How to calculate store cannibalization before opening a new location

Cannibalization rate is the share of a new store’s demand that would have gone to your existing stores anyway. Written out, it is transferred demand divided by the candidate’s total demand. Transferred demand is the dollars (or visits, or trips) that move from a store you already operate to the new one once it opens.

A simple example makes the units concrete. Say the model forecasts the candidate at 1,000,000 dollars in first-year sales. Of that, it estimates 250,000 dollars would otherwise have rung up at your two closest stores. The cannibalization rate is 250,000 divided by 1,000,000, or 25 percent. The remaining 750,000 dollars is demand the candidate pulls from competitors or from households nobody was serving well. These figures are illustrative round numbers, but the arithmetic is exactly what a real estate committee will want to see.

The number that matters for the network is net new sales, not the headline forecast. In the example, the candidate adds 750,000 dollars to the system and quietly moves 250,000 dollars from one pocket to another. A site that forecasts well but cannibalizes heavily can look like growth on its own scorecard while barely moving the portfolio.

Five things drive the calculation. You need the candidate’s trade area and your existing stores’ trade areas, built from drive time rather than a radius, because roads and travel time decide who actually reaches each store. You need a measure of demand inside those areas, usually households and category spend, so overlap can be weighted by how many customers it represents instead of how much map it covers. You need each affected store’s current sales, so a percentage can be turned into dollars. You need a sense of relative pull between stores, since a customer in the overlap does not split fifty-fifty when one store is closer or more convenient. And you need the candidate’s own demand forecast, which is the denominator in the rate.

The calculation runs in a few steps. Draw the candidate’s drive-time trade area and the trade area for each nearby store of your own brand. Find where they intersect, then weight that intersection by the demand inside it rather than its size, so a small overlap over a dense, high-spend area counts for more than a large overlap over empty land. Inside the contested area, split demand between the candidate and the existing store using their relative attractiveness and travel cost, which is what a gravity model does. Sum the demand assigned away from each existing store, convert it to dollars using that store’s sales, and you have transferred demand. Divide by the candidate forecast for the rate.

For the deep version of this math, including the gravity and Huff treatment and how daypart splits change the answer, the Geod blog on cannibalization analysis walks through it in full. This page stays at the level of the formula and a worked example.

A lot of tools report a trade area overlap percentage and let readers treat it as the cannibalization risk. The two are different quantities. Overlap is geometry: the share of two catchments that intersect on a map. Transfer is behavior: the demand that actually moves. Published examples show how far apart they sit. CARTO has documented a case where 4.7 percent of one store’s area accounted for 19.3 percent of its population, because the overlap fell over a dense neighborhood. Placer.ai has shown a roughly 28 percent overlap producing about 17 percent customer churn. Overlap is an input you weight by demand and convert through a transfer model. It is not the answer on its own.

A defensible forecast breaks the candidate’s demand into three buckets. Transferred demand comes from your own stores and nets to zero across the portfolio. Competitor-capture is demand pulled from rival brands, which is real growth for you. Net-new demand is spend from households that were underserved before the store opened. Adding competitor-capture and net-new gives the portfolio gain; transferred demand is the part you subtract before celebrating. Keeping the three separate is what lets you compare a site that forecasts 900,000 dollars of mostly net-new sales against one that forecasts 1,000,000 dollars with a quarter of it lifted from a neighbor.

A committee number should travel with its reasoning. A defensible cannibalization output names the worst-affected store and the dollars it stands to lose, shows the candidate forecast split into transferred, competitor-capture, and net-new, reports demand-weighted overlap alongside raw overlap so the two are never confused, and carries the data sources and vintage behind every figure. With that on the page, a reviewer can challenge an assumption instead of taking a single percentage on faith. Geod produces this as an explainable brief built on a network-gravity model, weighting overlap by demand and reporting net-new against transferred for every candidate.

You can approximate cannibalization in a spreadsheet to screen out obvious problems. Pull a radius around the candidate, note which of your stores fall inside it, and flag the ones that look close. That is enough to drop a few sites from a long list. It is not enough to approve one, because a spreadsheet cannot build drive-time areas, weight overlap by demand, or split contested customers by relative pull. A purpose-built tool does those steps consistently and attaches the sources, which is what turns a rough flag into a number a committee will stand behind.