GIS site selection: how to find the right site when you don't have one yet

Most of the GIS work around sites assumes you already have one. A landowner or an agent brings it to you, and the question is whether it stacks up. That is a site appraisal, and we will cover it in next week’s post.

Site selection starts a step earlier. You know what you need, whether that is a venue with space for your units, a plot for a small housing scheme or a building near enough to your customers, but you do not know where it is. The work is finding the few locations that fit out of the hundreds or thousands that could, and doing it in a way the people signing off the decision can follow.

This post sets out how that works in GIS, step by step, and where it tends to go wrong.

Start with the brief, not the map

The criteria for a site come from how the business makes money, so the first conversation is about the operation rather than the data. What does a good site do for you that a poor one does not? Where have your best and worst sites been so far, and what was different about them?

The answers then have to be turned into things that can be measured. “Near enough to our customers” becomes a count of adults within a fifteen-minute drive. “Easy to get to” becomes the road distance to the nearest junction or station. “Enough space” becomes a minimum floor area or plot size. Each one becomes a layer or an attribute that can be calculated the same way for every candidate, which is what makes a comparison fair.

The criteria then split into two groups. Hard constraints rule a site out entirely: a minimum plot size, a flood zone you will not build in, a designation that makes your use unworkable. Preferences make a site better or worse without ruling it out, such as a larger catchment, better visibility or a lower rent. Projects lose time when that split is never made and everything is treated as a must-have, because the shortlist comes back empty, or when the preferences are never weighted and the ranking becomes an argument after the fact.

It also helps to write down where each threshold came from. A fifteen-minute drive time chosen because it matches your existing customers’ travel is defensible; one chosen because it sounded reasonable will be questioned the first time a promising site falls just outside it.

Work out what is available

This is the step most site selection projects skip, and it matters more than any of the scoring that follows. The datasets used to screen land describe what is true about it: its flood risk, its designations, its boundaries, who owns it. None of them says whether you can have it. A ranking of land that is not for sale, to let or open to an approach is a well-made map of places you cannot go.

There are two main ways to build a candidate list that avoids this.

The first starts from what is on the market. Commercial listings, agent instructions and auction catalogues all give a dated, unambiguous signal that a site is available, and for operators placing units in other people’s premises, the list of host venues serves the same purpose. The list is narrower, because it misses anything off-market, but everything on it can be had.

The second starts from the land and works back to the owner. HM Land Registry’s INSPIRE Index Polygons show the indicative extent of registered freehold titles,¹ and its dataset of UK companies that own property in England and Wales lists titles held by companies and corporate bodies, although it excludes land owned by private individuals and overseas companies.² For any individual title, the register itself costs £7 to download.³ This is closer to how land promoters work, approaching owners whose land fits the criteria whether or not it is listed. It reaches further, but it brings obligations the screening work does not: writing to a named private landowner is direct marketing, so UK GDPR applies to how their details are found, stored and used.⁴

Sometimes the list already exists, because the client brings their own estate, a set of venues they have relationships with or sites put forward by agents. That is fine, as long as everyone knows the shortlist can only be as good as the list it was drawn from. Whichever route you take, settle it before any scoring starts, because it changes what the shortlist means.

Rule sites out first

Hard constraints work as a sieve. Each one is applied to every candidate, and any site that fails is set aside with the reason recorded. Running the most decisive and cheapest checks first keeps the later, slower work focused on fewer sites.

Flood risk is a common example. National planning guidance steers development towards areas at lower risk of flooding through the sequential test,⁵ so for many uses a site in Flood Zone 3 is either ruled out or carried forward with a clear flag. Designations such as Sites of Special Scientific Interest, scheduled monuments or Green Belt play a similar role, depending on what you plan to do.

The list of rejected sites is worth keeping. Criteria change: the minimum size drops, the budget rises or a new format of unit needs less space. When that happens, you can re-run the sieve against the sites you already have rather than starting again, and you can show why a site someone remembers from an early meeting did not make the cut.

One rule matters more than any other here. Where data is missing, the site should be marked as unknown, not as free of the constraint. A local authority that has not published its conservation areas to the national platform has not removed them.

Score what is left

The sites that pass the sieve are then ranked against the preferences. Each criterion is converted to a common scale, so that a catchment of 40,000 adults and a road distance of 800 metres can be compared, and each is given a weight that reflects how much it matters. The weighted scores are added together to produce a ranking.

