When analyzing the ranking distribution trend of products, if the ranking values under the search results are directly visualized, it is difficult to obtain effective information from the generated chart (because the ranking values of different products fluctuate too much, some products have very good sales and the ranking is in the single digits, and some products have not been sold for many days and the ranking is in the millions), so the ranking values need to be pre-processed.
The common ranking processing method is to take the logarithmic function of the ranking, which means that the ranking value of the product (category ranking) is the value calculated after taking the logarithmic function.
The reason for taking the logarithmic function is based on the following two points:
In the field of e-commerce and even the Internet, “long tail distribution” is very common.
On the Internet, from downloads of songs and software, clicks on web pages to sales in online stores, all show the characteristics of long tail distribution. The head of the curve is high, and as the sequence number increases, the curve drops in vain, but the curve does not fall quickly to zero at the tail position, but is extremely slowly close to the horizontal axis.
Take Taobao, a well-known domestic e-commerce platform, as an example. Search for the keyword “laptop” through the front-end webpage of the “Taobao” PC terminal, and then select the sales volume from high to low.
The monthly sales volume of the first 100 search-ranked products is recorded in sequence, and a product sales bar chart can be obtained.
The horizontal axis of the bar chart is different search rankings, and the vertical axis is the monthly sales volume of different search-ranked products. The shape of the image is very similar to the “long-tail distribution”.
For the “long-tail distribution”, the logarithmic function can effectively analyze the relationship between the sales volume of products in different orders in the distribution, and the simulated image of the logarithmic function is also very similar to the image of the “long-tail distribution”.
Therefore, in the analysis of product ranking data on the Amazon platform, choosing the logarithmic function to analyze the ranking is a very effective data processing method.
In addition to dealing with the problem of “long-tail distribution”, the logarithmic function can also effectively predict the relationship between sales volume and ranking.
On the Amazon platform, sellers cannot observe the actual sales of different products through the front desk (the domestic “Taobao” e-commerce platform can directly see the monthly sales data of different products), so how to predict sales through rankings has become the most concerned issue for many sellers. The research results found that “the relationship between log sales and log ranks is close to linear”, that is, “the relationship between sales data using logarithmic functions and ranking data using logarithmic functions is close to linear”@. Therefore, applying logarithmic functions to the ranking data on the Amazon platform can help operators effectively predict the actual sales of products.