The Client and Setting
The client ran a small number of extremely large format hypermarket stores, with each store having a very large number of categories. The stores were all in one city, and the client had an interesting scheme in place to capture customer data.
However, the client struggled to understand the reasons why the customers visited its stores, and that hampered their marketing campaigns. The firm frequently resorted to mass media campaigns rather than more targeted ones.
Client Mandate
We run SMS campaigns each week. However, these plans are not really customised to the customers. We capture very rich data of each customer, and would like to know why they are visiting the stores. Based on this, we think we can offer much superior campaigns and get more bang for our buck.
Can you help us understand why our customers are visiting the stores, and what kind of offers can help us to retain them and get them to shop with us more frequently?
Problem Solving
I started with what was a “manual” and clearly unscalable process. I sat down with some senior leaders of the firm and we looked through a small random selection of anonymised customer bills. The challenge for these leaders was to try and tell as much as they knew about the customer based on these bills.
It was remarkable how much we could tell about a customer simply based on what she had shopped (apart from tagging a customer to a bill, the chain captured no other data). This gave a strong hint that through robust data analysis it was possible to know much more about the customer, and to use that to send far more customised offers than what the firm was already offering.
Some simple segmentation and clustering analysis delivered clear customer clusters, based on why they came to the store. Based on this, the clients and I together developed frameworks to map customers to the available offers, and thus be able to offer campaigns tailored to the reasons they shopped rather than subjecting them to generic mass marketing.
Results and Outcomes
The customised campaigns based on the segmentation were immediately implemented. Each week, we deliberately chose not to use the analysis to determine the offers to be sent to a random proportion of customers.
This A/B test of the marketing campaigns showed a 5 percentage point increase in the likelihood of a customer who received a customised campaign in returning to the store in the following week.