Chipotle books about 100 basis points of comparable-sales drag from new restaurant openings and expects the affected restaurants back inside 12 to 13 months. The number is forecast and budgeted before any of it reaches a comparable-sales line. Krispy Kreme had no measurement of shared demand along the road network when it was opening stores, and the same effect turned up in its 2004 and 2005 sales reports, after the stores were already trading, with several franchisees going bankrupt. The transfer was the same event in both cases. Only one company had a number on it beforehand.
When Store Overlap Is a Budgeted Cost
The moment overlap becomes a problem is rarely the moment the second store opens. It arrives one or two quarters later, when somebody asks in a review why the older store is down and nobody has an answer prepared. A forecast number arriving on schedule is a cost of expansion, while the same number arriving unannounced gets treated as a failure of site selection.
Starbucks ran both versions of that in the same decade. Its history is the clearest available argument for doing the cannibalization analysis up front. The clustering was deliberate for years, and for reasons the company stated. With almost no advertising budget in the early period, Starbucks used storefronts as billboards and clustered them to intercept commuters. Its own annual filings described a level of cannibalization of existing stores by new stores that management believed the incremental sales and returns justified. That is a documented, priced-in decision.
By 2008 the pricing had stopped working. Howard Schultz returned as chief executive with the stock down 42%. The company announced the closure of roughly 600 underperforming US stores, and overexpansion and neighboring locations competing were among the named causes. By 2017 the average Starbucks had almost four other Starbucks inside a mile.
Why Cannibalization Is Invisible on a Pin Map
Most of the damage in early screening comes from two beliefs. The first is that two pins far enough apart cannot be competing. The second is that a 3-mile ring around each of them answers it. Both beliefs fail because customers reach a store along roads that no circle describes.
How Road Networks Change Trade Area Overlap
Stores several miles apart can share most of their trade area when the road network funnels the same commuters past both of them. A pair a mile apart can barely compete when a river with one bridge, a rail corridor, a highway median or a one-way morning commute puts them in separate catchments. Pin positions carry no information about any of that.
The eyeball test produces both kinds of error at once. Sites get rejected for looking crowded when the traffic pattern keeps them apart, and others get approved for looking spaced when both are on the same arterial. Either way the error stands until the store is trading, roughly 18 months after the decision that caused it.
Why the 3-Mile Ring Fails
The 3-mile ring is the inherited default in retail site selection. Practitioners now argue against it more consistently than against anything else in the method. A circle presumes every approach direction is equally likely, and commuting patterns are never symmetrical. The assumption would hold only in a market where road layout had no effect on how people travel. Two candidate sites with identical 3-mile rings routinely reach materially different numbers of people.
Trade area analysis also varies by format in ways one radius cannot express. Convenience formats draw from roughly a mile, or about a 5-minute drive. Everyday goods rarely draw anyone from more than 15 to 20 minutes away. Specialty and destination retail reaches 5 to 10 miles or further, with the exact reach moving by brand. Because those ranges span an order of magnitude, a standard radius is the wrong instrument.
A drive time polygon traces reach along the actual road network, stretching down a highway and stopping at a barrier. Where two of those polygons overlap, the shared ground has been measured along roads rather than drawn with a compass.
Cannibalization Analysis in Maptive
Draw Matching Drive Time Polygons
Since radius circles are fast and show where to look, the screen starts there. Open Map Tools, click the Distance Radius Circles tool, then either click an existing store marker and choose Draw Radius or type the address into the starting location box. Set the proximity distance and the color, then click Add Proximity Radius. Repeat at the same mileage for every store in the market and look at where the circles stack.
The finding comes from the Drive Time Polygon tool, in the same Map Tools menu. Its starting location accepts a typed address or a point picked off the map, and both the travel time and the band color are set in that panel before Add Drive Time Polygon renders it. Anything up to four hours calculates precisely. Set at the same travel time from each store in the pair, the two bands overlap on ground that either store reaches for the same amount of driving.
The first time an analyst does this on a market they already know, the polygons usually come out asymmetric in a way the circles never showed. One store is on the inbound side of a commute and the other is not. Polygons can be selected individually or grouped. Location data exports from one polygon or from all of them in xls, xlsx, csv or tsv.
What the Numbers Inside the Overlap Show
Clicking a band opens the pop-out where its metrics readout gets configured. Against a pair of overlapping polygons rather than a whole territory, three readouts do the work, namely total customer records, summed sales volume and a group count of store or format types. Comparing those three between the two polygons is as close to a transfer estimate as the platform gets, since nothing inside it computes a Huff model or scores cannibalization.
