Saltar al contenido
Volver a guías

BlackPrint · Guides · Site selection

Sales cannibalization: what it is and how to measure it

The share of a new store's sales that comes out of your own stores. The map usually says it is huge; measured with how people actually move, it is almost always much smaller. Here is the method, with numbers from measured cases in Monterrey and Querétaro.

Updated November 10, 2026 · BlackPrint Team

91.7% → 0.7%cannibalization risk on the map, against the risk measured with real behavior

The essentials, in 20 seconds

  • Cannibalization is the part of a new store's sales that is not new: revenue taken from your own nearby stores
  • Measured on a map it misleads: in a Monterrey study, geometric overlap gave 91.7% and real visits 0.7%
  • The investment case goes against incremental sales — gross minus what is transferred — not against gross

01 · The missing number

The sales you take from yourself almost never make it into the investment case.

When a site is evaluated, three things are almost always measured: the traffic passing in front, the demand in the area and the competition already there. The fourth is missing — and it is the one that decides whether the project pays.

4thquestion of the investment case: the one almost nobody measures

The four questions about a site · the fourth is rarely presented

Flow

How many people and cars pass in front, by hour and by day.

Demand

How many households and how much category spend sit in the catchment.

Competition

Who sells the same thing nearby, and how far away.

Cannibalization · the fourth

How much of that sale you will take from yourself. It almost never appears in the investment case.

▲ THE READ
  • It is missing for two reasons. The number looks bad and nobody wants to present it; and measuring it properly requires knowing where your current stores' customers really come from — data that is not in the point of sale.
  • What follows is predictable: the opening is approved on projected gross sales, a year later it is judged against consolidated sales for the market — which did not grow as modeled — and nobody knows exactly why.

02 · Gross vs. incremental

With the same gross sales, the worst candidate leaves the chain half of what the best one does.

Gross sales are everything the new store's register will record. Incremental sales are what the chain gains on top: gross minus what was transferred from your own stores. An opening that projects 100 and transfers 60 is not a 100 opening — it is a 40, and the investment has to be judged against that 40.

With identical projected gross sales of 100, a high-traffic corridor added 82 in incremental sales and a residential site only 41, because 59 was transferred from the chain's own store (BlackPrint method example).

41 of 100what the chain keeps in the worst candidate, against 82 in the best

Three candidates projecting the same gross sales (100) · what each adds and what comes out of your stores

CandidateTransferred from own stores ←→ Incremental sales
High-traffic corridorincremental 82 · the best
18
82
Mixed-use areaown store 1.4 km away · −21 pts
39
61
Residential areaown store 700 m away · −41 pts
59
41
▲ THE READ
  • All three project 100 in gross sales. Ranked by what they add, the residential site (41) is worth half the corridor (82).
  • The gap between best and worst is 41 points of incremental sales, with identical gross projections — an investment case built on gross presents all three as equivalent.
  • The rule: incremental = gross × (1 − transfer). Transfer is not guessed: it is estimated with the weighted overlap in the six steps below.

Illustrative example with indexed figures to explain the method. Not a client case.

03 · The overlap ladder MEASURED CASE

The map said 91.7%. People's actual behavior said 0.7%.

Overlap can be measured four ways, and each gives a different number for the same case. The first is the easiest to compute and the most misleading. This is how the four came out in a site study for a home-improvement chain in Monterrey, July 2026. The last three use a mobility panel: an anonymous, aggregated sample of mobile phones.

In a 2026 Monterrey site study, map-based overlap put cannibalization risk at 91.7%, while actual shared visits measured with mobility data were 0.7% (BlackPrint).

91.7% → 0.7%two orders of magnitude when behavior replaces geometry

Four ways to measure the same cannibalization risk · from most theoretical to most real

Level 1 · What the map saysof estimated spend within 20 min that falls in territory already covered
91.7%
Level 2 · Where the customers liveof the nearest store's visitors live in the site's 15-min area
26.3%
Level 3 · Who would have it closerof those customers would reach the new site faster
11.8%
Level 4 · Actual shared visitsof the current store's visitors also visited the site's surroundings
0.7%
▲ THE READ
  • Each rung measures something different, which is why it drops: it is not the same number corrected four times. From level 1 to level 4, 99% of the apparent risk disappears.
  • The first number is the ceiling of the risk, not the risk: it assumes everyone who could reach the new site stops going to the current store. Nobody behaves like that.
  • Only the last one describes behavior. A committee that decides on the first number cancels openings that made sense.

