, isn't just on sales in the most recent hour. Each hour before that most recent hour carries some weight, decreasing as each hour's sales gets older, until it finally drops off the algorithm at some unknown age.
To achieve the maximum results, each hour's sales over the entire span of the algorithm should be greater than the hour before it. This will keep any decreased hour's sales from affecting the algorithm in a negative way for many days.
Can we, as publishers, manipulate that to happen? No. But, what we can do is adjust our advertising schedule to get it as close as possible to those ever increasing sales numbers. We don't know when to start, because Amazon's algorithm is a more closely guarded secret than the recipe for Bush Beans. So, we start as far in advance as we possibly can.In a Kindle Countdown Deal, that's seven days. I oly use five, to have better control and a lower price.
Let's look at the algorithm like the mathematical equation that it is. I'm sure it's longer than 100 hours but for this example, let's use five hours. The most recent hour of sales counts full value, we'll call that baseline 1. Regardless of what percentage subsequent hours are, we can be assured each is lower. So, let's assume the hour before counts as .8 times the number of sales for that hour and the hour before that counts as .6 and so on down to zero.
If your sales fluctuate up and down, you might have something like this:
Last hour: 7
1 hour before: 5
2 hours before: 9
3 hours before: 4
4 hours before: 6
Total: 31
In this example, these sales would be calculated by: (7x1)+(5x.8)+(9x.6)+(4x.4)+(6x.2)
This yields a "score" of 19.2.
However, if you move those hour's of sales around so that each hour is more, but the total number of sales is the same, the resulting "score" will be higher.
Last hour:9
1 hour before: 7
2 hours before: 6
3 hours before: 5
4 hours before: 4
Total: 31
In this example, the "score" would be calculated as: (9x1)+(7x.8)+(6x.6)+(5x.4)+(4x.2)
This yields a "score" of 21.0, nearly a full value higher than the fluctuating sales "score".
It doesn't matter how long the algorithm is, if sales increase every hour, the final "score", which compares your book to every other book on Amazon, will be higher than if you had an equal number of sales across the same time period, but with the typical highs and lows we see on our daily graph.
No, it's not possible to achieve this perfect climb in sales, but the closer we can get to it, the better our "score". To achieve that, I monitored the email times of the 13 advertisers I planned to use. Actually, I monitor the delivery times of dozens, but it was these 13 that I chose, based on results and cost.
By placing ads with advertisers that deliver their book recommendations in a specific order, I was better able to control that hourly graph and came quite close to a steady rise in sales over the first three days before the BookBub ad came out. True, there aren't many advertisers that email in the middle of the night, but sales are slower then anyway.
In conclusion, each of us needs to take control of our advertising dollars and make them work in a manner that creates better results. I hope this simple mathematical explanation helps you out.
Semper Fi,
Wayne
