What Goes With Ranking in Google Maps? We Studied 795 Searches

As of November 2026. Search data pulled October 3-5, 2026. Every number in this post is computed by one script directly from the raw search responses and saved web pages, and the de-identified data is free to download at the end.

The short version: we studied 795 Google Maps searches across 8 kinds of local business. In the 337 big-metro searches where we measured it most fully, the businesses in the top 3 were, more often than the businesses ranked far below them, closer to the person searching, filed under the exact Google category being searched, and holding more reviews. Outside law firms, links showed little relationship with Maps rank. In 30 downtown searches, the top-3 business had more clean referring domains (links from real, non-spam websites) than the low-ranked businesses in 16. That is 53%, with a 95% interval of 36% to 70%, so we cannot tell it apart from chance.

Law firms were the exception. For personal injury and divorce lawyers, links had the clearest association with winning of any group, and that held up when we re-ran the test in ten more cities.

This is an observational study: it shows what goes with ranking, not what causes it. A business that collects reviews usually does other things well too. We say where the evidence is strong, where it is weak, and what we could not measure.

The numbers at a glance

All six rows come from the 15-metro dataset described below; the clean-link row covers the downtown searches in the first five metros only. For each search we compared the businesses ranked 1 to 3 with the businesses ranked 15 to 20 in the very same search.

What we measured

We pulled Google Maps results through the DataForSEO API, the way a person would see them from a given spot on the map. A search is one keyword searched once from one map pin. Maps rankings change with where the searcher stands, so we searched from downtown and from pins several miles out.

The study has two datasets.

• Dataset A, 15 big metros, 6 niches. "Personal injury lawyer," "divorce lawyer," "plumber," "roofing contractor," "dentist" and "real estate agent," in Chicago, New York, Los Angeles, Houston and Phoenix first, then in a pre-registered replication in Miami, Dallas, Atlanta, Denver, Philadelphia, Seattle, Tampa, Charlotte, Las Vegas and San Diego. We wrote down what would count as a pass or a fail before pulling the replication data. 337 usable searches, 6,708 listings.
• Dataset B, 4 niches from big metros to rural towns. "Roofing contractor," "plumber," "HVAC contractor" and "jewelry store" in the same 15 metros, 10 mid-size cities, 10 small towns and 30 rural towns drawn at random from the Census list of places. 573 usable searches, 11,173 listings.

The two datasets share 115 searches. Dataset B's big-metro roofing and plumbing searches are the same raw search results as dataset A's. Counted once, that is 795 unique searches (337 + 573 - 115) and 15,597 unique listings. Each chart and table below uses one dataset, so no search is counted twice inside any figure.

Who counts as a low-ranked comparison. We call them controls: businesses ranked far down the same search. The rule differs by dataset:

The short-pack rule in dataset B applies to 25 of its 573 searches (6 in big metros, 13 in small towns, 6 in rural towns), where Google simply shows fewer businesses.

For each listing we recorded what Google shows (category, reviews, photos, hours, distance from the searcher), what the linked website says, and the links pointing at that website. Paid Maps ads were removed, because an ad is not a ranking (12 in dataset A's six niches, 67 in dataset B).

The main statistic. Within each search, we ranked every listing on a factor (reviews, say) and measured how closely that order matched the Maps order, then averaged across searches. We call it the within-search score. 0 means no relationship; 1 would mean the factor lines up perfectly with rank. A dentist in Manhattan is never compared with a roofer in Phoenix. The 95% intervals resample whole keyword-and-city groups, because the pins of one keyword in one city are not independent.

1. A "#1 in Maps" belongs to a spot on the map, not a city

The business ranked #1 downtown almost never stayed in the top 3 a few miles away:

• In the first five metros, it kept a top-3 spot at pins 3 to 17 miles out 0 times in 60 checks (95% interval 0% to 6%).
• In the next ten metros, it kept it 0 times in 209 checks 7 miles out (0% to 1.8%), and 3 times in 24 at the 4-mile pins we used where 7 miles out was open water.

So when someone says a business "ranks #1 in Maps," ask: from where? The fair measure is a fixed grid of search points, checked the same way every month.

2. Distance tracks rank about as closely as reviews

In 276 of 337 searches (82%, interval 77% to 86%), the top 3 sat closer to the searcher than the controls did. Top-3 businesses were a median 0.87 miles from the searcher; controls in the same searches were 1.59 miles away. In the denser first five metros the gap was 0.35 miles against 1.18.

