Consulting & Innovation
Solutions & Products
Infrastructure & Operations
About us & News
Solutions & Products

Artificial Intelligence in Returns Management

Product Recognition for Easier Management of Returns

Artificial Intelligence (AI) is finding its way into more and more processes and is indispensable for many companies today. AI can simplify entire processes, which leads to considerable cost savings in the long term. This example illustrates how Artificial Intelligence can increase efficiency in the online and mail order returns process.

Project Overview

Background Situation

In the case of returns, original article numbers or other required information may be missing. A direct allocation of the article is then not possible and the identification must be carried out manually. Seasonally changing collections and great similarity of products make this assignment difficult.


Unique recognition of products through an intelligent digital object recognition service.


Arvato Systems has developed a service that assigns an item to be returned using a probability indicator. The product is recognized and identified via a video recording. Thanks to neural networks, this is also possible in different lighting conditions, angles and backgrounds to the article image. An analysis of material or other additional features such as weight is also possible.


Processing large amounts of data in a short time
Time savings through efficient processes
Cost savings
Reduction of error rates in the returns process
Relief of routine tasks
Faster processing of returns

Arvato Systems has developed a Machine Learning (ML) based system for the automatic recognition of items in returns. The application can recognize thousands of articles and associated attributes by image recognition. This “Product Detection” system takes a recording of a returned article as input and provides a set of data for that particular item as output, such as the article number, size, weight, etc. This help is especially useful for small or many similar items (such as jewelry), which are difficult to be identified by the human eye. The closest matching items for the return are pre-selected by the AI and displayed on a screen.

We have introduced the first steps of this POC case in our newsletter before. Click here to learn more about updates.