The project provided us with a deeper knowledge of the products and shopping behaviours that are associated with purchases of large amounts of calories in the retail sector. These insights support Nesta’s work in designing large-scale interventions to create a healthier retail food environment.

A critical goal of our healthy life mission is to improve evidence on diets. We must understand what food and drinks people are purchasing and consuming to help us better analyse the nation’s diet and know how this is impacting increasing obesity rates. In the UK, the retail sector is the single largest source of calories purchased. According to a study in The British Medical Journal, purchases of food and drinks to be consumed at home make up between 61% and 75% of all food purchases.

We analysed a large dataset of households' food and drink purchases to better understand what characteristics and shopping behaviours are most strongly associated with purchases of large amounts of calories in the retail sector. This builds on our previous work to select food categories that could be promising targets for reformulation. We wanted to understand the nation’s current diet and identify how it needs to shift to deliver a reduction in obesity rates. Our previous work identified the reduction in calories that we would need to see at population level to halve the obesity rate but it did not analyse what we are purchasing and how this is impacting obesity levels.

We investigated questions such as which groups have the largest calorie purchases, what patterns of food shopping behaviours are associated with largest calorie purchases (such as frequent shopping trips or infrequent large shops) and what types of products tend to be bought together resulting in high-calorie shopping baskets.

This project is divided into two phases.

1. During an initial exploratory data analysis phase, we ranked household groups based on the level of calories they have purchased. We then analysed these groups based on the household characteristics and what food and drink products they purchased.
2. Phase two used regression analysis to predict high-calorie purchases from a combination of geographical (for example, the location of the household), household (as in, the age of main shopper) and shopping behaviour (eg number of trips) factors.

Throughout our research, we looked at what typical high-calorie baskets looked like using a method called network analysis. This helped us to group commonly purchased food and drink categories together in order to identify and predict who purchased them.