Insights
Catchment areas

What a catchment area actually is, and why a radius is not one

The question. How far out should I be counting, and counting what?

Almost every site report opens with a circle. Five hundred metres, a ten minute drive, three miles. The circle gets drawn because it is easy to draw, and it survives because nobody downstream asks what it was meant to represent.

What it was meant to represent is where the customers come from. That is a question about behaviour: which way people walk out of the station, which road they will not cross, which centre they were going to anyway. None of those things are round.

So it is worth being precise about what a catchment is, because three different definitions are in daily use and people rarely say which one they are holding. Once that is settled, drawing a defensible boundary is straightforward. It is only slower than typing a number into a radius box.

A high street, with the part this page is about picked out.

The same street, and the boundary somebody drew around it.

In one sentence

Catchment area
The area a place actually draws its customers from, bounded by how people travel to it rather than by distance from it.

Written for

Three different things get called a catchment

The first is a share of trade. Primary, secondary and tertiary catchments split the area by how much of a place's turnover comes from each, and they are drawn backwards from data about who actually visits. This is the definition retailers mean, and it is the only one of the three that is about customers.

The second is a travel time. An isochrone is the area you can reach from a point within some number of minutes, on foot, by bus or by car. It is a supply-side statement about the network, not a demand-side statement about people, and it makes no claim that anyone in it is coming.

The third is an administrative boundary: a ward, a set of census areas, a town centre boundary from a local plan. These get used because the population statistics arrive already cut that way, which is a reason about the data rather than a reason about the place.

Most disagreements about a catchment are two people using two of these three. Say which one you mean in the first sentence, and if you have to mix them, keep the counts separate rather than adding them together.

  • ONS statistical geographies

    Every national statistics office publishes the boundaries its own data is cut by, as downloads: ONS output areas and built-up areas in the UK, Census Bureau tracts and block groups in the US, Eurostat and the national institutes for the European equivalents.

    What it misses. Definitive for the boundary itself, and silent on whether it matches anything a customer does. These units are drawn for roughly equal population, so they cut across the middle of a shopping street without hesitating. Check the vintage before using one: boundaries are redrawn between censuses and the old and new versions are not comparable.

  • The local plan's town centre boundary

    Your council's planning policy pages, in the adopted local plan and its proposals map.

    What it misses. The boundary that actually governs planning decisions, so it matters regardless of whether it matches trade. It is also often a decade old and drawn around where the centre was, not where it has moved to.

A radius fails in a specific direction, not randomly

Two things break a circle. Barriers remove people who are inside it: a dual carriageway with one crossing, a river, a railway cutting, a gated estate, a hill that is fine downhill and not uphill. Desire lines add people who are outside it: a station exit that feeds one street, a school run, a bus corridor that stretches the real catchment a kilometre along a line and nowhere at all across it.

Because barriers subtract and desire lines add somewhere else entirely, a radius is not merely imprecise. It is biased. It counts population that will never arrive and misses the strip that already does, and the error does not cancel out when you make the circle bigger. It gets worse, because a bigger circle picks up more of the neighbouring centre that those people go to instead.

This is why a radius flatters a site. It reports a population, all of which is implicitly treated as reachable, and reachability is the thing it cannot see.

  • An isochrone tool

    Any routing service that draws travel-time areas rather than distance rings.

    What it misses. Knows the road and path network, which is most of the way there. It does not know that the underpass is avoided after dark, that the level crossing is shut twice an hour, or that the path is a private estate road.

  • OpenStreetMap

    Community-maintained map data, browsable in any viewer built on it.

    What it misses. The quickest way to see the barriers: rail lines, watercourses, severance roads and where the crossings are. Coverage of paths and crossings is excellent in some towns and thin in others, so treat an absent path as unknown rather than as absent.

  • The national mapping agency's open data

    Ordnance Survey in Great Britain, IGN in France, the USGS and state GIS portals in the US, and the equivalent agency elsewhere. Most publish some geometry openly under their own licence terms.

    What it misses. Reliable geometry for the things that do not move, and the licence is the first thing to read rather than the last. The open products are deliberately generalised, so fine detail like a specific crossing is not always in them, and none of it says anything about behaviour.

