3 Point Summary

 

• High sales volume of a product does not necessarily equate to good quality

• Number of exposure greatly influences the sales of a product

• Therefore, when judging a product, it is vital to consider the number of times it has been shown or exposed to the customers, along with its sales volume

 

 

Impressions, or how often a consumer can view your product is vital in advertising.  We use CTR, also known as click-through rate, the ratio of the number of clicks versus the number of impressions as one metric to measure its efficiency. But, even though an ad has few clicks does not indicate its success. If you spend a lot of money and resources on a particular ad, it is natural that the number of clicks will increase.

 

Advertising costs are a finite resource. Therefore, it is essential to evaluate the performance of an ad through various indicators that consider impressions such as CTR, RPM, and eCPM and to optimize costs.

 

Is this a concept that only applies to the world of advertising? How do you judge a product's quality on the company’s website? Do you rely solely on sales volume when considering comprehensive marketing strategies, selecting products for the “BEST” category, or placing them on the main page banner? If so, please read this article to the end!

 

 

It may not necessarily be a good product just because it sells a lot

 

First, let’s examine the characteristics of product sales. Like the number of clicks in advertising, product sales should be considered with the number of times the product has had the opportunity to leave impressions on customers. The reason is that the sales volume of a product is greatly influenced by how much exposure it receives from customers. 

 

To identify the factors that affect product sales, we created a machine-learning model that predicts product sales based on various product characteristics. This model predicts sales for the next week based on the data from the week prior. Interestingly, it not only indicates sales volume but also reveals which data was nutritionally used in predicting the sales volume, similar to a nutrition label.

 

The chart below shows, in order, the characteristics that helped predict sales volume. As you can see, last week's product impressions are crucial in predicting sales for the following week.

 

Aren’t you curious about how accurately the model predicts sales? Stay tuned for the next article!

 

For an intuitive understanding, you can visualize the relationship between impressions and sales using a chart. While each point represents an individual product, the X-axis represents exposure, and the Y-axis represents sales. A trend line was also added to define the relationship between the two axes. Although there may be slight variations across different sites, in most e-commerce platforms, we observe a  positive correlation between product impressions and sales. The average correction coefficient was calculated to be 0.68. Furthermore, when calculated based on sales revenue instead of sales volume, the correlation becomes stronger at an average of 0.75.

 

As seen from the results of the machine learning model and the chart, products that sell well generally have higher impressions. When you think about it, it's pretty obvious. However, many site managers overlook this factor because while one can easily access product sales records, data on product impressions are not.

 

Therefore, in most e-commerce platforms, products are sorted into the “BEST” category based on sales volume. It may seem the best option if you don’t have access to product impressions. 

 

However, sorting products solely based on sales volume creates a “rich get richer” structure. Products with high sales will receive even more impressions, while products that initially had low sales will remain in a low impressions zone. This doesn’t necessarily mean the products with common impressions were bad; it could simply mean they didn’t have enough opportunities for exposure. If you truly want to select and sell quality products, this structure is not considered the most effective.

 

 

From this perspective, product B, which had higher impressions than product A, had lower sales. Wouldn’t it be a significant loss to continue advertising product B? What if we could detect a situation and give more opportunities to product A instead of product B? Furthermore, how beneficial would it be if this type of product-exposure optimization were applied site-wide?

 

Manage Product Impression 

 

Impressions, just like advertising, are a limited resource. Not all products can receive high impressions. Locations such as the big banners at the top of the site or the top of a page are where the most impressions occur. Locations such as the big banners at the top of the site or the top of a page are where the most impressions occur. Properly placing good products in those prominent areas with exposure is necessary to increase sales.

 

Is it somewhat vague? Let’s do a simple but exciting calculation.

 

 

For example, there is a site that has 4 big banners located at the top of the main page. Products displayed in the first banner will receive the most impressions, while products placed in the later banners will receive fewer impressions. The indicators of the four products currently displayed are as above chart. Assuming all product prices are the same at $10K and considering the metrics provided for the four products displayed on the top big banner (1, 2, 3, 4), the generated revenue from the big banner arrangement is $2.6K.

 

At first glance, the products are well arranged in order of sales. However, considering the number of times the banner was exposed to customers, the situation is more favorable for the products displayed on the big banners regarding their conversions per exposure. What would happen if we consider the sales order and conversion per impressions and arrange the products accordingly?

 

 

Considering impressions is just the beginning.

 

If we rearrange the products in the order of D, C, B, and A, which corresponds to the order of high conversion per impressions, the revenue generated would be nearly double the initial revenue at $4.9K! We made no significant effort and only adjusted the product display order! Unfortunately, this example is an extreme case. Increasing exposure often leads to a decrease in the conversion rate per exposure, making such a dramatic improvement challenging. Additionally, the revenue generated from big banners accounts for an average of 3-5% of the total revenue.

 

However, what is certain is that if product exposure optimization is applied not only to big banners but to the entire site, a noticeable increase in revenue can be expected. Of course, the ideal method would be personalized exposure for all products. Some Datarize clients are utilizing the Datarize API to implement personalized product exposure. Even if personalization is not possible, by effectively applying the strategies I have introduced so far, you can take the first step towards growth based on date.