Huff Model — Probabilistic Retail Gravity Model
Huff Retail Gravity Model · Also known as: Huff Gravity Model, Probabilistic Retail Gravity Model, Huff Trade Area Model, Huff Çekim Modeli
Proposed by David Huff in 1964, the Huff Model is a probabilistic spatial interaction model that estimates the likelihood that consumers located in a given geographic zone will choose to shop at a particular retail outlet. It extends deterministic gravity models by assigning each consumer zone a probability of patronage across all competing stores, weighting store attractiveness (typically measured by floor area) against the friction of travel time or distance. The model is widely used in retail site selection, trade area delineation, and market share forecasting.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use the Huff Model when you need to estimate retail market shares or trade areas across multiple competing locations given spatial data on store sizes and travel times. The model assumes that consumer choice depends only on store attractiveness and travel friction, that the attractiveness exponents are stable across the study area, and that consumers act probabilistically rather than deterministically. It is best suited for retail and service facility planning with moderate numbers of competing sites. For contexts where attractiveness is multidimensional or consumer heterogeneity is large, extended versions incorporating additional attributes or discrete choice models such as Multinomial Logit may be preferable.
Strengths & limitations
- Produces continuous, probabilistic trade area boundaries rather than rigid Thiessen polygons, more closely reflecting actual consumer behavior.
- Handles competition among multiple stores simultaneously within a single consistent framework.
- Requires only two types of input data — a measure of store attractiveness and a measure of travel impedance — making it feasible with commonly available GIS and census data.
- The exponent parameters can be calibrated from observed patronage data, allowing the model to adapt to local consumer sensitivity and store-type characteristics.
- The model assumes that attractiveness is fully captured by a single size variable (e.g., floor area), ignoring price, product assortment, brand reputation, and other qualitative factors.
- Estimating the lambda and alpha exponents requires observed patronage survey data, which may be costly or unavailable; misspecified exponents can lead to substantially incorrect market share estimates.
- The model does not account for consumer heterogeneity: all consumers in a zone are treated as identical, whereas real populations differ by income, mobility, and preferences.
- Results are sensitive to the geographic zoning scheme used; coarse zones can mask intra-zone spatial variation and bias trade area estimates.
Frequently asked
How are the exponents lambda and alpha estimated?
Lambda and alpha are typically calibrated by fitting the model to observed patronage data — for example, household survey responses indicating which store respondents actually visit. Ordinary least squares regression on linearized forms or maximum likelihood estimation on the probability model can both be used. When primary survey data are unavailable, meta-analytic defaults drawn from comparable retail contexts are sometimes borrowed, though this practice introduces uncertainty and should be acknowledged explicitly in reporting.
What is the difference between the Huff Model and Reilly's Law of Retail Gravitation?
Reilly's Law assigns a hard boundary between two competing cities or stores: all consumers on one side go to one destination and all on the other side go to the other. The Huff Model replaces this deterministic rule with probabilities, distributing consumer patronage continuously across all competing destinations. This makes the Huff Model more realistic for multi-store markets and allows it to produce probabilistic market share estimates rather than binary zone assignments.
Can the Huff Model be applied when there are more than two competing stores?
Yes — unlike Reilly's Law, the Huff Model is designed for markets with any number of competing outlets. The denominator of the patronage probability formula sums the attraction scores of all stores accessible from a consumer zone, so adding or removing stores automatically updates the probabilities for all remaining alternatives. This makes the model particularly useful for realistic urban retail landscapes with dozens of competing facilities.
Sources
- Huff, D. L. (1964). Defining and estimating a trading area. Journal of Marketing, 28(3), 34–38. DOI: 10.1177/002224296402800307 ↗
How to cite this page
ScholarGate. (2026, June 2). Huff Retail Gravity Model. ScholarGate. https://scholargate.app/en/spatial-analysis/huff-model
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Location-AllocationSpatial analysis↔ compare
- Radiation ModelSpatial analysis↔ compare
- Spatial Interaction ModelSpatial analysis↔ compare