A correlation matrix shows the relationships between the categories visited in a store. For every pair of categories it gives the probability that shoppers who stop at one also stop at the other, which helps identify cross-selling opportunities and the best location for secondary displays.
Shoppermotion builds it from anonymous trolley and basket journeys, tracked passively from the entrance to the checkouts, and compares how often two categories are visited (or bought) in the same trip with what you would expect by chance. Representative results need at least 10,000 journeys from the previous three months. The time correlation matrix adds the order of the visit: in one example, the correlation from Fruits and Vegetables to Frozen Food is 40%, but only 11% in the opposite direction.
The matrix is rarely symmetrical. In one store, 20% of Vegetable shoppers also stopped at Bakery, while 100% of Bakery shoppers stopped at Vegetables, and 38.9% of shoppers who stopped at Seafood also spent relevant time near Tofu. In another example, one in every two Bakery shoppers passed by Dairy, but only 28.8% of Dairy shoppers walked through Bakery, which suggested placing fresh bread next to butter.
Typical uses are placing secondary displays of a category inside or next to its strongest partner, retargeting shoppers who visited a category without buying, and designing bundles backed by real behaviour. The correlation matrix article and the case study for a grocery chain show how.