Artificial intelligence (AI) for the management of returns in online retail

Returns are unavoidable in online retail but, with the help of artificial intelligence, they can be significantly reduced and more efficiently managed. Forecasting product returns using machine learning is the starting point for all kinds of organizational and strategic measures.

Online shopping is very popular among many consumers, however, on average, every sixth order ends up being returned. The return rate is particularly high for fashion and accessories. Almost 40% of all orders in this sector are returned fully or partially to the online retailer (EHI Retail Institute, 2019). Such a high return rate generates enormous costs and is in conflict with efforts to become more sustainable. Understandably, new and, above all, technology-based approaches to lower return rates are eagerly welcomed.

Predicting returns with the help of artificial intelligence

Let’s get one thing clear: despite great hopes in the multitude of applications with artificial intelligence (AI) – this innovative technology will not be able to fully solve the problem of returns. Practical examples (e.g. Westphalia DataLab, 2020), however, indicate that AI-based solutions help in significantly lowering return rates and realizing efficiency gains within the company. Forecasting product returns using machine learning is the basis of one of the promising approaches. Special algorithms are trained using historical data and learn with every new return. As well as the historical sales and returns figures, further types of data are also used as a basis for the information. This includes information on the customer’s characteristics (e.g. age, place of residence, return behaviour), information about their basket (e.g. order sum, discounted articles, number of the same articles in different sizes) and what is known as external data (e.g. the weather, holidays and celebration days). Machine learning methods using this data are not only able to generate forecasts for the future, but are also able to identify the relevant drivers of returns, i.e. answering the question “under which circumstances are articles most often returned?”

Putting AI-based product return forecasts to best use

Using precise product returns forecasts means that procedures and resources pertaining to the operative processing of returns can be planned better and efficiency gains can be realized. For example, warehouse staff – whose numbers can often be flexibly increased or reduced thanks to external recruitment agencies – can be optimally planned. Having appropriate staff shift planning prevents bottlenecks in the warehouse and increases customer satisfaction as they can expect their returns to be processed quickly and a fast refund of the sales price. With the information gained regarding the drivers of returns, preventative measures could also be introduced with the aim of lowering the returns rate. For example, incorrect/inappropriate sizing labels on clothes could be corrected, the interdependence of products in the online shop could be meaningfully indicated and personalized digital sales advice could be developed. With AI, the possibilities for generating real added value are endless, it just takes courage to go down this road.

Due to the current situation, Connecta Bern will again be held as a digital event in 2021. Connecta is renowned for shining a light on the diverse nature of digitization and this year will be no different with content presented across the three formats of Connecta Blog, Connecta TV and Connecta Talk. Find out more here: www.swisspost.ch/connectaTarget not accessible