  • Your own feet

    One hour on a weekday and again on a Saturday.

    What it misses. The only source that shows where the footfall stops. Slower than everything else here and worth more than most of it.

Draw the boundary, then justify every edge

The method that works is short. Start from an isochrone at the travel time that matches the trip you care about. Then move every edge by hand, and write one sentence per edge saying why it is where it is. The sentences are the deliverable as much as the shape is.

If an edge has no sentence, it is still a radius, and you have simply drawn it more slowly. The test is whether somebody who disagrees with the boundary can find the sentence they disagree with, argue about that, and leave the rest of the shape standing.

Record the trip the boundary is for, at the same time. A convenience catchment and a Saturday comparison-shopping catchment for the same unit are different shapes, and a boundary with no stated trip will get reused for the wrong one within a month.

Only then count what is inside it

Counting comes last, because every number downstream is a count of what falls inside the line. Get the line wrong and each figure is confidently, invisibly wrong by the same amount.

There are two sides to count. The demand side is people: resident population, and separately the workday population, because a centre with forty thousand commuters and eight thousand residents trades nothing like the reverse. The supply side is what is already trading, by category, as shares rather than raw counts, because a share is the only version comparable with another area.

  • The census, down to small-area geography

    England and Wales ran theirs in 2021, with Scotland and Northern Ireland on their own timetables; the US runs the decennial census plus the annual American Community Survey; most European countries run a register-based census through their national institute. All of them publish resident population and some form of journey to work.

    What it misses. The most reliable population count that exists, and a snapshot that recedes every year, so check how old the one you are using is. Several 2020 and 2021 rounds were taken under pandemic restrictions, which visibly moved the travel to work and student location answers.

  • The property tax roll

    The Valuation Office Agency's rating list in England and Wales, the Scottish and Northern Irish equivalents, and the county assessor's parcel records in most of the US. Each carries a description and a value per property.

    What it misses. Excellent for what a unit is and roughly what it is worth. It is a tax record, so it lags reality, describes the property rather than the operator, and its categories serve assessment rather than retail analysis. Verify a sample against the street before trusting the list.

  • The company register

    Companies House in the UK, the Secretary of State register in each US state, the Handelsregister in Germany, and so on. Most jurisdictions publish company name, officers and some filing history in some form.

    What it misses. Tells you who is actually behind a fascia, including when six shops turn out to be one operator. What it holds varies enormously by jurisdiction, it says nothing about an unincorporated trader, and a registered address is very often an accountant's office rather than the shop.

The short version

Where this goes wrong

One catchment reused for every question

The area a convenience store draws from and the area a cinema draws from are different by an order of magnitude, and the area a scheme's own staff come from is a third shape again. A boundary saved in a file with no note about the trip it was drawn for will be reused for all three.

Letting the data pick the boundary

Population statistics arrive cut into census geographies, so it is tempting to build the catchment out of whole census areas. That decision is made by the shape of the download, and it routinely puts a whole neighbourhood on the wrong side of a road nobody crosses. Draw the shape you mean, then apportion the statistics into it.

Drawing it once and never walking it

Boundaries age quietly. A new crossing, a closed footbridge, a bus route moved two streets, a competing centre opening: each moves an edge, and none of them announce themselves in a dataset. Walk the boundary when the answer starts to matter.

Now do it

Or have it run for you

An area you drew, counted the same way every time you ask

  • The boundary is a shape you draw on the map, so barriers and desire lines are yours to reflect rather than a radius to accept.

  • Every business inside that shape is collected with its category, rating and review volume, which is the supply-side count without the afternoon of searching.

  • Category shares are computed rather than eyeballed, so the same boundary is comparable against another area and against itself next quarter.

  • The boundary is stored with the project, so a repeat run counts the same ground rather than an approximation of it.

What is actually inside the line I drew?

Area Analysis

The full mix by category with shares, the rating distribution, and the themes that recur across the operators trading inside your boundary.

How does this compare with somewhere I know?

Gap Analysis

Your area against up to three others on mix, performance, customer voice and momentum, which is what makes a share meaningful rather than merely precise.

Draw your area and see it

Sign up free. The first area is on us, and it takes a couple of minutes to set up.

Get started free

More insights