Returns go forward: using operations research to reduce transports in e-commerce

‘In this innovative concept the return is sent to a next customer, where the next customer receives a discount for accepting such a return’

The C2C return concept 

Almost everybody has returned items ordered online when these did not meet their expectations. This often happens in the case of clothes, as these have to fit. The customer returns them back to the seller, often for free. In turn, the internet seller receives the items, checks them, and stores them, ready to be sold again. Transporting and handling these returns is an enormous challenge for e-commerce sellers, especially in the clothing sector. All these transports create congestion and emissions, next to the handling costs at the warehouse. The question is how to address this issue effectively. 

To answer this question, the C2C concept was proposed by Carl van Heijst from consultancy company Bearing Point. Together with Eline Tetteroo, a student in the Master's programme Econometrics and Management Science, we further developed and optimised the C2C concept. In this innovative concept the return is sent from the returning customer directly to a next customer, where the next customer receives a discount for accepting such a return.  

In theory this is a nice idea, but to get it to work many operational details need to be specified. For instance, the return sender should be motivated to dispatch the return quickly. Additionally, do the discounted items cannibalise existing sales or would they lead to extra sales? This last issue would be crucial to an e-retailer to assess whether the concept is cost-effective.  

Optimisation for profitable C2C returns 

We guided Eline in translating the problem into a mathematical model and solving it with reinforcement learning (an AI optimisation tool), considering a continuous sale of a product. We tested the model on demand and return data from a large Dutch fashion retailer to determine the settings that yielded the best results. Additionally, we developed a second model that focused on an individual customer to gain deeper theoretical insights into the problem. 

It appeared that the stylised customer-based model also gave good advice in case of multiple customers and that it outperformed the reinforcement learning method. The profitability depends on several factors though: how many customers participate in the concept, to which extent customers accept the returns, the cost of sending it to the next customer, and the time allowed to find another customer. In general, the results are promising. The so-called C2C concept yields an increase of 34% in expected profit in the most optimistic setting, while in pessimistic scenarios it is still 6% better. Finally, the amount of work at the e-commerce warehouse is reduced substantially, which is especially useful in busy periods. 

From research to reality 

A lot of dedication and persistence went into the research: developing and analysing the new model, redoing the calculations, and rewriting and polishing the paper. After some setbacks, we finally got our results published in a highly respected journal, a nice achievement for the researchers involved: Sena Eruguz, Oktay Karabag, Wilco van den Heuvel, and Rommert Dekker. Carl van Heijst even quit his job and created the start-up company It Goes Forward to commercialise the concept. Presently, the C2C concept is active at a webshop for sustainable clothing, and we hope many companies will follow.