Inside Google Maps: 72 ranking signals and the architecture behind local search

When a business appears successful Google Maps, the listing you spot is the extremity merchandise of a overmuch larger system.

Behind it sits a canonical geographic entity assembled from aggregate information sources, connected to the Knowledge Graph and the web, scored by respective ranking systems, filtered done geographic and semantic retrieval, personalized for the user, and yet passed to a rendering motor that decides what tin really look connected the map.

Our squad precocious obtained a binary exposing a non-public scope of Geostore, the strategy Google uses to correspond geographic entities. We crossed it pinch Maps protocols, web traffic, the web index, mobile services, style tables, on-device components, and Google’s 2024 leak.

The recovered worldly includes:

• 72 Geostore ranking signals.
• 793 information root providers.
• 446 section hunt intent types.
• 50,998 Mapcore styles.
• 12,936 explanation styles.
• 10,936 searchable Geostore declarations.

The ranking signals will astir apt pull attention. But they’re only 1 layer.

The architecture astir them tells a much important communicative astir really Google understands places and what local SEO whitethorn go arsenic Maps turns into a conversational product.

The first point to understand: A listing isn’t the entity

A useful intelligence exemplary starts pinch Geostore. Google represents geographic objects internally arsenic Features. A Feature tin beryllium a business, building, road, city, station, area, transit element, aliases moreover a 3D object.

For an establishment, the entity tin incorporate identity, geometry, root information, websites, business-chain relationships, Knowledge Graph references, concepts, and ranking information.

The acquainted Maps listing is assembled later. What a business proprietor edits successful Google Business Profile isn’t needfully what Google maintains internally arsenic the entity.

Google builds a canonical practice of the spot that tin incorporated information from aggregate sources, past changes successful geometry, and link to different Google identifiers, including the Knowledge Graph instrumentality ID (MID).

For section SEO, the entity is the much useful portion to consider. The listing is the interface. The entity sits underneath it.

Google combines information from 793 providers

One of the astir revealing parts of Geostore is its provenance system. 

A business doesn’t simply person 1 source. Its sanction mightiness travel from 1 provider, its telephone number from another, its class from another, and its geometry from location other entirely.

The corpus exposes 793 root providers, on pinch mechanisms for provenance, priority, trust, and conflation.

Conflation is the process utilized erstwhile respective sources picture the aforesaid entity and disagree.

Geostore contains generic mechanisms that tin prime 1 value, merge respective values, aliases harvester them. It besides models spot levels ranging from blocked aliases untrusted sources to trusted and super-trusted ones.

This gives a different mentation to a communal section SEO problem: changing a section successful Google Business Profile doesn’t guarantee that Google’s canonical practice instantly becomes that caller value.

The edit becomes different portion of grounds entering a strategy that whitethorn already person competing evidence.

For businesses struggling pinch persistent incorrect attributes, copy information, aliases changes that many times revert, this architecture helps explicate why the problem tin beryllium harder than editing a listing.

Dig deeper: The section SEO gatekeeper: How Google defines your entity

The archive makes the leak overmuch much useful

The binary itself gives america structures, section numbers, and complete enumerations. Google’s March 2024 archiving often gives america thing different: prose explaining what those structures mean.

We published the 2 together. The resulting archive contains 10,936 Geostore declarations that tin beryllium searched by connection name, package, section type, archiving text, tag number, status, and different properties.

In reverse-engineering work, an soul sanction tin easy go a mentation erstwhile it circulates done the SEO community. The archive lets you inspect the underlying evidence.

If a awesome exists, you tin find its declaration. If a section existed successful the 2024 documentation, you tin publication the associated description. If a section has been stripped from the newer customer scope, its protobuf tag still leaves a numbered hole.

The 2024 leak gave america galore descriptions of Google’s systems. The newer binary gives america overmuch much of their existent vocabulary.

Together, they supply a much useful image than either root does connected its own.

