Strategic Retail Location

Gravity Modeling (or Spatial Interaction Modeling)

A Spatial Interaction Model (SIM) considers the likelihood that a shopper in a given location (a) has a need for a good or service, and (b) chooses to acquire it from a particular store.  That likelihood depends on attributes of the customer (e.g. age, income), attributes of the store (e.g. floor space, availability of parking and public transit, advertising, pricing), and its accessibility to the customer.  Thus all stores get at least a few patrons from even the most distant locations, given enough time; but the majority of shoppers are those who reside or work nearby.  This is consistent with reality—for example a Los Angeles boutique gets the odd customer from Tulsa, Toronto, or even Tokyo; but for the most part its clientele is drawn from the neighbourhood and nearby towns.

The SIM is mathematically analogous to gravitational laws regarding forces that keep planets in their orbits, hence the original term, gravity model. Some researchers took the concept to extremes and proposed bizarre hypotheses of social physics. For this reason, and other technical reasons, the term “spatial interaction model” is preferred.

A key feature of the SIM is the Distance Decay (DD) profile.  This is a curve that summarizes how patronization of the store declines over distance.  For the first 1–2 km the store dominates the market.  As distance from the store increases, customers are more likely to consider other options, and the likelihood of patronization of that store declines.  There are differences between the DD profiles of businesses.  A shopper may drive 20 km for a good deal on a major appliance, up to 10 km for weekly groceries, but for just a carton of milk the limit is probably 3 km.  Therefore a convenience store has a completely different profile from an appliance store.  The compactness and degree of urbanization of the community, distribution of competing and complementary opportunities, shopping traditions, etc, make a difference.  Even for a single retail chain, some locations have profiles significantly different from others, even within the same city.  The model specification itself may differ, depending on the nature of competitive forces.

The SIM doesn't rely on boundaries, and it does not cast behaviour in black and white.  The destination-choice process is modelled in terms of probabilities, allowing residents of a given area to distribute their patronage over a selection of stores, near and far.

Each business needs a set of SIMs specified and calibrated to suit its practice. The SIM is the foundation of location evaluation and sales forecasting.  Moreover, the intensive local study required to calibrate the models, and to understand the individual deviations from it, are critical in interpreting the results of subsequent numerical analyses.

Back: Retail Site Selection Models

© Digital Geographic Research Corporation Contact us