Large-scale loyalty card data in health research

Abstract

Objective: To study the characteristics of large-scale loyalty card data obtained in Finland, and to evaluate their potential and challenges in health research.

Methods: We contacted the holders of a certain loyalty card living in a specific region in Finland via email, and requested their electronic informed consent to obtain their basic background characteristics and grocery expenditure data from 2016 for health research purposes. Non-participation and the characteristics and expenditure of the participants were mainly analysed using summary statistics and figures.

Results: The data on expenditure came from 14,595 (5.6% of those contacted) consenting loyalty card holders. A total of 68.5% of the participants were women, with an average age of 46 years. Women and residents of Helsinki were more likely to participate. Both young and old participants were underrepresented in the sample. We observed that annual expenditure represented roughly two-thirds of the nationally estimated annual averages. Customers and personnel differed in their characteristics and expenditure, but not so much in their most frequently bought items.

Conclusions: Loyalty card data from a major retailer enabled us to reach a large, heterogeneous sample with fewer resources than conventional surveys of the same magnitude. The potential of the data was great because of their size, coverage, objectivity, and long periods of dynamic data collection, which enables timely investigations. The challenges included bias due to non-participation, purchases in other stores, the level of detail in product grouping, and the knowledge gaps in what is being consumed and by whom. Loyalty card data are an underutilised resource in research, and could be used not only in retailers' activities, but also for societal benefit.

Keywords: Health behaviour; SWOT; food expenditure; loyalty card data; participation; purchases; sales data.

Figure 1.

Mean annual expenditure by age and gender in the two samples. Observed age–gender specific means are denoted by markers, and solid lines indicate weighted loess curve fits.

Figure 2.

Most purchased product groups (top 10, ranked by expenditure) among customers and personnel.

Figure 3.

Total expenditure of customers during calendar year. Daily sums (grey solid line) and centred seven-day moving average (black solid line). The vertical reference lines are positioned at Good Friday, 1st of May, Midsummer Eve and Christmas Eve.