Your local rank is not a number.
It is a surface.

I.

The comforting fiction of “we rank #3”

Ask a business owner where they rank for their most important keyword and you will usually get a number. “We’re third.” That answer feels precise. It usually is not. Google says local results are mainly based on relevance, distance and prominence. Distance explicitly refers to how far each business is from the customer who is searching.^[1] If the searcher’s location changes, one of Google’s stated local-ranking considerations changes with it. A business therefore should not be treated as possessing one universal local rank for a keyword. It has observations tied to locations.

That distinction is the foundation of serious local-rank measurement: business → query → coordinate → observed result. Remove the coordinate and you have compressed away part of the phenomenon.

II.

Geography already has a name for the problem

Geographers routinely work with measurements whose values vary over space. Penn State’s GIS curriculum describes spatial autocorrelation as the degree to which attribute values are related over space, and emphasizes that global measures can hide the specific local areas contributing to a pattern.^[2] Columbia’s overview of geographically weighted regression describes spatial non-stationarity: relationships can vary by locality rather than remaining constant everywhere.^[4]

DecodeLocal is not claiming Google uses Moran’s I, LISA or geographically weighted regression to rank businesses. Those are different analytical systems. The value of the academic literature is methodological: it teaches us not to assume that a spatially varying observation is constant everywhere.

Figure 1 — one keyword, two models. The single-position model reports “rank #3” for the whole city. The spatial model keeps one observation per coordinate: #1 at coordinate A, #2 at B, #6 at C, absent at D. The scalar compresses the market; the spatial model preserves it.

III.

A local ranking is better understood as a field of observations

Suppose a dentist is #1 near its clinic, #3 several kilometres north, #8 farther away and absent across another part of the market. Which rank is correct? All of those observations may be correct for their respective search origins. The error appears when a reporting system discards the coordinates and keeps only a single number.

Penn State’s Local Indicators of Spatial Association material makes a directly useful measurement point: a global whole-map statistic can identify an overall pattern but still fail to tell the analyst where the meaningful local pattern occurs.^[3] Local analysis exists because the where can matter as much as the aggregate.

Here is what that looks like when it is measured instead of asserted — a real rank surface from a real market:

IV.

Why the centre pin can be accurate and still mislead

Imagine measuring rainfall over a city with one perfectly calibrated gauge. The gauge can be accurate at its location and still be inadequate as a description of the entire city. This is the difference between an inaccurate observation and inadequate sampling. A local rank taken from one coordinate can be perfectly correct at that coordinate while being an incomplete description of the broader market.

One gauge can be perfectly calibrated and still not describe the storm. Accuracy and adequacy are different failures.

V.

Grid resolution is a measurement decision

More points are not automatically better. A grid that is too sparse can hide commercially meaningful variation; a grid that is unnecessarily dense can increase cost without adding much information. Resolution should be chosen in relation to market size, density, competition, physical geography and the decision the measurement must support.

A defensible system discloses the geometry instead of pretending there is one universal grid size. Report point count, point spacing or construction method, geographic extent, attempted versus successful observations and the query conditions used. Resolution determines the scale of geographic variation the instrument can detect — which is exactly why it belongs in the disclosure, not in the marketing.

VI.

The visibility footprint is more useful than the vanity rank

Once enough observations exist, a richer entity appears: the visibility footprint. Instead of asking only where the business ranks, ask across what proportion of the measured market it ranks first, in the top three, in the top ten, or not at all. Preserve the raw point observations and then calculate summary metrics on top of them.

• First-position coverage = valid points where the business ranks #1 ÷ valid measured points.
• Top-three coverage = valid points where the business ranks 1–3 ÷ valid measured points.
• Visibility boundary = the geographic transition from strong visibility to weak or absent visibility.
• Competitor overlap = coordinates where two businesses both maintain meaningful presence.

This is the measurement CitySweep implements: it maps local rankings across the entire city, keeps every observation’s coordinate, and computes the footprint on top of the raw points rather than in place of them. You can inspect the measurement geometry behind the published example sweeps — market, keyword, run date and point coverage disclosed.

VII.

Relevance and prominence deform the distance pattern

If distance were the only consideration, the nearest eligible business would always win. Google explicitly says local results are based on a combination of relevance, distance and prominence, and a farther business can still be selected as a better match.^[1] Therefore the observed surface should not be assumed to form a neat circle around the business. Irregularities are information: they tell the analyst where distance alone is not sufficient to explain the observed results.

VIII.

Measure change with the same instrument

A baseline becomes useful only when it can be repeated. The cleanest comparison holds the business, query, coordinates, geographic extent and result definition constant while changing the measurement date. If the grid, keyword universe or business identity changes, disclose it. A before-and-after should report points improved, unchanged and declined, first-position coverage, top-three coverage and missing points — and it should never allow an improved average to hide deteriorating zones.

A single-point instrument has the same discipline at smaller scale: a 13-point rank check establishes a baseline around one business, and that baseline can expand into a citywide sweep when the decision needs the whole market.

IX.

What a defensible local-rank report should disclose

Business measured; query measured; date; geographic boundary; attempted points; successful observations; grid spacing or construction method; result depth; business matching method; missing observations; comparison method if applicable. This is the difference between a screenshot and an instrument.

X.

The practical doctrine

Google gives us the premise: distance is one of the main local-ranking considerations.^[1] Spatial analysis gives us the discipline: local variation can be lost inside a global summary.^[2] The conclusion is simple. One point can answer “what happened here?” A grid can begin answering “what happens across the market?” Those are different questions. Stop asking one pin to describe a surface.

For the operating context around this measurement — what Google ranks on and how the local market got fiercer — read the 2026 local SEO field manual.