Validating Bluetooth logging as metric for shopper behaviour studies

• Trip length data measured using Bluetooth auto-logging was found to be very similar to the data measured manually using systematic sampling.
• Samples of shoppers obtained using Bluetooth auto-logging are similar to the shopper population in terms of demographics with the exception of underrepresenting older population.

The ability to track shoppers as they move through retail environments using signals emitted by their communication devices kindles the interest of practitioners and researchers. This data collection method is cheap and has the ability to supply big data for shopper insights. However, this non-probabilistic sampling method can possibly under- or over-represent certain groups of the shopper population. This study assesses the validity of the data describing the length of shopping trips and representativeness of the sample of shoppers carrying Bluetooth-enabled devices. The authors track unique Bluetooth logs in-store and compare to simultaneously collected data from a manual, systematic sample of 324 shoppers observed and interviewed in the same supermarket. A comparison of the results obtained from the two samples (auto-logging and manual systematic) drawn from the same population indicates automated Bluetooth tracking produces very similar (r=.92, p<.001) trip lengths to that observed manually. Basket size, spend and occupation of Bluetooth trackable shoppers are similar to those with no Bluetooth-enabled devices. These findings present compelling evidence that the Bluetooth auto-logging method holds great potential for retail practice and research. An expected under-representation of the oldest demographic (66 y.o. and over) in the Bluetooth discoverable sample calls for complementary methods of data collection to minimise representation bias in real-time tracking technologies for shopper research. The benefits of using auto-logging data describing shopping trip length for retail practitioners and researchers are discussed.