Propensity estimation provides a means of converting the characteristics and size of the population at a small area level, such as a census output area or postcode sector, into estimates of spending power. This information can be said to provide a bottom-up method of calculating market size, as the propensity estimates by area can all be added together to estimate the size of the macro market. Demand estimates are therefore also one of the key components that feed into spatial interaction models, which model flows from the small area level to areas of supply. Usually estimating propensity involves combining small area lifestyle or demographic information with client spending patterns or market research data. However, this usually assigns most spending power to the residential location of customers.

Although many shopping trips can still reasonably be thought to originate from the home address an increasing amount of purchases are made from the workplace or combined with larger shopping trips, which distort the typically shopping pattern for some products or services. Techniques for demand reallocation, which reassign a portion of the residential demand to the workplace and the location of major shopping destinations, are able to address this problem and allow us to use more realistic demand surfaces in our spatial interaction models. Demand reallocation models typically use travel-to-work flows and surveys of major shopping patterns and are calibrated against observed customer spend data.

Other situations call for different adjustments to be made to the simple residential-based demand. In the case of demand elasticity for example, demand is related to the availability of supply outlets, and new demand may be generated by increasing the supply. One situation in which this may occur is when new outlets are introduced in an under-saturated market. Opening a bingo club in an area with little other leisure activity, for example, may generate more income than can be expected from the market’s demand estimates.

Because accurate propensity estimates are key to the success of the outcome of a spatial interaction model it is important to keep this data current. An important aspect of building propensity surfaces is, therefore, making sure the model tracks any changes to small area population size and composition as well as any changes to regional spending patterns. In practice this means that the ability to produce accurate population projections is a vital and often underestimated aspect of propensity estimation.