Half the people around the median town live in 6 of its 25 map squares: why Google Maps ranking grids can misstate a dental practice's reach

Getting an NHS dental appointment is becoming slightly easier, but not for everyone. In the GP Patient Survey 2026 national results, carried out by Ipsos for NHS England from January to April 2026 and published on 9 July 2026, 81% of people who had tried to get an NHS dental appointment in the previous two years had managed it, up from 78% in 2025. Nearly one in five who tried still could not. People looking for a dentist often start on Google Maps, and Google Maps answers differently depending on where the searcher is standing. Agencies and software tools sell dental practices "local ranking heatmaps" to show this. This study asks whether those heatmaps are being read correctly.

The short answer is that they usually count the wrong thing. A heatmap counts map points. Patients live in some squares and not others.

"A field and a housing estate count the same on a ranking grid. They don't count the same in your appointment book," said Jack Henty, co-owner of Bollington Dental Practice and founder of Practice Radar.

What a rank grid is

A rank grid is a set of Google Maps searches for one term, such as "dentist", each made from a different point on a grid laid over the area around a practice. Each point records where the practice appeared when searched from there. The grids in this study are 5 points by 5, one mile apart. Each point stands for the one-mile square of ground around it, so the whole grid covers 5 miles by 5, reaching two and a half miles out from the practice in every direction.

Each square is given one of three results. Top three means the practice was in positions 1 to 3 on the map, the places Google shows first. 4 to 10 means it was on the map further down, or in the ordinary top ten results. Outside the top ten means anything else, including not found at all.

A typical report then says something like "top ten in 15 of 25 points, 60% coverage". That sentence treats every square as equal. We counted the people instead. Every neighbourhood in England has an official population estimate from the Office for National Statistics, and a point where its population is centred. Each neighbourhood's residents were added to the grid square that contains that point. The result is the share of residents, not squares, in each band.

1. People are bunched up around every town we measured

For each of the 56 English towns Practice Radar measures, we laid a 5 by 5 grid of one-mile squares around the town's centre point (a practice's address, the town's official centre, or the middle of its practices) and counted the residents in each square.

If people were spread evenly, the busiest 5 of 25 squares would hold 20% of them. In the median town they hold about 48%. The middle half of towns sits between 40% and 57%. The least concentrated, a London district, is still at 27%. In 43 of the 56 towns the busiest five squares hold more than 40% of the residents, and in 24 they hold more than half.

Put another way, half the people in the median town's grid live in just 6 of its 25 squares. In 50 of the 56 towns it is 8 or fewer. A point count gives the other 17 to 19 squares, where the other half of the people are spread thin, the same weight as the 6 to 8 that matter most.

Town size explains most of it. Around smaller towns the grid reaches out into countryside, so a few squares carry almost everyone. In cities and dense suburbs people are spread more evenly, and a point count is less wrong. But even the most built-up third of our towns sits well above an even spread.

For readers looking at their own region, the same measure by English region is below. Treat it with care: the towns are the ones Practice Radar happens to measure, 30 of them in the North West, and several regions have only two or three. Differences between regions mostly reflect which towns are in the sample, not the region itself.

2. Real grids, counted two ways

Bunched-up people only matter if they change what a grid says. So we re-counted real grids: four practices Practice Radar measures, five searches each ("dentist", "dental implants", "emergency dentist", "invisalign" and "teeth whitening"), run on 19 September 2026. That is 20 grids. In 3 of them the practice was outside the top ten on every point, so they read the same either way. That leaves 17 to compare.

In 9 of the 17, the share of residents in at least one band differed from the share of points by more than 5 percentage points. In 6 it differed by more than 10. The error runs both ways. The point count flattered the practice's top-ten reach in 11 grids and understated it in 6. At the extremes, one grid's point count overstated top-ten reach by 23 points, and another understated it by 13.

In these four practices' grids, the lumpier the map, the bigger the gap. Where people were spread fairly evenly across the grid, points and residents told almost the same story. Where most people lived in a handful of squares, the largest gap averaged about 10 to 12 points, against 1 to 4 where people were evenly spread.

3. The squares that matter are the busy ones you are missing

The useful question is not "how many squares am I green in". It is "where do lots of people live, and am I missing there". In the typical grid of the 17, the practice was outside the top ten in 4 of the 5 most populated squares. Those missed busy squares held a median 41% of everyone in the grid.

