Have you ever ordered something online and then decided to return it? You’re not alone. While returns for in-store purchases average in the single digits, returns from e-commerce sales average 30%, with even higher rates in certain categories. This creates an interesting problem for both retailers and manufacturers that sell online: in addition to predicting a customer's original order, manufacturers and retailers need to be able to forecast expected returns.
Why has this been a difficult problem to solve?
Streamlined returns management starts with an advanced ability to forecast returns, which is not an easy problem to solve. Forecasting returns effectively is a two-step process: first, you must predict the sale itself, and next, you must predict that the sale will be reversed in the near future. The second event cannot occur if the first doesn't take place. Given the proliferation of new product introductions, ever-changing customer expectations and differences across markets and channels, this can easily become a mind-boggling problem to solve.
Unlocking the value of returns management
Online shopping has resulted in a dramatic spike in returns, forcing companies to look at their reverse supply chains. Just as it is important to manage the efficient flow of goods and services from a manufacturer to the customer in a traditional supply chain management context, it is equally important for the digital supply chains of tomorrow to manage the reverse flow effectively.
Additionally, managing returns requires supply chain resources — such as labor, transportation, and storage capacity — and an ill-designed returns operation can result in significant value leakage and margin losses. Learning how to manage returns well is becoming an increasingly important priority as companies introduce more new products into the market and attempt to make it easier for customers to do business with them by offering more choices and “no questions asked,” easy returns.
New technology ushers in new opportunities
Interestingly, physics holds the answer to the returns management problem. In physics, there’s a concept called lagged signal processing, where physicists observe two connected events that occur in sequence — Event A followed by Event B, a lag-time later — each with a different probability. Thanks to massive advances in computing technology, decision and data science algorithms can crunch through such calculations rapidly and predict the likelihood of a sale, as well as the likelihood the product will be returned a certain number of days later. With this sophisticated forecasting techniques on hand, companies can plan for their end to-end supply chain operations more holistically, and predict their actual revenues more reliably.
Machine-learning algorithms in this space are constantly evolving and are using transfer functions to establish the relationship between sales and returns. Such algorithms can perform updates to refine patterns based on the latest trends observed. The use of lagged signal processing in forecasting returns is now in advanced stages of research and development in the labs of leading supply chain software companies and is ready to be commercialized soon.
Machine learning has already come a long way and continues to make rapid advancements every year, improving the ability to make better decisions as we move more and more into the digital world. Forecasting returns is just one way machine learning can be leveraged to power your digital supply chain. Learn how it can also enable intelligent segmentation, drive insights from unstructured and structured data, and support image recognition for shelf logistics.