Weighting matters more than the raw overlap area. A marker-density heat map goes underneath the polygons. Select the heat map style for marker density, click Select Sample, choose Specific Group, pick a column holding the category values and then a specific value from it, then click Add Heat Map. Radius, opacity and intensity stay adjustable afterward. Move the intensity slider before drawing a conclusion, since a threshold set too low turns a moderate cluster into an apparent hot spot. Where the analysis has to attribute shared demand to a countable unit, the boundary tool does the same job by ZIP code or census boundary. The forecast at the end of all this belongs to the analyst, built out of the record counts and sales sums the export contains.
What Overlap Percentage Measures
A 25% overlap figure gets taken for 25% of revenue at risk. The two numbers are unrelated. Overlap measures geographic exposure. A 25% figure says roughly a quarter of the existing store’s catchment area is shared with the proposed site. Turning that exposure into transferred revenue depends on how many households are inside the shared ground, how the roads run through it, then which store is more attractive when they arrive. A large shared polygon over empty ground is worth less attention than a small one over a dense corridor.
How to Calculate a Cannibalization Rate
The post-opening measure is the decline in the existing store’s traffic after the new store opens, divided by the new store’s traffic. The division is simple. Whoever picks the baseline decides whether the answer means anything. If the whole market fell 5% over the same window, that 5% comes out before anything is attributed to the new store, and seasonality has to be controlled the same way. A brand that skips both steps will book market softness as cannibalization, or book cannibalization as market softness. Both errors come out of the exercise looking like a precise figure, and a precise figure rarely gets queried.
The site-selection variant runs on estimates, dividing estimated sales lost by the existing store’s sales before the opening. Both feed the figure a cannibalization analysis exists to produce. Net incremental sales is the share of the new store’s revenue that is genuinely new demand rather than demand that moved across town.
What an Acceptable Threshold Looks Like
There is no universal threshold. A brand that adopts somebody else’s has borrowed a conclusion without the inputs that produced it. The acceptable rate is whichever one still leaves the new store with a positive return after the damage to the existing base, given that market’s rent and revenue per visit.
The working band in practice is 15% to 30%. That band is industry convention with no published study behind it.
Category explains most of the spread. Quick-service restaurant brands work at the top of the band, near 25% to 30%, where visit frequency is high enough that the cannibalized store still covers its own economics. Apparel and specialty retailers stay nearer 15% to 20%, since each lost visit takes far more revenue with it. Growth stage moves a brand a few points inside that range without moving the range. Franchise systems handle the same tension through the protected territory of a franchisee, where the agreement itself fixes the acceptable distance or share before any individual site comes up.
How the Huff Model Turns Distance Into Probability
David Huff adapted the gravity model to retail trade areas in 1963. It remains the standard because it assigns no hard boundary at all. Every neighborhood gets a probability of shopping at each store.
A store’s draw rises with its attractiveness, measured as square footage or assortment or a composite score, and falls as travel time rises. A distance-decay exponent sets how steeply. Each neighborhood’s probability of choosing a given store is that store’s attractiveness-over-distance value divided by the same value summed across every competing option.
Adding a sibling store adds a new term to the denominator for every household nearby. That lowers their probability of visiting the original store even when nothing about the original store has changed. Distance to the nearest sibling is therefore the wrong quantity to ask about. The question the model answers is how much of the surrounding probability moves, and from which neighborhoods.
The magnitudes are smaller than intuition suggests, and they compound. A 2018 analysis of the 2008 Starbucks closures put the average cannibalization imposed by one neighboring outlet at 1.2% within a mile and 0.4% at one to three miles. Those figures are dated, drawn from one event, and should not be read as current benchmarks. What they imply moves with the number of neighbors inside a mile before it moves with the distance to any one of them.
Franchise Encroachment and the Map as Evidence
When the cannibalized store belongs to a franchisee, everything above changes character. Chain revenue and the franchisor’s royalty base can improve while the operator’s income falls. One transaction is then a success on one balance sheet and a loss on another. In almost every US state there is no statutory protection against same-brand franchise encroachment. The franchisee’s position rests on whatever territory protection the franchise agreement grants, plus the implied covenant of good faith and fair dealing that courts read into contracts generally. That covenant has protected some franchisees and not others.
Some brands write the protection down. Choice Hotels created a formal impact policy in 1992 and revised it in 1999, giving an existing franchisee the right to object to a same-brand property inside a 15-mile radius. A stated radius is a blunt instrument by the standards of everything above. It is still better than an argument conducted with no rule at all. The distinction between exclusive and protected territories determines how enforceable that language is.
For franchise development that changes what the analysis is for. A franchisee’s counsel will read the overlap map looking for the flaw in it, and a hand-drawn circle does not survive that reading. Export the drive time polygons with their counted record sets and keep the dated file. The map itself gets redrawn every time the data refreshes.
