Skip to content
Shyenova

Networks

Why a radius is not a catchment

Drawing a ten-kilometre circle takes ten seconds and is wrong in a specific, predictable direction. The lease it justifies runs a decade.
Published
21 May 2026
Reading time
6 minutes
Topic
Networks
Author
Shyenova

The standard way to assess a site is to draw a circle around it, count the population inside, and call that the catchment. It is fast, it is intuitive, and it embeds an assumption that is false almost everywhere: that people can travel in a straight line in every direction at equal speed.

They cannot. A river with no crossing for six kilometres removes half the circle. A highway with restricted turns removes a quarter. A one-way system makes the morning catchment differ from the evening one. And a competitor sitting two kilometres closer to a dense pocket inside your circle takes that pocket regardless of what the geometry says.

The error is systematic, not random

If radius error were random it would matter less. It would average out across a portfolio of decisions. It does not. Circles overstate catchment almost everywhere, because barriers subtract and nothing adds. The overstatement is largest exactly where it hurts most: in dense urban geographies with rivers, rail lines and controlled-access roads, which is where the expensive sites are.

MethodPopulation reachedvs drive-time
10 km radius412,000+38%
15-minute drive time298,000,
15-minute, net of nearer competitor186,000−38%
Illustrative comparison for the same candidate site.

The second line is what the road network permits. The third is what a demand model allocates once alternatives are accounted for. Between the first and third figures sits a factor of more than two, and the decision is usually being made on the first.

What to use instead

  1. 01Drive-time isochrones over the actual road network, rather than radii. This alone removes most of the error and requires no proprietary data.
  2. 02Calibration against where your existing customers genuinely come from. Geocode your own transaction data and compare the observed catchment to the modelled one. This is the single most informative input and the one most often unused.
  3. 03A demand allocation model rather than nearest-site assignment. Gravity models allocate by relative attractiveness and distance, which is closer to how people actually choose.
  4. 04Cannibalisation, subtracted explicitly. The relevant number is never how much a new site will do; it is how much it will do that you were not already getting.

The number that is usually missing

Of those four, the last is the one most often absent from a site paper and the one most likely to change the decision. In an established network, ten to thirty per cent of a new site's revenue is commonly transferred from sites you already own. A candidate projected to add twenty per cent may add four.

Gross revenue justifies the lease. Net revenue justifies the investment. Site papers routinely present the first and describe it as the second.

None of this requires exotic data. Road networks are freely available and good in urban India, isochrone routing is a solved problem, and the most valuable input (where your existing customers live) is sitting in your own transaction records. What it requires is the decision that a ten-second circle is not adequate justification for a ten-year lease.

Start a conversation

Tell us what is not working.

Most useful first conversations are thirty minutes and involve no proposal. Describe the problem you are looking at and we will tell you honestly whether it is one we can help with.

Or write directly to hello@shyenova.com