Huff Probability Model

The Question

“Given 289 competing restaurants within 400m, what is the probability that a customer from any given area will choose our restaurant?”

The Formula

Huff Model (1964, extended by Birkin & Clarke)

P(i→j) = S_j · d_ij^−β / Σ_k S_k · d_ik^−β

P(i→j) = probability customer in zone i chooses store j
S_j = attractiveness (size, reviews, brand, cuisine uniqueness)
d_ij = distance or travel time from zone i to store j
β = distance decay parameter
Σ_k = sum over ALL competing stores

Why This Formula

The Huff model is built for competitive environments. With 289 restaurants in 400m, counting nearby people is meaningless. You need the fraction you’ll capture. The denominator (all competitors weighted by distance) makes it realistic.

Source: Birkin, M. & Clarke, G. — Retail Geography

How Company Input Changes It

• S_j (attractiveness) depends on your business concept — a unique niche scores higher
• β depends on your restaurant type — convenience food has faster decay than destination dining
• Price point from your pricing strategy affects perceived attractiveness

Tip

The Restaurant Licences data provides every competitor’s location and type. The Demographics provide the population per zone. These are the model inputs.

Key Insight

At β=2.0, a restaurant 300m away is 9x less attractive than one at 100m. In Sheung Wan’s narrow streets, being on the right street matters enormously. Side streets (like Wa In Fong East) reduce casual discovery.