The within-search score for closeness was 0.28 (interval 0.25 to 0.31), about the same as reviews. Distance is not something a link or a review can change. If a whole side of town never sees you, what you can change is where your real address is, for example a real, staffed location there.

3. Primary category: the cheapest thing you control

97% of top-3 businesses had a primary Google category that matched the search, against 82% of controls (a gap of 15 points, interval 12 to 19). The gap was widest for law: 92% of top-3 law listings matched against 62% of controls, many of which were filed under a generic "Attorney" or "Law firm" category instead of "Personal injury attorney" or "Divorce lawyer."

This matches the industry's expert survey: Whitespark's Local Search Ranking Factors puts the primary category first for the local pack. Our data cannot prove that changing a category moves a ranking. But it takes minutes and costs nothing, so it is the first thing to check.

4. Reviews: the strongest factor a business can work on

Of the factors a business can change, review count had the strongest association with rank. The within-search score was 0.28 (interval 0.24 to 0.32) across 337 searches, and it was positive in 281 of them. The top 3's median review count beat the controls' in 258 of 337 searches (77%). Top-3 businesses had a median of 87 reviews; controls had 25.

The relationship held in every niche group, with real differences in size. It was weakest for real estate agents. In our first five metros it did nothing for dentists (a score of 0.00 over 15 searches), but in the next ten metros it was clear for dentists too (0.34 over 47 searches).

A single average can hide a lot of spread, so here is the score search by search. Reviews and closeness lean clearly positive; referring domains cluster around zero.

One data note: Google's data leaves the rating out entirely for a listing with no reviews (453 of 6,708 listings in dataset A). We counted those as zero reviews. Dropping them would hide exactly the weakest competitors.

Practitioners widely report that recent reviews matter more than the lifetime total. A one-time snapshot cannot test that, so treat the monthly pace of the businesses in your top 3 as the bar to beat.

5. Photos and the page your listing links to

Top-3 businesses carried a median of 24 photos against 9 for controls (within-search score 0.21, interval 0.18 to 0.25).

The page the Google listing links to showed a clear pattern too. We read the saved landing pages of 2,389 top-3 and control listings in dataset A:

• 65% of top-3 businesses named the searched service (or a close synonym, such as "roofer" for "roofing contractor") in the page's title tag, against 49% of controls.
• 53% named it in the page's main heading, against 36%.
• The top 3 mentioned it a median of 11 times in the body text; controls, 4.

That association remained after we accounted for reviews, raw referring domains and distance within each search (a score of 0.22, interval 0.14 to 0.30, over 178 searches). It is still a correlation: businesses that invest in their listing also tend to write better pages. But it is free to fix.

6. Each niche has its own shape

The same four factors, niche by niche. Six niches come from dataset A (15 metros); HVAC contractors and jewelry stores come from dataset B (metros to rural towns). Roofers and plumbers appear once, from dataset A, so no search is counted twice.

A few things stand out. Real estate agents and dentists lean hardest on distance. Lawyers have the highest referring-domain scores (0.22 to 0.24); HVAC contractors (0.18) and plumbers (0.14) show smaller ones; for roofers, dentists, real estate agents and jewelry stores the score is about zero. Jewelry stores are the odd one out overall: Google fills their results with pawn shops, gift shops and big-box stores, so reviews and photos track rank only weakly as pulled.

What the industry sells vs what we measured

A lot of local SEO is sold as links. Here is what this data says about them.

Links, outside law. In the 20 non-law downtown searches in our first five metros, the top-3 business had more clean referring domains in 9 and the control in 11 (45%, interval 26% to 66%). Median clean referring domains: 23 for the top 3, 22 for controls. Across 239 non-law searches in dataset A, the within-search score for raw referring domains was 0.05 (interval 0.01 to 0.09): above zero, but small. In dataset B's link sample (240 top-3 businesses and 240 controls), the top 3 had a median of 25 clean referring domains against 15. As a ratio that is 1.22 times more, with an interval of 0.90 to 1.64. Among businesses with similar review counts in the same town size and niche, that gap was no larger than chance would produce (p = 0.98).

What this does and does not show: outside law, the samples rule out a large link gap between the top 3 and the low-ranked businesses (in dataset B, a gap above about 1.6 times). They cannot rule out a small one (HVAC contractors are the clearest case, at 0.18 on raw referring domains), and they cannot say what one new link would do for one business.

That fits the outside evidence. Whitespark's survey ranks the quality of links to the landing page #25 for the local pack, and links to the whole domain #31 for the pack but #3 for regular organic results. The one published before-and-after case we found, from Sterling Sky in 2024, saw a strong link move a site into the organic top 3 while its local pack position moved by one spot. That is one site with no control group.

