Creating smoother customer experiences for Jumbo

Is it possible for online food shopping to be fast, easy, and inspiring? Or is that just wishful thinking? Jumbo wanted to know if new technologies, like machine learning, could make the customer experience smoother and more enjoyable. But where was the best place to start?

Putting machine learning to the test.

Machine learning can be used to build a recommendation engine to create personal and relevant product lists for each customer.

As supermarket purchases tend to be fairly repetitive, Jumbo’s recommendation engine mostly recommends previous purchases which are not yet in the customer’s basket. This ensures that the potential purchases are additions to the weekly order instead of replacing a chosen item.

Jumbo’s recommendation engine also explored the added value of machine learning for the company. For example, three different recommendation methodologies were tested for the ‘Just try’ section of the Jumbo app. This allowed the company to see which list was the most relevant, delivered the best engagement, and boosted sales.

So that customers wouldn’t get bored by always receiving the same recommendations, we implemented a feedback loop. This tracked the recommendations that the customers had already seen.

It is important to note that for the recommendation engine to work properly, it needs large volumes of high-quality data. Jumbo’s recommendation engine used data from customer profiles and online purchasing behaviour.

100% data-driven, 100% relevant, 100% personal.

Personal recommendations are the way forward. Over 30% of users who saw personalised recommendations added one item (or more), increasing the average order size by several euros.

Some extra figures include:

• 30% more additions to the basket
• 270% increase in conversions compared to an unpersonalised list
• 1.9 more products per person on average
• Short- and long-term increase on the average voucher amounts

What's a recommendation engine?

As the name suggests, it’s a way of providing the customer with relevant recommendations. It works by looking at the customer’s past purchases, as well as the additional purchases done by customers with similar tastes, to suggest relevant, additional products that the customer might be interested in buying.

Recommendation engines are also used by streaming services, including Netflix and Spotify, and online retailers such as Amazon and Bol.