Detecting Address Uncertainty in Loyalty Card Data

Abstract

There is a fundamental need to better appreciate the dynamics and uncertainty of large consumer datasets, particularly if they are to be utilised to model social and geographical phenomena. This research, the first to utilise a major UK retailer’s loyalty card dataset, presents a novel data-driven approach for quantifying uncertainty in consumer addresses. Uncertain cases were identified through the linkage of locational and behavioural attributes and consumer mobility patterns recorded at a small area level. Such methods are not only important for the reliable adoption of large commercially generated datasets in research, but also for retailers if utilising this information to inform location-based marketing strategies. Results are contextualized with dynamics in the general population, demonstrating comparable relationships with Census migration patterns.