Every benchmark we publish comes from a collection run we can point at. This page describes exactly how those numbers are produced, what we exclude and why, and what the figures cannot tell you. If something here does not justify a number you have seen on one of our pages, that is a bug and we want to hear about it.
For each category we run a Google Maps search of the form “{category} in {city}” across 40 US cities, signed out, twenty results deep in each city. The city list is fixed so that quarter-on-quarter comparisons are comparable, and it spans both the largest metropolitan areas and mid-sized cities, because those turn out to behave differently.
Results are collected through the DataForSEO SERP API rather than by scraping, and the ratings, review counts and primary categories we report are Google’s own displayed values, not our interpretation of them.
Collection is repeated quarterly. Any benchmark whose most recent reading is more than 180 days old stops rendering on the page that cites it, rather than sitting there looking current.
A Maps query for a category does not return only that category. A search for a dentist returns dental insurers, practice-management consultancies and the occasional optometrist. They rank, and they are real competitors for the click, but they are not the bar a practice has to clear. Averaging them in drags every figure down and makes the benchmark useless for the person reading it.
So each listing is checked against the primary category Google itself assigns, and kept only if it belongs to the vertical. Every page reports how many listings were excluded on that basis, so you can see the size of the effect rather than take our word for it.
Sponsored listings are stored and counted separately. They never enter a median, a percentile or a percentage. We report how many cities showed ads at the moment we looked, and nothing stronger, because Maps ads are an auction that turns over through the day and one snapshot per city cannot support a claim about the rest of the week.
Listings with no displayed review count are excluded from review statistics but counted in the sample, and both numbers are shown.
Review counts in a local category are heavily skewed. A handful of listings with several thousand reviews will drag a mean well above anything a typical business would recognise. The median and the quartiles describe the distribution people are actually competing in.
For the same reason we do not publish minimums. A single junk profile anywhere across 40 cities sets the minimum, and one listing is not a finding. Where we describe the low end of a top-three position we use the 10th percentile, which one outlier cannot move.
Figures are reported nationally and split into two bands: cities in metropolitan areas above 2.5 million people, and everything below. Population figures are US Census Bureau CBSA estimates. A band with fewer than twenty listings is not reported at all, because two numbers side by side invite comparison and a thin one does not deserve it.
They are not a forecast. A benchmark says what businesses currently ranking look like. It does not say that reaching those numbers will make you rank, because it cannot separate cause from correlation. Proximity, which we cannot control for from a city-level query, does a great deal of the work.
They are a snapshot. One collection per city per quarter. Maps results vary by the hour, by the searcher’s exact position and by device.
They describe a city, not your street. A city-level query returns results centred on the city. Your own ranking varies across the radius around your location, sometimes dramatically, and only a grid measurement around your specific address will show that.
The scope, plainly. Every figure we publish comes from 153 categories measured across 40 US cities, collected 1 September 2026. Each category page states its own sample size and collection date, so you can check any figure against the run that produced it.
Category classification is Google’s. A business categorised oddly by Google is categorised oddly in our data too. Our filter can drop an insurer that calls itself an insurer. It cannot drop a consultancy that has set its primary category to the vertical.
If a number on one of our pages does not match what you can see, tell us. We would rather correct it than defend it.