Level 3 rises to 25.9% if travel-time ties are counted.

04 · Your network's floor MEASURED CASE

Your stores already share 13.3% to 16.8% of their customers. That is the floor, not zero.

One comparison almost nobody makes changes the reading completely: how much your existing stores already overlap with each other. In the same Monterrey study, the network already operated at that level of overlap — and worked. The question is not "does it overlap?" — it always does — but "does it overlap more than my own network already tolerates?"

In the same Monterrey study, existing stores already shared 13.3% to 16.8% of their customers; the candidate site's 26.6% overlap was 1.6 times that internal maximum (BlackPrint, 2026).

1.6×the new site's overlap against the network maximum: 26.6% vs. 16.8%

Overlap of origin areas · the new site against what the network already tolerates

Candidate site ↔ nearest store1.6× the internal maximum
26.6%
Existing stores with each other · maximumthe ceiling of normal in this network
16.8%
Existing stores with each other · minimumthe floor of normal in this network
13.3%
56,791households in the buying profile within a 10-min drive
$207.6 MMXN total yearly market: households plus passers-by
3,213phones a month in front of the site, vs. 3,162 at the network's best store
Top 4%among the zones with most car traffic in the country
▲ THE READ
  • A 26.6% overlap sounds high only against zero. Against the 16.8% the network runs on every day, it is 1.6 times the internal maximum: high, measurable and decidable.
  • The site moves more passers-by than the network's best store, measured the same way. That, plus 0.7% shared visits, was the full case for opening.

05 · Two kinds of demand

Passers-by barely cannibalize. Households do.

This is the part that changes the result most and gets overlooked most. Resident demand comes from households living inside the catchment; pass-by demand comes from the flow already moving past the point — people who were going that way anyway and buy out of convenience.

2kinds of demand that do not cannibalize the same way

Resident vs. pass-by demand · how each splits and how much each cannibalizes

Resident demand · cannibalizes a lot

Comes from catchment households and is split by closeness to home. Two of your stores reaching the same households fight over the same pesos. Measured with households, income tier and spend by block.

Pass-by demand · cannibalizes little

Comes from the foot and vehicle flow that already passes and is split by access and visibility. These trips already happen and belong to no store. Measured with foot and vehicle counts by hour and day.

▲ THE READ
  • A unit on a high-traffic corridor can open relatively close to one of your own without transferring much.
  • Two stores in purely residential areas fight over the same households even when further apart.

06 · The method METHOD

Six steps from gross sales to incremental sales.

Where people come from today, where the new site would reach, what sits inside the overlap and how much spend is leaking out. In that order.

6steps, from the current store to incremental sales

How it is measured, step by step

01 · Measure the real catchment of each current store

Not with a radius: with the observed origin of the people who actually visit each store, summarized in distance percentiles.

02 · Model the candidate's reach with isochrones

Curves of equal travel time on the real road network, walking and driving. Asymmetric by definition: they stretch along corridors and stop at barriers.

03 · Compute overlap weighted by demand

Geometry alone is of little use. Each census block in the overlap is weighted by households, income tier and estimated category spend. A 30% overlap in area can be worth 8% or 55% of demand.

04 · Separate resident from pass-by demand

The flow already passing the point belongs to no store and cannibalizes much less.

05 · Estimate transfer and compute incremental sales

Projected gross sales minus what is transferred. That is the number for the investment case.

06 · Validate against consumption leakage

If much of the spending born in the area leaks out, there is a local capture gap and the estimated transfer goes down.

▲ THE READ
  • The step that moves the result most is 03: the same overlapping area can be worth 8% to 55% of demand.

07 · Consumption leakage MEASURED CASE

When 79.4% of spending leaves the area, a new store recovers sales instead of cannibalizing them.

Leakage is the share of consumption that originates in an area and is spent outside it. When it is high, local demand exists but nobody is capturing it from inside.

In a measured site in Peñaflor, Querétaro, 79.4% of local consumption was spent outside the area, so a new store would recover leaked spend rather than cannibalize (BlackPrint, 2026).

