What each number on the Explore screen means, how to read it, and where it stops being reliable. Every score described here is computed from the same data the product runs on.
Can this area support a fitness business at all, setting aside who is already operating here?
Demand is a composite of the population and demographic character of the area you have selected, normalized so a score means the same thing in a dense urban tract and a spread-out suburb. It is deliberately blind to competition. Mixing supply into a demand number makes both harder to reason about, and the two questions have different answers often enough that they are worth keeping apart.
The inputs are weighted rather than averaged, because they are not equally predictive. Raw population is the obvious one and the least useful alone: a large, spread-out, older area and a compact, younger one can hold identical numbers of people and support very different businesses. Each input also saturates, because past a point more of a good thing stops adding opportunity. Beyond a certain size the difference between a large catchment and a very large one barely matters to a single studio working on a ten minute drive time, because nobody is crossing the whole of it to reach you. That is why a genuinely excellent area lands in the 90s rather than at 100.
80+: strong. The population and demographic profile support a fitness business comfortably.
60 to 79: moderate. Viable, but concept and price point carry more of the risk.
Below 60: weak. The addressable population is thin for the format.
Read these as bands. A 72 and a 75 are the same answer. A 72 and a 45 are not.
How crowded is this area compared with everywhere else in the country that looks like it?
This is a percentile, not a count. A score of 50 means competition here sits at the national median for areas of similar population and density. Higher means more crowded than the typical comparable market, lower means less.
A raw count cannot answer the question an operator is actually asking. Twelve gyms is crushing in a small town and unremarkable in Manhattan, and no absolute number distinguishes those without knowing what normal looks like at that density. So the local ratio of actual to expected supply is compared against a reference distribution built from samples across qualifying metros, weighted by population so that dense markets do not dominate purely by having more people in them.
The reference distribution and your score are computed over the same geometry, deliberately, even when you are looking at a drive-time area on the map. Comparing a score measured one way against a curve built another way produces a number that looks precise and means nothing.
The radius is also smaller than you might expect, and that is a correction from experience. Built over a wide radius the curve flattened: every dense urban core came out near the top of the scale and the score stopped discriminating between genuinely different neighbourhoods. A tighter comparison restores the range where it matters most.
Near 50: a median market. Competition is what you would expect here.
Above 50: heavier competition than comparable areas.
Below 50: lighter competition than comparable areas.
High competition is not automatically bad. Operators cluster where the business works, and an empty market is sometimes empty for a reason. Read it against demand: heavy competition with strong demand is a fight worth having, while heavy competition with weak demand rarely is.
How many fitness locations are here, and how many would you expect for an area this size?
The competition line shows both numbers together, for example 86 nearby against roughly 41 typical. The first is a count of what we have classified inside your selected area. The second is what an area of this population would normally hold, derived from a per-capita rate computed across our own database rather than taken from a published industry figure.
Using our own data matters more than it sounds. Trade estimates are national, self-reported, usually a year or more behind, and they mostly describe revenue or membership. A site selector needs physical locations on the ground. A rate computed from the locations we have actually classified reflects the market as it is now and moves as the data refreshes.
This pair is the rawest thing on the screen, and often the most useful. The scores above compress it into something comparable across markets; this shows you the arithmetic they are compressing.
Which categories are underrepresented here, relative to what an area like this usually supports?
A category being absent is not by itself interesting. Most categories are absent from most places, because most places are small. The question is whether a category is thin relative to what an area of this size and character would normally carry, which means the same actual-against- expected comparison, run per category rather than in aggregate.
Each category line shows both figures, so a category reading 9 against 6.1 is oversupplied and one reading 1 against 2.0 is thin. Categories are flagged as underserved only when the shortfall is large enough to be worth acting on, rather than whenever actual dips below expected, because small differences at these counts are noise.
The most important caveat on this page. A category can be missing because nobody has tried it, or because somebody tried it and it did not work. The data cannot tell those apart, and the second is more common than operators expect. A gap is a question worth asking, not an answer.
It is also worth checking the gap against the demographics beside it. A format missing from an area whose population would never sustain it is not a gap you want to fill.
Provides: Population, age distribution, household income
Dataset: American Community Survey five-year estimates
Granularity: Census tract
Coverage: All 50 states and the District of Columbia
Provides: Businesses, addresses, ratings, categories, websites
Method: Commercial listings data, verified and classified in house
Coverage: All 50 states
Provides: The shape of a five, ten or fifteen minute journey
Method: Real road network routing, not a straight-line radius
Modes: Walk, bike and car
Provides: Which of 13 categories each location belongs to
Method: Brand and keyword rules first, a language model for the remainder, human review for anything low confidence
Why it matters: Gap analysis is only as good as the categories underneath it
A five mile circle and a fifteen minute drive are different shapes in every city with a river, a highway, or a downtown. Circles cross water nobody can drive over and ignore the interchange that puts a suburb eight minutes closer than the neighbourhood behind it. Members do not travel in circles, and the distinction decides whether the people you are counting can actually reach the door.
| Data | Refreshed |
|---|---|
| Demographics | With each annual Census release |
| Location listings | On search, then cached |
| Scores | Recomputed on search |
| Drive-time areas | On search, then cached |
Every model is wrong somewhere, and the useful thing is knowing where. This section is the one we would read first.
Parking, visibility from the road, co-tenancy, signage rights, which direction traffic runs at six in the morning, and the lease itself are not in this data, and they are frequently why a high-scoring location fails. These scores tell you whether the market is there. They cannot tell you whether the building is right.
Brand strength, price point and operating model are not inputs. An area scoring 65 can be a strong site for an established regional brand with a membership base to draw on, and a poor one for a new premium concept at the same address.
Four boutique studios with twelve reformers each and one forty thousand square foot big box count as five locations, and they are not five equivalent competitors for the same member. The data also cannot distinguish a thriving operator from one about to close, so a market that looks saturated is sometimes saturated with businesses on the way out.
Demographics are survey estimates, and their margins of error are widest exactly where they matter most: small areas and small populations. Below roughly a thousand people in the selected area, scores are suppressed rather than reported, because a number there would imply a precision the underlying data cannot support.
Most locations are categorized by deterministic rules and are effectively certain. The remainder are read by a model, and anything it is unsure about is queued for a human. A small number of edge cases are wrong at any given time, usually businesses that genuinely straddle two categories. If a category count looks wrong for an area you know well, it is worth telling us, because that correction becomes a rule.
Non-US locations are excluded at ingest rather than scored badly, because a partial international dataset is worse than none. Depth also varies inside the US: some metros carry pre-scanned competitive data while elsewhere analysis runs on demand. The coverage page lists which is which.
PlaceGap for Fitness v0.1.0-alpha