Maps and organic are different contests. (Organic results are the regular blue links under the map.) Across the six study keywords searched from downtown, only 10 of 246 Maps top-3 businesses with a website (4.1%, interval 2.2% to 7.3%) also ranked in the organic top 10 for the same search. Links are a well-documented lever in organic results. For Maps, outside law, we found little relationship.

None of this says links are worthless. They help organic rankings, and they track Maps rank in law (next section). It says that for Maps outside law, the evidence does not support making links the first thing a local business pays for.

The law exception

Personal injury and divorce lawyers were the one group where links clearly went with winning.

• In the first five metros, the top-3 law firm had more clean referring domains in 7 of 10 downtown searches (median 176.5 against 98).
• In the ten-metro replication, it had more in 17 of 20 (85%, interval 64% to 95%; a sign test gives p = 0.003), with a median of 212.5 against 73.5.
• Across 97 law searches, the within-search score for referring domains was 0.23 (interval 0.18 to 0.27), against 0.05 for every other niche in dataset A.

Law also had the strongest review effect, tied with plumbers and roofers (0.34), and the widest category gap. If you market a law firm in a large metro, the data supports building links alongside reviews and category. Even here this is a correlation: firms with money for links also tend to have more reviews, better pages and better locations.

Small towns follow the same pattern, with a lower bar

Dataset B asked whether any of this changes outside big cities. Mostly it does not. Reviews tracked rank about equally in big metros and outside them (within-search score 0.30 against 0.28, intervals 0.25 to 0.35 and 0.25 to 0.32). Distance mattered a little more outside metros (0.29 against 0.24).

What changes is the bar. Take the weakest of the three top-3 businesses at each search. Its median review count was:

The intervals resample whole towns, so they show how much these medians could move with a different draw of places. The rural figure is clearly below the big-metro and mid-size figures; its interval still touches the small-town one, and the other three overlap each other. In rural towns, 26% of top-3 listings had fewer than 10 reviews (84 of 318), and in 58 of 106 rural searches (55%) at least one top-3 listing had fewer than 12. A rural business faces a much lower review bar than one in a city, though a low count alone does not put anyone in the top 3: distance and category still apply.

What to do first

In this order, cheapest and best-supported first:

1. Measure from a grid, not one spot. Check your keywords from a fixed set of points around your service area, at the same time each month.
2. Set the primary category to exactly what you want to rank for. Copy what the top 3 at your own search points use.
3. Ask every customer for a review, every month. Match or beat the monthly pace of the businesses in your top 3. No incentives and no gating; Google's policies forbid both.
4. Name the service on the page your listing links to. Put it in the title tag, the main heading and the body copy, written for people.
5. Add real photos until you are at least at the level of the businesses beating you.
6. Use a real address in the area you want to win. If a whole side of town never sees you, the answer is a real, staffed location there, not more links.
7. Then links: first for organic rankings, and from the start if you are a law firm in a competitive metro.

Methodology appendix

Dates. Dataset A's first five metros were pulled on October 3, 2026, 20:55-21:38 UTC; the ten-metro replication on October 4, 00:10-00:29 UTC. Dataset B's new searches were pulled on October 5, 20:05-20:17 UTC; its big-metro roofing and plumbing searches reuse dataset A's October 3-4 pulls. Each search was pulled once, on desktop, in English, asking for the top 20 Maps results.

Pins and zoom. First five metros: downtown plus two named neighborhoods or suburbs 3 to 17 miles out (Chicago: Lincoln Park and Oak Park; New York: Brooklyn and the Bronx; Los Angeles: Hollywood and Santa Monica; Houston: Spring Branch and Pasadena; Phoenix: Scottsdale and Chandler), at map zoom 14. Next ten metros: downtown plus pins 7 miles north, south, east and west, or 4 miles where 7 miles out is open water (Miami east, Seattle west, Tampa west, San Diego west), at zoom 14. Dataset B: big metros use the same pins; mid-size cities (100,000-500,000 people) have a center pin and two pins 3 miles out at zoom 13; small towns (10,000-50,000 people, no place of 100,000+ within 20 miles) have three pins at zoom 12; rural towns (1,000-9,999 people, no place of 50,000+ within 25 miles) have one center pin at zoom 11. Wider zoom in smaller places was a design choice, so "town size" here means population plus map area, not population alone. Of the 30 rural towns drawn, 28 produced at least one usable search.

