Michelle Morris is Professor of Data Science for Food at the University of Leeds and Deputy Director of the Healthy and Sustainable Places Data Service. In this new three-part blog series, Michelle explores the role of academic-industry partnerships in driving food system change, unpacks the opportunities and limitations of consumer purchase data, and shares practical lessons for researchers seeking to build productive industry collaborations.
Consumer purchase data have changed our ability to understand food purchasing behaviours at scale. This is the second in a series of articles that draw on insights from a recent paper I wrote with Nilani Sritharan Group Head of Healthy and Sustainable Diets from Sainsbury’s, and some of our team members, bringing together both academic and industry perspectives on how consumer purchase data can support food system transformation.
Each day, millions of transactions are recorded through supermarket tills, loyalty cards and online shopping platforms. For researchers, these smart data offer a rich source of information about what households buy and how purchasing patterns change over time.
However, while these datasets are incredibly valuable, they cannot answer every question. Understanding what sales data can and cannot tell us is essential if we want to use them responsibly.
What supermarket sales data can tell us
Consumer purchase data provide an objective record of food and drink purchases. Unlike traditional dietary surveys, these data are generated through everyday shopping activities rather than participation in research, which removes some of the reporting biases associated with self-reported food intake.
Sales data can help us understand:
- Purchasing patterns across large populations
- Any purchasing pattern differences between demographic groups
- Changes in behaviour over time
- Responses to interventions and policies
- Seasonal and geographic trends
- The healthiness and sustainability of food purchases
One of the greatest strengths of these data is scale. A single retailer may hold information on millions of customers and billions of product purchases covering many years. This allows researchers to investigate important questions that would be difficult using traditional dietary assessment methods alone.
Evaluating real-world interventions
Sales data are particularly valuable for evaluating interventions in real-world settings. In my Nutrition and Lifestyle Analytics team at the University of Leeds we’ve been using purchasing data to:
Investigate if price incentives encourage customers to buy more fruit and vegetables and try a wider variety
Sainsbury’s reduced selected fruit and vegetable prices to 60p for four weeks in January 2020 and January 2021 across 101 stores nationwide. To assess the impact, we analysed national sales data and loyalty card transactions from customers in Yorkshire and the Humber, South East, East Midlands and West Midlands, linking purchases to nutritional data and Eatwell Guide categories.
Sales of discounted fruit and vegetables increased by 78% in January 2020 and 56% during the January 2021 Covid-19 lockdown compared with 2019. In both years, sales rose well above expected seasonal levels, showing that the 60p Fruit and Veg trial substantially increased fruit and vegetable purchases.
Customers who engaged with the promotion also shifted towards healthier purchasing patterns, buying a greater proportion of fruit and vegetables and fewer discretionary foods in their baskets, in line with the Eatwell Guide.
Understand if in-store signposting could encourage customers to switch to healthier alternative products at the point-of-sale.
In this trial at Lidl, healthier alternative products in eight categories (including cereals, tuna, chicken, fries, granola, rice, cheese and coleslaw) were highlighted in-store to encourage healthier choices at no extra cost. Swaps were identified based on lower calories, fat, saturated fat or sugar, or higher fibre content, and were the same price or cheaper than the original products.
The four-week trial ran across all Lidl stores in February 2021. Analysis of sales data from 133 stores in two English regions compared purchasing patterns before, during and after the intervention. Results varied by category: sales of healthier cereals increased by 32% and healthier coleslaw by 71% above predicted levels, while sales of both original and healthier products increased for fries, rice and cheese. No meaningful changes were seen for chicken, tuna or granola.
Evaluate the impact of the High Fat, Sugar and Salt (HFSS) product placement restrictions legislation
We evaluated legislation restricting the prominent placement of foods high in fat, sugar and salt (HFSS) in supermarkets, using sales data from Asda, Tesco, Morrisons and Sainsbury’s, as well as feedback from shoppers and retailers.
We found that the legislation led to a significant reduction in HFSS purchases, with an estimated two million fewer HFSS products sold per day after its introduction.
The findings demonstrated that changing the retail environment can encourage healthier purchasing at scale, even when many shoppers are unaware of the changes taking place in-store.
Because these data are collected continuously, they provide an opportunity to study behaviour before, during and after interventions take place, which makes them useful for evaluating natural experiments.
What sales data cannot tell us
It’s important to remember that a purchase record is not the same as a consumption record – people do not always eat what they buy.
Sales data cannot tell us:
- Who consumed the food
- How food was shared within a household
- How much food was wasted
- What food was purchased from other retailers
- What food was eaten outside the home
For example, a household may purchase relatively few fruit and vegetables from one retailer because they buy fresh produce elsewhere. Looking at data from a single retailer alone cannot capture the full picture.
Similarly, household purchasing data should not be used to make assumptions about the dietary intake of specific individuals.
Context matters
Another challenge is interpretation. Sales data can highlight patterns, but they do not always explain why those patterns exist. A change in purchasing behaviour could reflect economic pressures, policy interventions, seasonal factors, marketing activity or broader social trends.
This is why sales data should be interpreted alongside other sources of evidence.
The future of food system research
Consumer purchase data are not a replacement for traditional dietary assessment methods. Instead, they are a powerful complement. When combined with nutrition expertise, population health research and robust analytical methods, they provide unique insight into how food systems function and how they might be improved.
The challenge is not deciding whether sales data are useful. The challenge is understanding how to use them appropriately, recognising both their strengths and their limitations.
In the next article in this series, we provide a practical guide for academics who are interested in pursuing industry collaborations.
Article contributors and declaration of interests
- Michelle A Morris is an inventor and shareholder at Dietary Assessment Limited but has not received payment from the company to date. Michelle has worked with multiple large retailers who have shared data for research purposes.
- Nilani Sritharan and Maddie Thomas work for Sainsbury’s Plc.
Alice Kininmonth, Emma Wilkins and Victoria Jenneson have worked with multiple large retailers who have shared data for research purposes.