That is a smaller, sharper target than "improve the heatmap". A practice does not need to turn every square green. It needs to show up in the handful of squares where about half the people around it live.

"The question worth asking of any heatmap is how many people live in the squares where we show up. That tells you more than the percentage on the front page," said Jack Henty.

What to do about it

You do not need any particular tool to use this. Next time a grid or heatmap report lands, do four things.

1. Ask whether the headline figure is weighted by people. If "coverage" or "visibility" is a count of points, treat it as a rough picture, not a measure of reach.
2. Find your busy squares. Open the grid next to an ordinary map. Circle the squares covering housing estates, dense terraces and the high street. In most towns that will be 5 to 8 squares out of 25.
3. Judge the report on those squares alone. Where are you in the top three, where are you 4 to 10, where are you missing. That is your real position.
4. Point the work at the busy squares you are missing. Google says it ranks local results mainly by relevance, distance and prominence. You cannot move your building, but you can make your Google Business Profile complete and accurate, keep earning reviews steadily, and make sure your website names the areas you serve. Then check the same squares next month, for the same search, so any change is the ranking and not the grid.

A caution for anyone comparing grids: a bigger grid is not a better one. Stretching it further out adds squares of countryside, or other towns' patients, and a point count gives them the same weight as your own high street.

How we did this

Rank grids. Twenty Google Maps rank grids run by Practice Radar on 19 September 2026 for four practices it measures in England, five searches each. Each grid is 25 points, one mile apart, centred on the practice, with up to 60 map results read at each point. Only pooled results are reported. Practices are not named, located or described, and no grid is shown on its own except the two largest gaps, which carry no location. One further grid market, Practice Radar's fictional demonstration town, was excluded because its data is a copy of a real market.

People. Every 2021 Lower layer Super Output Area (a neighbourhood of about 1,500 people) whose population-weighted centre lies within 3.6 miles of the grid's centre, from the ONS population-weighted centroids, with its ONS mid-2024 population estimate from Nomis. Each neighbourhood is counted at its nearest grid point if it lies within half a mile of it in both directions, which is the one-mile square that point stands for. 3.6 miles covers the grid's furthest corners. This is the same rule Practice Radar's product uses. The populations matched Practice Radar's own stored area files on all 5,937 neighbourhoods both contained.

Towns. The 56 English towns Practice Radar measures, each with the same 5 by 5 grid laid around the town's centre point. Regions come from reverse geocoding each centre point with postcodes.io.

Controls. A neighbourhood is counted at a single point, so a large rural one lands in one square even if its people are spread across two. To check this is not what makes people look bunched, we repeated everything at Census 2021 Output Area level (about 130 households each, 39,691 areas). The median town's busiest-five share moved from about 48% to about 47%, and the median town changed by 1.8 points. The widest single change was the most rural town, from 100% to 79%. For the 17 grids, the counts became 8 rather than 9 moved by more than 5 points, 5 rather than 6 by more than 10, and 10 rather than 11 flattered by the point count. In 47 of the 56 towns at least one of the 25 squares has no residents at all even at that finer level. The exact median also depends on small method choices, such as how a neighbourhood on the edge of a square is assigned, which is why we give it as about 48%. We grouped the towns by size (table above), because a smaller place is expected to be more concentrated.

What this is not. Residents are not searchers. Commuters and shoppers search from town centres during the day, and older and younger people search at different rates. Counting residents is a far better weight than counting squares, but it is still a proxy for demand. Four practices' grids are a small sample: they show the size of the error that can occur, not an average for the country. The 56 towns are the ones Practice Radar measures, not a random sample, and 30 of them are in the North West.

For editors

Free to reuse with attribution to Practice Radar, under Creative Commons BY 4.0. Sources: ONS mid-2024 small area population estimates and 2021 population-weighted centroids, Census 2021 (control), GP Patient Survey 2026 national results (Ipsos for NHS England), and Practice Radar Google Maps rank grids of 19 September 2026. Practices in the study are anonymised and will not be named. Contact: hello@practiceradar.co.uk.

Practice Radar is the tool Jack Henty built to measure local dental search markets, and counting grids by people is how it now reports them.