You might not consciously know what led you to buy those 10 items at the grocery store, but soon retailers might have an idea of what motivated your purchase. Maryland Smith’s Bruno Jacobs is on a mission to pair scanner data – the data collected at checkout – at major retailers with the “why” behind consumer purchases.

Jacobs, a marketing professor at Maryland Smith, worked with Dennis Fok and Bas Donkers of Erasmus University Rotterdam on the research, forthcoming in the journal Marketing Science. They introduce a new model that enables retail marketers to gain in-depth insights from purchase history data. Unlike previous research, their work accounts for the large and varied product assortments, while accounting for differences across customers and shopping trips.

“This research will allow us to provide retailers with insights about their customer’s buying behavior,” says Jacobs. “The new model takes away the limitations of scanner data which did not provide an explanation for why customers made their purchase decisions.”

Jacobs’ research also tackles a challenge in analyzing a retailer’s purchase history data – the number and variety of products sold. Normally, the large variety of products makes it difficult to understand and analyze purchase behavior. However, Jacobs and his co-authors were able to create a complex algorithm that incorporates three large dimensions simultaneously: product, customer and time.

“Most customers will normally purchase not even 1% of a retailer’s products,” Jacobs says. “This results in a lot of zeros in the data set and ultimately makes it more difficult to sift through the data and provide meaningful interpretations.”

Jacobs says that, in addition to the sparseness of the data, accounting for the timing of purchases in this new model was also a major challenge.

“Including the time component with data was complex, but it also made the model much more realistic,” he says. “If you are in a store on Saturday afternoon in the middle of December, your needs and preferences may be much different than if you’re in the store on Monday morning in July. Our model accounts for that.”

Taking these complex measures into consideration can ultimately help retailers improve their marketing efforts and better personalize communications to their customers.

Read the full research, “Understanding Large-Scale Dynamic Purchase Behavior,” in Marketing Science.