Grocery expenditure is responsible for around 10% of total household spend in the UK, making the grocery retail market worth over £200bn a year in 2021. The size of this market and the nature of retailing competition makes it important for retailers to make the right decisions. One such decision is the location of their stores for which there have been a number of changes in the location, format and channel of consumer interaction along with the methods that have been employed to determine new store location. In recent years it has been suggested that the spatial interaction model is the most appropriate method for estimating new store revenue and hence location. However, previous attempts to explore the performance of the spatial interaction model in grocery retailing have been limited by access to loyalty card data. In this thesis we show that these models are unable to account for the heterogeneity in store conditions and consumer behaviour to model total store revenue. Notably, we find that at the regional scale the size of the errors are such that these models are unlikely to be used consistently in practice for estimating store revenue or locating new stores. Furthermore, that the performance achieved in previous applications are unlikely to be consistently replicated. Thus our results demonstrate that the spatial interaction model in its current form is no longer appropriate for modelling grocery store revenue. It is anticipated that these results may become a starting point for the development and application of alternative forms of models and methods for predicting grocery retailing store revenue. Notably, such new methods must be able to account for recent changes in consumer behaviour such as convenience store shopping, multi-purpose trips and the growing influence of e-commerce, alongside changes in retailers interaction strategies.