79.4%of the area's consumption is spent outside it

Leakage and reach of a measured site · Peñaflor, Querétaro

79.4%of consumption born in the area is spent outside
≤0.48 kmdistance half the visitors come from · median
6.53 kmhow far the 90th percentile stretches
48%of origin households are tier C
▲ THE READ
  • A new store there recovers spending that now goes elsewhere, instead of taking it from your own.
  • Two areas with the same households and income support different store densities if their leakage differs. The question shifts from "is there room for another store?" to "how much spend is leaking, and where to capture it?"

08 · The reverse case PUBLISHED CASE

It wasn't saturation: 168 points were missing from the list.

One scenario gets confused with cannibalization and is its opposite: the chain never knew how many points there were. A pharmaceutical distributor in Querétaro measured coverage against its own account list: 260 registered points, 66% of them visited each cycle. By its own dashboard it was doing well. The territory had 428 pharmacies operating.

A pharmaceutical distributor in Querétaro had 260 points on its list while 428 pharmacies operated; adding the 168 missing ones raised coverage from 66% to 84% and sales by about $840k MXN a month (BlackPrint).

168points the account list never saw: 39% of the universe

Account list against territory · points of sale

Detected in the territorypharmacies operating · full universe
428
Registered account listown customer list · 61% of the universe
260
Outside the list39% of the real universe
168
260 → 428points: list vs. detected · +65% universe
66% → 84%visit-cycle coverage · +18 points
168new points outside the list, all visited
+$840kMXN incremental monthly sales · ≈ $10 M a year
▲ THE READ
  • Detection found a universe 65% larger than the account list: 168 of 428 points were not on it.
  • Not an execution problem but a denominator problem — and the 18 points of coverage came without adding staff or route days.
  • There was no cannibalization because the new points belonged to nobody. Rule this scenario out before declaring a market saturated.

Published BlackPrint pharmaceutical-distribution case, Querétaro. Same team, same route days.

09 · When it does pay

Cannibalizing on purpose can be the right call. Cannibalizing without knowing, never.

Cannibalization is not a defect in itself. In three cases an opening with high transfer is the right decision.

3cases where transferring sales pays

Three reasons to open even if it transfers sales

Defending the market

If you don't open on that site, a competitor will — and you lose the sale without recovering anything.

Decongestion

A store at capacity loses sales to saturation: queues, stock-outs, full parking. Part of what was "transferred" was already being lost.

Service level

When the customer's travel time is the real barrier, moving closer expands the reachable market even with overlap.

What never pays

Cannibalizing without knowing. The opening was justified with a number that didn't exist, and the error surfaces only after the lease is signed.

10 · The inputs

Five data sources — and only one lives inside the company.

What it takes to measure cannibalization with behavior instead of geometry.

5inputs to measure with behavior

What data is needed

Visitor origin

Real catchment of each current store. Aggregated and anonymous.

Road network

Walking and driving isochrones on the network that exists, not on a circle.

Households and income tier by block

Weight the overlap by real demand, not by area.

Foot and vehicle counts

Size pass-by demand and its rhythm by hour and day.

Internal company data

Stores, sales and ticket, to calibrate the model.

Frequently asked questions

What is sales cannibalization?
The share of a new store's sales that is not new for the chain: revenue taken from your own nearby stores. The investment case should be built on incremental sales — gross sales minus that transfer — not on gross sales.
Why does map overlap overstate the risk?
Because it assumes everyone who could reach the new site will stop going to the current store. In a 2026 Monterrey study, map overlap gave 91.7%, while actual shared visits measured with mobility data were 0.7%.
How is cannibalization measured with real behavior?
With the observed origin of current stores' visitors (aggregated, anonymous mobility data), isochrones for the candidate, overlap weighted by households and spend, separation of resident from pass-by demand, and a check against consumption leakage.
What should a new site's overlap be compared against?
Against the overlap your existing stores already have with each other. In the Monterrey case, the network ran at 13.3% to 16.8%; the candidate's 26.6% was 1.6 times that maximum — high, but a decision, not an alarm.
What is the difference between gross and incremental sales?
Gross sales are everything the new store records; incremental sales are gross minus what was transferred from your own stores. A store projecting 100 that transfers 60 adds only 40 to the chain.
Why does pass-by demand cannibalize less?
Because it comes from trips that already happen and belong to no store; it is split by access and visibility. Resident demand comes from the same households, so two nearby stores fight over the same spend.

Your territory

This analysis covers the city. Yours is about your corner.

If you have a decision in front of you — where to open, which site to lease, how much competition you walk into — we measure it with data for your exact area, not with city averages.