The most common approach to managing cannibalization risk in retail site selection is the exclusion zone: no new location within X miles of an existing store. That rule is simple to apply and simple to explain. It is also a poor approximation of how customers actually behave, which means it both over-blocks useful locations and under-protects against real risk at the same time.
The problem is that customers do not choose stores by drawing a circle on a map. They route to the most convenient option along their actual paths of movement, which depend on the road network, the location of their home and work, and the relative attractiveness of each store. Two stores that are four miles apart in a suburban grid might have nearly identical customer populations because they are both positioned along the same commute corridor. Two stores three miles apart in a geography cut by a river or a major highway might share almost no customers because the barrier makes the trip to the second store feel significantly more effortful than the distance implies.
A radius rule cannot see any of that. A gravity model can.
How Gravity Models Work
The gravity model for retail location draws from a physics analogy: the pull that a store exerts on a potential customer decreases with distance and increases with the store's attractiveness. Applied to cannibalization, it answers the question of how likely a customer who visits Store A is to visit a proposed new Store B instead, given the locations of both stores relative to where that customer lives or works.
In practice, a gravity model takes as inputs the trade area of the existing store (defined by where its current visitors come from), the location of the proposed new store, and a friction parameter that describes how quickly customer preference shifts with additional travel time or distance. The output is an estimate of what share of the existing store's visitor pool would likely shift patronage to the new location versus how many genuinely new customers the new location would attract from unserved areas.
The friction parameter is calibrated to your category. Convenience retail categories, where customers visit frequently and prioritize proximity, have higher friction. Destination retail categories, where customers are willing to travel specifically to visit a particular store, have lower friction. A specialty fitness retailer and a quick-serve lunch spot in the same geography would produce very different cannibalization estimates for the same proposed site, because the willingness to travel differs significantly.
Why Radius Rules Fail in Both Directions
A fixed-radius exclusion zone makes errors of both types. It will block a site that is within the exclusion distance from an existing store even when the road network means the two trade areas barely overlap. And it will approve a site outside the exclusion distance when the road network funnels both stores' customer populations along the same corridor.
Consider a hypothetical scenario: a 25-location chain evaluating a site two miles from an existing store. The two-mile mark falls just inside the exclusion zone, so the site is blocked under the radius rule. But the two stores are on opposite sides of an interstate interchange, and the drive-time between them is over 12 minutes. A gravity model would show minimal customer overlap because the friction of the route effectively separates the two trade areas. The radius rule rejected a genuinely complementary location.
Reverse the scenario. A candidate site is 3.2 miles from an existing store, clearing the three-mile exclusion zone comfortably. Both stores are positioned along the same suburban retail corridor, and drive time between them is under 7 minutes with no barriers. The gravity model shows that a significant share of the proposed site's projected customer base lives in the same catchment as the existing store. The radius rule approved a site with real cannibalization exposure because it measured geography, not customer behavior.
What the Model Tells You
A gravity model output for cannibalization analysis is not a yes/no answer. It gives you an estimated transfer rate: of the new store's projected annual visits, what share would come from customers who currently visit an existing location? And of those transferred visits, how many would result in lost revenue versus simply shifting where a customer fulfills an existing intent?
That distinction matters for the pro forma. A customer who previously drove to Store A and now goes to Store B instead represents a revenue transfer, not a revenue gain. The new store's volume needs to be adjusted to exclude that transferred business from the net new calculation. If 30 percent of the new store's projected visits are cannibalized from an existing location that is performing well, the economic case for the new store needs to be built on the remaining 70 percent, plus whatever incremental frequency lift the network proximity generates, minus the revenue the existing store loses.
Some cannibalization is acceptable, even expected. Opening a second location in a growing metro will inevitably draw some customers who previously drove farther to reach the first location. If the new location brings in enough genuinely new demand to justify the lease, the cannibalization is a cost of network expansion rather than a reason to block the site. The model tells you what that cost is. You decide whether the overall math works.
Where Gravity Models Still Require Judgment
Gravity models are more accurate than radius rules, but they are not perfect predictors. The friction parameter is estimated from historical data and category benchmarks, not measured directly for your specific brand in your specific markets. The existing store's trade area may not perfectly represent where its customers live, because foot traffic data captures device visits to the store but not home origins with perfect precision.
More fundamentally, a gravity model predicts customer redistribution under static conditions. It does not fully account for the market response effects of a new opening: some customers who were loyal to a competitor may now switch to your brand because you are now more convenient for them, which is a positive effect that partially offsets the cannibalization of your own stores. Modeling that dynamic requires additional assumptions about brand switching rates in your category.
The right way to use gravity model output is as a tighter input for the cannibalization estimate, not as a definitive prediction. It replaces the radius rule's binary block-or-approve logic with a quantified estimate that can be incorporated into the pro forma alongside other assumptions. The judgment call about whether the net economics justify the lease still belongs to your team. But the judgment is better when the cannibalization estimate is based on how customers actually move rather than on how far apart two pins are on a map.