Amazon sales analytics

4 minute read

Measuring performance as precisely as possible on Amazon is of increasing interest for many CPGs.

Why?

Because third party (3P or Seller) sales on Amazon have grown over 50% since 2020. Selling as a 3P became possible with Amazon in 1999 and they now account for 58% of their sales. 

However, since Amazon offers data to brands on their first party sales only, brands are in the dark as to how a significant portion of their products, categories and competitors are performing. This article offers a review of AMZ’s features and a summary of the challenges with Amazon sales analytics. 

The basics of Amazon

Amazon is in a state of constant evolution, so here’s a quick refresher on the AMZ essentials: analyzing sales, estimating sales, ASIN and filters.

It’s important to note that the expansion of Amazon into CPG products is quite recent. As consumers become more comfortable buying these categories of products on AMZ and they proliferate, extracting value from the data AMZ provides is becoming more complex. 

 

Table of Contents

Analyzing Amazon sales

Amazon is unique among platforms, retailers and marketplaces which means unique technology is required to analyze it. 

There are six ways in which AMZ differs from other retailers:

quick commerce visual

1. Sheer data volume and volatility

The number of products available on Amazon is enormous, with products sold by new or regional manufacturers that are not found on traditional retailer sites. New products are added every day and, since there can be many 3P sellers for a single product, there is often competition for the lowest price, creating high volatility in prices.

2. The presence of 3P sellers and competition for the Buy Box

Products may be sold by Amazon (1P) or through Amazon’s platform by other sellers (3P). Manufacturers can choose one model or the other, or a hybrid model with some of their products (or the same one) sold by Amazon and others sold directly or via authorized resellers. Whatever model they choose, CPGs are in competition for the Buy Box with 3P sellers and cannot choose who can or can’t sell their products. 3P sales are an unavoidable reality on AMZ and, as mentioned, an increasingly important one. The second article in this two part series outlines the potential benefits of 3P.

3. Sophisticated product pages

Amazon product pages are unique and rather complex. They include up to nine images, bullet points, a description and enhanced “A+” content, including high resolution images, videos, and benchmark tables. Customer ratings and reviews are widely consulted and are an important part of the product page, as is the questions and answers section.

4. Specific rules for visibility and the strategic importance of search

Visibility on Amazon depends primarily on search. Amazon’s A9 algorithm is complex and considers keyword relevancy, but emphasizes sales velocity and customer satisfaction. There are specific rules to respect for optimizing the product page. Price and product availability are also key. Finally, Amazon employs badges such as “Amazon’s choice” or “Best seller” which are impactful in terms of conversion.

5. Open but limited access to data

Amazon provides data on 1P sales, as well as data on page traffic and conversion. However, data on global product sales (1P and 3P combined) is not available. 

6. Proprietary product identification

Amazon uses its own system, ASIN (Amazon Standard Identification Number) for identifying products. ASINs are different on Amazon core and Amazon Prime Now and sometimes there is more than one ASIN for one single EAN.

Machine learning for estimating AMZ sales

The current generation of AI modelling provides item level sales and share measurement. It gives granular visibility on all CPG categories and products. 

The optimal view provides an estimate of a CPG’s market share in addition to competitor’s.

An accurate estimation requires dedicated models per country and per category. A generic machine learning model for all categories will lead to inaccuracies. People don’t buy chocolate and cleaning products, for instance, in the same way. 

Since Amazon’s platform changes constantly and features such as “Recommended Products’ change weekly, the new generation of machine learning modelling adapts to those changes by scraping the platform several times per day. 

Amazon and ASIN

All of the six features above require specific technologies that solution providers develop to analyze AMZ, however it’s useful to briefly mention the last one in particular because it ties them all together. Here’s how:

As discussed, the number of products on Amazon is huge, approximately 12 million. For a solution provider to get an accurate view of market share, they need to track all of those products. Since AMZ has a non-standard way of identifying items, a solution provider needs to match their own identification system with ASIN. 

Why?

Because when Amazon provides 1P data on sales and market share it needs to be automatically integrated with the 3P data from the solution provider so that manufacturers can have a global view of their sales and market share. 

Filters for analyzing Amazon

Several filters are needed to provide the analyses than can manufacturers to accurately monitor their KPIs;

  • Categories: by category, sub category or segment
  • Retailers: choice of retailers available
  • Brands: choice of brands to analyze
  • Periods: Filter by week or by period-the user can choose to show the comparison to the previous year, to the previous period, or from desired starting to end point
  • Display: by results displayed by volume or by value
  • Number of products: percentage or absolute value

Our other article in this series looks at the five families of analytics needs manufacturers have, as well as exploring measuring market share.

analytics amazon data impact by nielseniq