How Amazon reduces the return rate with AI tools

Fatih-Kağan Taşkoparan

Share this article

Anyone who likes to order clothes online knows the problem: which size is the right one? In fact, the problem with online retail has always been that sizes and fits vary for all buyers, depending on the item. People often order their standard size and then one item or another turns out to be larger or smaller. So far, however, the e-commerce industry has been very user-friendly when it comes to returns: More than 90 percent of retailers in Germany still offer returns free of charge. In an international comparison, this is not a matter of course; the rate in the United States is only 56%, and in China it is only a quarter of all retailers.

However, with increased transport costs and the current inflation problem, the industry in Germany is also faced with the question of how returns can best be minimized in order to save costs. Amazon recently presented its concept for minimizing returns in this context: Smart solutions designed to help customers choose the right fit size and thus improve the shopping experience at the same time.

Amazon starts where it counts

One of the most important aspects of online retail is returns management in the clothing segment. This is because suppliers incur enormous costs here, which they largely have to bear themselves. Amazon is considered a pioneer here and recently published a blog post explaining how artificial intelligence is helping the company to sustainably reduce its return rates. But that is not the only goal: by reducing returns and the associated cost savings, the technology should also help to ensure greater satisfaction among sellers and customers and offer an improved shopping experience. By using a total of four different AI tools, the system collects important data and evaluates it in a specially created database in order to ultimately offer personalized product and size recommendations for all users.

 "Deep Learning"

In short, all of the tools used should use previous reviews and recommendations on the fit of clothing from comparable customers as a source of information in order to make personalized recommendations. The basis for the AI program is the algorithm. Using deep learning, size ratios between brands and their respective sizing systems are taken into account and evaluated based on product ratings and fit preferences. The algorithm then generates an optimized size chart and product description, making it easier for customers to choose the right size.

Large Language Models (LLM)

The size table generated by the AI is mainly based on the analysis of customers' previous size recommendations for the product. This is where the so-called Large Language Model Tool comes into play, which analyzes these summarized recommendations from existing databases. For example, the LLM tool also helps to find out whether a product is described as "stretchy" from previous product reviews. This means that a smaller size may be sufficient. It would also be possible the other way around: the tool filters out from the product reviews that the respective product often runs small and therefore recommends that customers order this product one size larger.

Fit Insights Tool

This tool analyzes the reasons why customers return products. This helps online retailers to find out why product XY, for example, has a relatively high return rate. Information on customers' preferred sizes is also recorded. The tool then uses the insights gained to offer improved product descriptions. The algorithm anonymously groups customers with similar size and fit preferences and assigns selected products with a corresponding fit description to shopping profiles that exactly match this description. The AI tool uses selected product purchases based on the collected data on style, preferred size selection and rating and displays suitable product recommendations to customers with similar preferences. Amazon states that customers are more likely to shop online if certain sizes are already recommended to them. Amazon's databases are fed from a total of 19 countries.

Fit Review Highlights

The Fit Review Highlights (FRH) tool provides customers with an overview of their own fit and size when shopping online. The tool only evaluates reviews from online shoppers based on their own standard size. This means that if customer XY wears size M as standard, he or she will automatically be shown reviews in exactly this size. The tool thus enables personalized size advice for all customers. In addition, the underlying algorithm also uses predictive technology that automatically recommends larger sizes a few months later when buying trousers for a toddler or baby, for example.

Artificial intelligence in the industry - trend or already a reality?

The first signs of AI tools becoming established have already emerged through the use of Alexa, Google Assistant and Apple's Siri - smart home functions from the various providers are also already fully established. But with ChatGPT from OpenAI at the latest, AI tools have been made accessible to the masses in their full scope of use. Many companies are already using artificial intelligence to improve their transport routes in logistics, for example, or to better evaluate customer reviews - including Amazon. AI tools are neither a trend in everyday life nor in the world of online retail - rather, they are already an integral part of the e-commerce industry.