What does supermarket loyalty card data reveal about food purchase behaviours?
Supermarkets gather loyalty card data for marketing purposes but these novel data sources offer great potential for research and policy making. These projects in partnership with Sainsbury’s explore the use of transaction data to understand the food purchasing behaviours of our population.
Traditional dietary assessment methods in research can be challenging, with participant burden to complete an interview, diary, 24h recall or questionnaire and researcher burden to code the food record to obtain a nutrient breakdown. Self-reported assessment methods are subject to recall and social desirability biases, in addition to selection bias from the nature of volunteering to take part in a research study.
Supermarket loyalty card transaction records, linked to back of pack nutrient information, present a novel opportunity to use objective records of food purchases to assess diet at a household level. With a large sample size and multiple transactions, it is possible to review variation in food purchases over time and across different geographical areas.
Variation in fruit and vegetable purchasing patterns in Leeds: using novel loyalty card transaction data
This project uses supermarket loyalty card transactions for one retailer’s customers in Leeds, for 12 months during 2016. Fruit and vegetable purchases for customers who appear to shop regularly for a ‘complete’ shop, buying from at least 7 of 11 Living Cost and Food Survey categories, were calculated. Using total weight of fruits and vegetables purchased over one year, average portions (80g) per day, per household were generated.
Descriptive statistics of fruit and vegetable purchases by age, gender and Index of Multiple Deprivation of the loyalty card holder were generated. Using Geographical Information Systems, maps of neighbourhood purchases per month of the year were created to visualise variations.
Results reveal that households in Leeds purchase, on average, 3.5 portions of fruit and vegetables daily. This is higher in affluent and rural areas and with 22% of households purchasing more than 5 portions per day. Conversely in poor, urban areas 18% purchase less than 1 portion per day.
Research Team
- Victoria Jenneson – PhD Student, CDT in Data Analytics and Society
- Becky Shute, Sainsbury’s
- Dr Darren Greenwood, School of Medicine, University of Leeds
- Dr Michelle Morris – University Academic Fellow in Health Data Analytics, University of Leeds
- Professor Graham Clarke, School of Geography, University of Leeds
- Tim Rains, Sainsbury’s
Compliance with the Eatwell Guide: a case study using supermarket transaction records in Yorkshire and the Humber
This project from CDRC Research Fellow Stephen Clark, reviews the UK dietary recommendations, the Eatwell Guide, compared with loyalty card purchases for Yorkshire and the Humber.
As a proportion of the weight of all purchases, fruit and vegetable purchases are encouragingly close to the recommendations, with 31% purchased compared with 39% recommended. Surprisingly purchases of starchy products, such as bread and pasta, were below the recommended with 17% purchased compared to 37% recommended.
Meat and plant based protein products were similar to recommendations and more than twice as many dairy products are purchased compared to recommended. Perhaps unsurprisingly, sweet and savoury snacks like chocolate and crisps exceed recommendations with 17% of purchases by weight on these foods, compared to 3% recommended.
Research Team
- Stephen Clark – CDRC Research Fellow, University of Leeds
- Becky Shute – Sainsbury’s
- Victoria Jenneson – PhD Student, CDT in Data Analytics and Society
- Tim Rains – Sainsbury’s
- Dr Michelle Morris – University Academic Fellow in Health Data Analytics, University of Leeds
Future Work
We are excited to be collaborating with Sainsbury’s on this work and by the potential of these types of transaction data to understand the food purchasing behaviours of our population. We accept there are limitations to these data as they may not capture all food consumed and that individuals may buy from multiple retailers. However, compared to limitations of self-reported data such as recall bias, in addition to the burden of completing a food diary, limiting the scale of data collection, these novel data sources offer great potential in future research and policy making.
Dr Michelle Morris – University Academic Fellow in Health Data Analytics, University of Leeds
The work described here is in the early stages, full academic papers are forthcoming.