Jumbo x Data
How can retailers offer an online experience that is easy, fast, and inspiring? Can new technologies, like Machine Learning, help or hinder the online shopping experience? Jumbo wanted to inspire online customers with 100% personal, 100% data-driven, and 100% relevant recommendations. What are your recommendations for achieving Jumbo’s goal?
We recommended using Machine Learning to build a recommendation engine to compile personal, relevant product lists for each customer.
A recommendation engine stimulates engagement with relevant suggestions that increases the average coupon amount and number of conversions. Since purchasing behaviour in supermarkets is mostly repetitive, the focus of our recommendation engine suggests relevant products that the customer has not previously purchased to ensure that the extra items are additions to the weekly order.
A secondary objective was to explore the added value of Machine Learning applications for Jumbo online. This included testing three different recommendation methodologies under the ‘Just try’ list in the Jumbo app to increase the relevancy of lists, improve engagement, and boost sales.
Data is the driving force behind the engine and the Machine Learning application that ensured that performance continued to improve compared to logic-based models. The main data input sources were the online purchasing behaviour and customer profiles, including over 400 million interactions encompassing the details of each customer and the products they purchased over the preceding six months. The output was then available in real time. In order not to bore customers with the same product suggestions, we added a feedback loop that tracks which recommendations each customer has already seen.
The personal recommendations had a positive impact on the value of the average shopping basket, with over 30% of users who viewed a list purchasing something from their recommended list, increasing the average order size by several euros.