We are always trying to deep the knowledge about the Amazon marketplace. Today, we want to talk about the Best Sellers Rank or BSR. It is a tool that helps to estimate sales.
Firstly, it can be found on the page with product's details. You already know that items are usually listed in several categories. You can find a top level one or sub categories as well. One product can be found in many sub categories and sometimes even in the top categories. Algorithms pay attention to the top level categories commonly as there is no algorithm that can process 64.000+ sub categories. Maybe it will be created in future. :)
It is found out by calculating an item's number of orders compared to other items in the category. A certain amount of time is taken into account as well.
Long story short, all 'best selling' products will be in the first place. The second 'best selling' item will have #2 BSR, and so on.
We tried to find out how it is done exactly and tried to use machine learning, Linear Regression, in particular, to know what time frames Amazon analyzes. It hasn't given results concerning time frames, but we've learned a lot about BSR. Here is the list of our insights:
That's why these numbers tell us both how an item has been sold during its presence on Amazon and during the last couple of hours. But we can not know what exactly each BSR means until we know detailed history.
Here's an example
Let's pretend that we have 2 items: 'A' and 'B'. The first one has been selling at 40 units per day for 6 months, the second one had 1 sale for the last month.
Now let's imagine that 'B' starts promo activities and there are 120 sales for the last two days. It can seem that these items will have similar ranks, even if we won't consider that 'A' had 1200 sales, and 'B' just 120 for the last 30 days. And here is the result.
Looks interesting, doesn't it?
Since there are millions of items in any category, the number of sales can be the same, but ranks will vary.
If 'A' and 'B' have around 200 sales, but 'A' had 3 new sales in the last couple of hours, while 'B' had none, the first item will be ranked higher. A product that sold poorly is pushed down as BSR renews every hour.
Best Sellers Rank is a volatile metric and it's updated every hour. It can fluctuate significantly (from 38.000 to 98.000 and then again to 76.000) every hour. And this is what happens exactly.
And if BSR is estimated by the number of sales, its volatility, for example, for kettles will be higher in the morning, when people think of hot tea and order kettles. We doubt that the majority of people buy kettles at night. :)
The same is for the weekdays. People will likely order kettles during the working days, as they usually drink coffee before going to work. Some other orders connected to leisure will prevail on weekends.
That's why BSR can and can't tell us much at the same time. It fluctuates dramatically every hour and this movement has various causes.
Here are the key aspects that were very helpful:
- As you've already noticed, BSR is extremely unstable hour to hour. So, our algorithm had to be less volatile and change every month, but not every hour or day. Still, all changes that occur to an item, are taken into account. This makes estimation more accurate.
- BSR is updated to sales each night. This allows to keep up with current market trends.