Acquiring customer data isn’t a problem for retailers anymore. The new challenge is leveraging that data to create better customer relationships and grow a business. An increasingly common way many brands are approaching this task is to use machine learning. The process, which leverages complex models and algorithms that learn from data and make predictions and decisions based on it, has tremendous potential across a number of industries, and for retailers it has quickly gone from the nice-to-have category to a must-use tool to keep up with the competitive landscape and the changing demands of consumers.
Machine learning can be used by retailers to improve everything from pricing strategy to marketing practices, but one area where it can be extremely effective is customizing product assortment decisions on a local level. In traditional product assortment planning, a grocer leverages a national planogram, develops a large product universe and then makes small tweaks to customize it locally based on demographic, store location or customer purchase behaviors. With this method, one store might offer two vanilla yogurt SKUs, while another, larger store may offer six. Limited space means limiting assortment, and most of the time the decision to ax certain brands or products is made based on individual purchase behavior (vanilla yogurt one sells more than vanilla yogurt two, so two loses its place on the shelf).
Machine learning allows retailers to leverage customer data like demographics and week-to-week purchase habits to go beyond that process, replacing it with a more sophisticated method that considers the overall sales impact of assortment decisions, not just individual sales. At Precima, we use machine learning to improve this strategy in two key areas — general assortment decisions and the selection of new products to add to the shelf.
Optimizing Assortment Decisions
With the typical current product assortment model, stores must customize their product selections based on shelf space and basic data like individual product sales. The problem with this approach is that it doesn’t account for how removing a product may impact a shopper’s behavior and resonate across the entire store. Leveraging machine learning solves this problem by analyzing factors like substitutability and total-store impact, giving retailers the chance to better plan their assortment to not miss out on sales.
For instance, if a store removes a single offering of banana yogurt, data could indicate that shoppers will likely substitute that purchase for a different flavor. However, removing a six-pack of yogurt may mean shoppers won’t buy any yogurt at all. In this case, machine learning would notice this trend and recommend removing several individual flavors to save shelf space rather than nixing the multi-pack, maximizing profitability. A machine learning program can also help identify a “purchase halo” by drawing lines between how purchase behavior interacts across products. If shoppers are more likely to buy bread and peanut butter when they buy a yogurt six-pack, that six-pack is even more deserving of the real estate it takes up.
New Product Introduction
The second area where machine learning can help improve assortment strategy is new product introduction. Again, the traditional process for this practice is based largely on individual sales. If a certain type of yogurt is selling well, a grocer may introduce a new brand similar to that best-seller in hopes it will perform similarly. Machine learning can take this process much deeper.
By feeding a machine learning program the mountains of customer data at a retailer’s disposal, the retailer can begin to compare products by potential sales impact, rather than just sales. One product may be more transferable than another, one may be more likely to induce sales of accompanying products.
Improving your approach to product assortment can be a complex task, but one that is ultimately beneficial to sales in individual categories and across the entire store. Applying machine learning to your efforts from the beginning of the planning process can help to simplify your efforts, and to drive decisions that could never be achieved by considering just individual sales numbers or going “by the gut.” If you haven’t explored the power of machine learning yet, the sooner you start, the better off you (and your customers) will be.