Reasons to Accept Store Overlap on Purpose
Treating cannibalization as always a mistake costs more than any of the errors above. It stops teams from planning the overlap they should be planning.
Three versions are defensible. A blocking site takes a corner before a competitor can, which matters where corners are scarce. A capacity-relief site absorbs demand an existing store was already turning away through drive-thru stacking, register queues or parking turnover, so the transfer that registers as cannibalization in the sales report appears as recovered demand in the throughput report. The third case is a matter of format. Putting a pickup or drive-thru unit beside a full-service one splits customers by occasion, the same logic that separates healthy channel conflict from the destructive kind.
All three are frequently applied backward. A site chosen because it was the only one available in the pipeline gets described afterward as a blocking play, and the claim is rarely tested. A justification that appears only in the post-opening review, with no counterpart in the dated cannibalization analysis submitted with the site package, was written after the fact to explain a decision already taken.
Store-level comparable sales are not the whole ledger. Halo measurement across retail portfolios puts the online lift from opening a store at 6.9% in the affected market. A capacity-relief or format-split unit therefore adds an online order book that never appears on the comparable-sales line of the store beside it. The case for deliberate overlap is under-credited for exactly that reason.
Frequently Asked Questions
Cannibalization analysis measures how much of a new store’s sales come from existing stores in the same brand rather than from new demand. It is done by estimating trade area overlap between the proposed site and nearby locations, weighting that overlap by the demand inside it, and calculating the net incremental sales the new site would add to the network.
Store cannibalization is the drop in sales or visits at an existing location caused by opening another location of the same brand nearby. The underlying mechanism is trade area overlap, meaning the share of one store’s customers who could be served as easily by the other.
The post-opening rate is the decline in the existing store’s traffic after the new store opens, divided by the new store’s traffic, times 100. Strip out market-wide trend and seasonality before attributing anything to the new store. The site-selection variant divides estimated sales lost by the existing store’s sales before the opening.
Retailers work inside a 15% to 30% band taken from practice. Position within the band follows visit frequency. A quick-service brand recovers a cannibalized store on volume and can tolerate the top of the range, while an apparel retailer loses too much revenue per visit to go past the bottom of it. The number that governs any individual deal is the one that still leaves the new store profitable after the older store’s decline, calculated on that market’s rent and revenue per visit.
Distance is the wrong test and no minimum exists in the trade. The workable check draws the same travel time from both sites and counts the customer records inside the ground they share. Two locations four miles apart on one commuter corridor return a large shared count. Two a mile apart with a river between them and one bridge return almost none.
It is the share of one store’s catchment that also falls inside another store’s catchment, reported as a percentage of area. Area is the weak form of the measure. A 25% figure over farmland and a 25% figure over a dense commuter corridor describe two different amounts of money, so weight the shared ground by the households or customer records inside it before quoting the number to anyone.
Format decides it, across roughly an order of magnitude. A convenience store draws from about a mile, or five minutes of driving. Everyday goods rarely pull anyone from beyond 15 to 20 minutes. Specialty and destination retail reaches 5 to 10 miles and sometimes further. Applying one 3-mile ring across those three formats gets it wrong for at least two of them.
Both, in that order. Radius circles take seconds to draw and are good enough to say which pairs of stores deserve a closer look. The polygon is what the decision gets made on, since it follows the road network, extends down a highway and stops where a barrier stops the driver. A site package that presents the circle as the analysis will not survive a careful reading by anyone on the other side of the deal.
The Huff model, adapted from the gravity model by David Huff in 1963, assigns every neighborhood a probability of shopping at each store rather than drawing hard boundaries. A store’s draw rises with its attractiveness and falls as travel time rises. Adding a nearby sibling store lowers every surrounding household’s probability of visiting the original one.
Net incremental sales is the portion of a new store’s revenue that represents genuinely new demand rather than demand transferred from existing locations. It is the figure that decides the deal, because a store that only relocates existing revenue adds operating cost to the network without adding much to the top line.
No. Overlap is accepted deliberately to block a competitor from a corner, to relieve a store that is already turning demand away on capacity, or to run a differentiated format such as a pickup unit beside a full-service one. The distinction that matters is timing. Modeled during site selection, the transfer goes into the opening plan as a budgeted cost. Found later in a quarterly comparable-sales report, it has to be explained after the money is committed.
Franchise encroachment is a franchisee’s claim that the franchisor or another franchisee opened a location close enough to cause a measurable sales decline. In almost every US state there is no statutory protection against it, so the franchisee’s position rests on the territorial language in the franchise agreement and, in some cases, on the implied covenant of good faith and fair dealing.
