Skip to content

Distance Decay

Distance decay describes how much less likely a customer is to visit as distance increases. The API models this with a single beta (β) parameter that controls the rate of decay. Beta is shared across Huff and Gravity models — a higher beta means steeper drop-off (more local), a lower beta means people are willing to travel further.

The API uses two functional forms:

Exponential decay (convenience dining):

attraction = A × e^(-β × distance)

Power decay (destination dining):

attraction = A × distance^(-β)

Power decay has a heavier tail — customers at 2 km are less penalised than exponential. This matches destination dining behaviour where people seek out specific restaurants.

Beta is set by calcBeta() before any spatial model runs. The same beta is passed to both Huff and Gravity:

Service model / price bandβDecay typeRationale
Delivery0.3ExponentialFlat within 2 km delivery zone; distance matters much less
Takeaway0.5ExponentialWalk ~300 m willingly; steeper beyond
High-end HKD 200+2.5PowerDestination dining — patrons travel specifically for experience
Premium HKD 100–2001.5PowerSemi-destination; some distance tolerance
Budget / Mid (dine-in)0.8ExponentialConvenience: most customers within 400 m
  • Low β (0.3–0.8): A restaurant 1 km away still gets meaningful traffic. Good for delivery-first or destination concepts.
  • High β (1.5–2.5): Distance strongly suppresses traffic. Critical for budget walk-in concepts — being 200 m further than a competitor is a serious disadvantage.
  • Beta is fixed per service/price category — real decay varies by day of week, time of day, and individual preference
  • No separate beta for tourist vs resident behaviour
  • Does not account for physical barriers (harbour crossing, highway, etc.)
DateChange
2026-02-10Initial implementation with 5-tier beta parameterization