Two details make the difference between a ranking that holds up and one that does not. The first is measuring distance the way people travel. Census 2021 output areas, which hold between 40 and 250 households each,⁶ are small enough to count the population within a drive time or walk time along the road network. A straight-line buffer cannot see that a river or a motorway puts a site that looks close a long way from the people it needs to serve.

The second is testing the weights. Adjust them within a sensible range and watch what happens to the top of the list. If a small change reshuffles the top ten, the ranking is less certain than it looks, and the decision should rest on the analysis of individual sites rather than the score. Sites that stay near the top across a range of reasonable weightings are the safest picks, and showing that to the people signing off the decision is more persuasive than a single number.

Look at every site on the shortlist

A score tells you a site met the criteria. It does not tell you that the car park is now a building site, that the access road is a private track or that the neighbouring use would put your customers off. Datasets lag behind what is on the ground, so every shortlisted site needs a review against recent aerial imagery and, where it helps, street-level photography.

The review works best when every site is checked against the same questions and the answers are written down. A short analyst note against each site, saying why it scored the way it did and what is still uncertain, is often more useful to the people making the decision than the score itself. It is also the part a self-serve screening tool cannot give you, because it depends on someone having looked.

How it changes by sector

The method stays the same across sectors, although the criteria and the candidate list change.

Leisure and wellness operators placing units at host venues usually start from a list of venues, so availability is largely settled and the work is scoring catchment, access and the physical space each venue has. Retail and food operators planning a rollout need the same catchment analysis, with one addition: their own existing sites, because a new site that takes customers from an existing one adds less than its score suggests.

Housebuilders and land buyers sourcing across a patch rely more on ownership and planning history. Brownfield land registers, which every local planning authority in England must publish and review at least once a year,⁷ are a useful source of candidates, and previous refusals and appeal decisions on nearby sites show what the authority is likely to accept.

Data centres and other power-hungry infrastructure reverse the usual order, because grid connection is usually the deciding constraint. Screening starts from substation capacity and works outwards; our data centres page covers how we support that work.

Keep the shortlist current

A shortlist delivered as a spreadsheet starts to go stale almost immediately. New sites come to market, others are let or sold, and the team’s view of each site changes as conversations happen. If those decisions live in email threads, the reasoning behind them is lost by the time anyone asks.

For anything beyond a one-off exercise, a shared web map works better. Each site carries its score, its analyst note and a status that the team updates as they go, whether that is to approach, under discussion, dismissed or progressing to appraisal. We have built this for a venue-sourcing operator, who marks each shortlisted venue to approach or dismiss directly on the map, and for an infrastructure site-finder whose candidates move through screening stages towards detailed assessment. In both cases the decision sits on the map next to the evidence for it, which makes it easy to revisit when circumstances change.

Where site selection ends

Site selection produces a shortlist, not a decision. Its output is evidence against your criteria, set out clearly enough that the people responsible can choose between the finalists with confidence. Those finalists then need the detailed work: a site appraisal to pull together everything known about each one, followed by a site visit, surveys, a solicitor’s review of title and, where relevant, pre-application advice from the local planning authority.

What site selection changes is how much of that expensive work is spent, and on which sites. Instead of commissioning surveys on whatever came through the door first, you commission them on the three sites that fit best out of two hundred, and you can show why.

How we approach site selection

Our site selection case study shows a working site selection application built in three weeks, and our post on renewable energy site selection walks through the same two-pass method of sieving and scoring for solar. Our property and land page sets out how we support developers, landowners and operators, and the Knowledge Hub covers the underlying GIS methods. If you are new to the open datasets mentioned above, our post on open data in the UK is a good place to start.

If you know what you need from a site but not where to find it, get in touch.

References

1. HM Land Registry. Use land and property data: INSPIRE Index Polygons.
2. HM Land Registry. UK companies that own property in England and Wales.
3. HM Land Registry. Changes to fees for HM Land Registry’s information services.
4. Information Commissioner’s Office. Direct marketing and privacy and electronic communications.
5. Department for Levelling Up, Housing and Communities and Ministry of Housing, Communities and Local Government. Flood risk and coastal change.
6. Office for National Statistics. Census 2021 geographies.
7. Ministry of Housing, Communities and Local Government. Brownfield land registers.