Niches and keywords. Dataset A: "personal injury lawyer," "divorce lawyer," "plumber," "roofing contractor," "dentist," "real estate agent." Dataset B: "roofing contractor," "plumber," "HVAC contractor," "jewelry store." Two cash-home-buyer keywords were also pulled in dataset A's metros and are left out of every figure here (they rarely return 15 listings).

Exclusions. Paid Maps ads removed (12 in dataset A's six niches, 67 in dataset B). Dataset A: 390 planned searches in the six niches; 7 returned no results; 46 had fewer than 15 listings and so no controls; 337 usable. Dataset B: 620 planned searches, all returned; 47 had fewer than 11 listings; 573 usable. Landing pages: a page counts only if it was fetched as readable HTML (not blocked) and only one business links to that exact address; 2,389 page-and-group pairs qualified.

Clean-link filter. "Clean referring domains" counts one link per linking website, live links only, with a DataForSEO backlink spam score under 25, and none from 17 cheap top-level domains (.shop, .store, .site, .online, .space, .website, .xyz, .top, .icu, .click, .pw, .live, .cyou, .sbs, .lol, .fun, .tech) used by a spam network that links to local business sites whether they rank or not. In dataset A, clean counts cover the downtown pin only: every top-3 business and control in the first five metros, and in the next ten metros the top 3 plus three controls in each law search (budget limits; one plumber search too). In dataset B they cover a seeded random sample of 15 top-3 businesses and 15 controls per niche and town size (480 businesses; directories, social sites and websites shared by two or more businesses skipped). Raw referring-domain counts cover every listing with a website.

Category match. A primary category counts as matching when it names the searched service: "personal injury" for personal injury lawyers, "divorce" or "family law" for divorce lawyers, "plumber," "roof," "dentist" or "dental," "real estate," "HVAC," "heating," "air conditioning," "furnace" or "cooling," and "jewel."

Missing data. A listing without a rating object is counted as 0 reviews (453 of 6,708 listings in dataset A, 802 of 11,173 in dataset B). Distance is missing for 3 listings in each dataset (no coordinates). 892 of dataset A's listings have no website, so they have no link or landing-page data. The within-search score needs at least 8 listings with a value in that search.

Statistics. Within-search score: Spearman rank correlation between better Maps rank and the factor, across all listings in one search, averaged over searches. Intervals: 2,000-draw cluster bootstrap over keyword-and-metro groups (dataset A) or keyword-and-town groups (dataset B), seeded so every run gives the same numbers. Counts of the form "k of n" use a Wilson 95% interval. The landing-page result also removes, within each search, what reviews, raw referring domains and distance already explain (a partial rank correlation). The dataset B link test shuffles top-3 and control labels within groups of similar town size, niche and review count, 5,000 times.

Replication outcomes. Before pulling the ten-metro data, we wrote down what each first-round finding would need to count as replicated. Re-scored from this study's own numbers:

The one failure is informative: the first round said reviews did not separate dentists or real estate agents, and the replication found that they clearly do for dentists (0.34) and weakly for agents (0.12, right at the pre-set cut-off).

What we could not measure. Clicks and direction requests (Google does not share them), review recency over time, how long a business has been listed, and anything about AI answers. Link data comes from one commercial index (DataForSEO). Two side checks from our earlier write-up (directory listings found for each business's name, and whether top-3 businesses ran ads) used a separate pipeline we did not re-run for this version, so they are not reported here.

Independent check. A separate reviewer, who did not write this analysis, recomputed the numbers in this post from the raw data with their own code before publication.

Get the data

Free to use with a link back to this page:

• Every listing, de-identified (CSV, 15,597 rows) and its data dictionary. Business names, websites, Google place IDs, phone numbers, addresses and coordinates are removed. Smaller towns are replaced by a code (such as small-03) assigned in a random order that cannot be worked back from town names. We left out their Census region too, because no region, town size and niche combination has 10 or more searches. The searches themselves are public, so anyone who re-runs the same search could see the businesses ranking then. We remove direct identifiers; we do not claim more than that.
• Chart data: factor strength, by niche, clean links, headline comparisons, reviews by niche, review bar by town size, score distributions, review-count distribution, stability, replication, search counts. Summary cells built on fewer than 10 searches are left blank.

Citing this study? Please link to https://sceptermarketing.com/what-ranks-a-local-business-in-google-maps/ and give the "as of November 2026" date. Questions about the method, or want the numbers for a niche we did not cover? Tell us.