Cannibalization analysis software for retail site scoring

Open a store close to one you already run and some of its sales will come from customers who used to drive to the older location. That portion of the forecast was already yours. Cannibalization analysis software exists to size the transfer before a lease is signed, so the number you carry to committee reflects demand the network did not already have.

The math underneath varies by vendor, but the workflow is fairly consistent:

• It draws a drive-time or walk-time trade area for the candidate and for each nearby store, following the road network so that a river or an interchange counts the way it does on the ground.
• It locates where the candidate trade area falls inside an existing one, then weights that shared ground by the demand living and spending inside it.
• It converts the overlapping demand into an estimate of the sales that will shift from the older stores to the new one.
• It splits the forecast into demand that is new to the network and demand that is merely relocated, and flags the existing store that takes the biggest hit.

Most tools begin with trade-area overlap, the share of the candidate trade area that sits inside an existing store’s trade area. Overlap computes fast and maps cleanly. It answers a question about geography, though, when the question that matters is about revenue. Two trade areas can share a wide stretch of farmland where almost nobody shops, or they can share one dense commercial corridor that carries most of the sales. On a map those two situations can look nearly the same.

Transferred demand is what a committee is actually asking about: how many visits, or how many dollars, move from the stores you have to the store you are proposing. Weight the overlap by where customers and spend really sit and a 30 percent area overlap can mean almost nothing or almost everything. A tool that stops at raw area treats empty overlap and busy overlap as equals, so it can overstate the threat in one market and miss it in another.

• Travel-time trade areas drawn from the road network rather than radius rings, so physical barriers actually bend the boundary.
• Overlap weighted by population or spend, so a dense block of customers counts for more than an empty one.
• A forecast that reports net-new and transferred demand as separate figures instead of one blended total.
• A network view that scores the candidate against every nearby unit at once and names the store hit hardest, whether or not it is the closest.
• An export that lays out the inputs, the assumptions, and the estimate, so a committee can interrogate it.

Cannibalization features tend to sit in one of two camps. Turnkey scoring platforms wrap the trade-area math, the underlying data, and a forecast into a workflow you can run yourself. GIS platforms hand you the components and assume you will assemble the analysis. Neither is better in the abstract. They fit different teams, and different appetites for in-house spatial expertise.

• Geod. Runs a network-gravity model and breaks a site forecast into demand transferred from your own stores, demand captured from competitors, and demand that is net-new. It names the worst-affected existing store and exports an explainable PDF brief. Built to run without a GIS team.
• GrowthFactor. Automates site scoring, with its cannibalization read coming from trade-area overlap against your current network.
• SiteZeus. Forecasts site performance and folds cannibalization into its predictions and what-if scenarios.
• Buxton. A long-running analytics provider whose cannibalization view generally rests on analog and forecast-based methods built on customer data.
• Esri Business Analyst. Ships a Measure Cannibalization workflow that quantifies trade-area overlap inside a full GIS, which assumes some GIS fluency.
• CARTO. A spatial-analytics platform where you assemble the cannibalization model yourself from its data and functions, with all the flexibility and the work that implies.

A spreadsheet can approximate cannibalization with distance bands and a few rules of thumb, and for one obvious overlap that may be all you need. The cracks show fast. A spreadsheet cannot trace a real drive-time boundary, cannot weight overlap by where the demand actually sits, and cannot re-score the candidate against every neighbor each time the network shifts.

The harder problem is provenance. Ask a spreadsheet how much of the forecast is net-new and the answer is tough to source and tougher to reproduce six months later. Purpose-built software keeps the geography, the weighting, and the assumptions in one place and shows the steps behind the figure, which is what a reviewer ends up wanting.

Where many tools report overlap and leave the revenue read to you, Geod begins with a network-gravity model. It builds drive-time trade areas for the candidate and every nearby store, then allocates demand across them the way shoppers choose between stores in practice. That allocation lets it split the candidate forecast three ways: sales transferred from your own existing stores, sales captured from competitors, and sales that are genuinely new to the network.

The split is what separates knowing that two trade areas overlap from knowing what the new store costs the rest of the chain. Geod points to the existing store that loses the most, lays out the assumptions behind the estimate, and exports an explainable PDF brief a real estate committee can read and push back on. What you carry into the meeting is a number you can defend instead of a heat map you have to talk your way through.

The tool descriptions and the comparison table above summarize each vendor’s publicly stated capabilities, checked against their public product and documentation pages as of June 2026. Vendor features change, and the exact internals of any forecast are rarely published in full, so treat these as broad characterizations and confirm the specifics with each vendor before you rely on them.