An extended Huff-model for robustly benchmarking and predicting retail network performance

• An extended Huff model is applied to the Belgian grocery market.
• Only easy acquirable input data is used.
• Results are validated on multiple levels: block, store and brand level.
• Inclusion of brand related store attractiveness drivers ensures model robustness.
• Incorporating more spatially dynamic drivers improves model accuracy.

This study proposes a modified Huff model that takes directly into account spatial competition between stores of the same brand, brand attraction based on actual brand performance and spatially variable substitution. The model uses only publicly available or easily acquirable data as input, whereas model output is extensively validated on various levels. These levels include comparison of modeled and real market shares on block, store and brand level for the Belgian food market. Results show that multi-objective optimization of model parameters yields comparable results on block level to other models in the literature but improved results on store and brand levels, thereby ensuring model robustness. This robustness also enables the application of the model for various business purposes as store location determination, leaflet distribution optimization, store and store concept benchmarking, without loss of spatial generality.