Cannibalization in multi location retail is the redirection of visits from an existing store to a new one when their origin zones overlap. The usual safeguard is an exclusion zone: no new site within X miles of an existing store. It is easy to apply and explain, but poorly reflects customer behavior, blocking useful sites while missing genuine risk.
Customers do not select stores by tracing a circle on a map. They choose the most convenient option along their actual travel paths, shaped by the road network, where they live and work, and each store's relative appeal. Two stores four miles apart in a suburban grid may serve nearly the same customer population if both sit on one commute corridor. Two stores three miles apart, separated by a river or major highway, may share very few customers because reaching the second store takes more effort than the distance suggests.
Radius rules miss this. Gravity models capture it.
Gravity Model Method
A retail gravity model uses a physics analogy: a store's pull on a potential customer weakens with distance and grows with store attractiveness. For cannibalization, it estimates how likely someone who visits Store A is to visit a proposed Store B instead, based on both stores' positions relative to that customer's home or workplace.
In practice, the model uses the existing store's trade area, defined by where current visitors originate, the proposed site's location, and a friction parameter showing how quickly preference changes with added travel time or distance. It estimates the share of the existing visitor pool likely to shift to the new store, versus the customers the site may attract from previously unserved areas.
The friction parameter is calibrated to the category. Convenience retail, where visits are frequent and proximity matters, has higher friction. Destination retail, where shoppers will travel for a particular store, has lower friction. A specialty fitness retailer and a quick serve lunch spot in the same geography would yield very different cannibalization estimates for the same site because willingness to travel differs.
Why Radius Rules Fail Both Ways
A fixed radius creates both kinds of error. It can reject a site inside the exclusion distance even when the road network leaves the trade areas with little overlap. It can also approve a site outside that distance when roads funnel both stores' customers along the same corridor.
Consider a 25 location chain reviewing a site two miles from an existing store. Because two miles is inside the exclusion zone, the radius rule blocks it. Yet the stores sit on opposite sides of an interstate interchange, and drive time between them exceeds 12 minutes. A gravity model would indicate minimal overlap because route friction separates the trade areas. The rule rejected a genuinely complementary site.
Now reverse it. A candidate 3.2 miles from an existing store clears the three mile exclusion zone. Both locations sit on the same suburban retail corridor, with drive time under 7 minutes and no barriers. The gravity model finds that a significant share of the candidate's projected customers lives in the existing store's catchment. The radius rule approved real cannibalization exposure because it measured map distance, not customer behavior.
What the Model Reveals
A gravity model does not produce a yes or no cannibalization answer. It estimates a transfer rate: of the new store's projected annual visits, what share would come from people who currently visit an existing location? Of those transferred visits, how many mean lost revenue, and how many simply shift where an existing intent is fulfilled?
That distinction belongs in the pro forma. Someone who once drove to Store A and now visits Store B represents transferred revenue, not a gain. The new store's volume must exclude that business from the net new calculation. If 30 percent of projected visits are drawn from a well performing existing store, the case rests on the remaining 70 percent, any incremental frequency lift from network proximity, and the revenue lost at the existing store.
Some cannibalization is acceptable, and often expected. A second location in a growing metro will draw some customers who once traveled farther to reach the first. If the new site brings enough genuinely new demand to support its lease, that transfer is a network expansion cost, not a reason to reject it. Modeling does not remove cannibalization; it puts a cost on it. You decide whether the overall math works.
Where Gravity Models Need Judgment
Gravity models are more accurate than radius rules, but they are not perfect forecasts. Their friction parameter comes from historical data and category benchmarks, rather than direct measurement for your brand in each market. The existing store's trade area may also miss some customer origins, since foot traffic data records device visits at the store without perfectly identifying where those visitors live.
More fundamentally, a gravity model forecasts redistribution under static conditions. It does not fully capture the market response to a new opening. Some customers loyal to a competitor may switch to your brand because the new site is more convenient, partly offsetting cannibalization of your own stores. Modeling that effect requires additional assumptions about brand switching rates in your category.
Use gravity model output as a tighter cannibalization input, not a definitive forecast. It replaces the radius rule's binary block or approve choice with a quantified estimate that can sit in the pro forma beside other assumptions. Your team still decides whether the net economics support the lease. That decision improves when the estimate reflects customer movement, not map distance.