Oyster Rank contains 72 ranking signals

Geostore has its ain ranking system. Internally, it’s called Oyster Rank. We recovered a complete visible enumeration of 72 signals, including:

• Google reviews.
• Web query volume.
• Listing impressions.
• Listing opens.
• Direction requests.
• Website clicks.
• Chain membership.
• Wikipedia signals.
• Popularity.
• Prominence.
• Landmark information.
• Road usage.

Out of 72 values, 25 are explicitly marked deprecated.

The important limitation is that we recovered the awesome names, not their existent weights.

The schema shows a pipeline successful which earthy observations are extracted, normalized, and mixed into the Feature’s rank. But the coefficients that would show america really overmuch each awesome contributes are extracurricular the scope we recovered.

So SIGNAL_GOOGLE_REVIEWS proves that reviews beryllium to the Oyster Rank vocabulary. It doesn’t beryllium that reviews presently transportation a peculiar weight successful a Maps search.

The 72 signals aren’t the Google Maps algorithm

This is astir apt the astir important explanation for SEOs. Oyster Rank appears to qualify the value of the entity wrong Geostore. A personification query still has to spell done further systems.

Maps must understand what the personification means, place a geographic context, make candidates, measure semantic relevance, and service a last consequence set.

A simplified pipeline looks much like:

Geostore entity -> query knowing -> semantic matching -> campaigner procreation -> surface science and value -> reranking -> results

There are further complications. We besides recovered a abstracted scorer moving wholly offline connected the device. It has 8 signals crossed 13 tiers and is chopped from some Oyster Rank and server-side Places ranking.

There isn’t a azygous Maps ranking formula. Different scoring and retrieval systems run astatine different stages.

Turning the 72 Oyster Rank signals into a checklist of 72 Google Maps ranking factors would miss astir of the architecture.

Local hunt doesn’t person a fixed radius

We besides tested the geographic furniture directly. A communal section SEO exemplary imagines Google looking wrong a predefined radius astir the personification and ranking the businesses recovered wrong it.

Our measurements show thing much dynamic. Using the aforesaid root successful Paris, the geographic footprint changed considerably depending connected the query.

A dense query, specified arsenic pharmacie produced a acold smaller hunt area than a marque query, specified arsenic Carrefour.

The situation matters, too. The aforesaid drugstore query expanded dramatically erstwhile tally successful a sparsely populated agrarian area.

Google appears to accommodate the campaigner abstraction to some the query and what exists astir the user.

We past removed geographic weighting from the aforesaid engine. Across 5,083 calls and 86,584 results, the median region moved from 6.87 km pinch surface science to much than 4,000 km without it.

More interestingly, the non-geographic bid remained highly stable.

This suggests surface science is doing much than reordering the aforesaid database of candidates by distance. It changes what the retrieval strategy considers successful the first place.

Distance is still basal successful section SEO. But “I’m closer, I should rank higher” is an incomplete model.

Dig deeper: The proximity paradox: Beating section SEO’s region bias

Maps and the web are connected done entities

The relationship betwixt Maps and classical web SEO whitethorn beryllium 1 of the astir consequential findings successful the corpus.

Geostore Features tin link to the Knowledge Graph done a MID. On the web scale side, documents tin besides transportation MIDs.

Google has a furniture called webref that associates documents pinch entities and stores information, including topicality, confidence, geographic metadata, and document-level scores.

The narration besides useful astatine the document-ranking level. The recovered structures picture a comparative ranking awesome betwixt different documents for the aforesaid entity, on pinch properties specified arsenic whether a page is an writer page, patient page, aliases reference page.

This creates a very different measurement of reasoning astir a shop locator aliases location page. Its domiciled whitethorn widen beyond ranking for queries specified arsenic “shoe shop Paris.” The archive tin go grounds astir the underlying entity.

The SEO nonsubjective is past partially to make it easy for Google to establish:

• Which entity the archive describes.
• How overmuch of the archive is really astir that entity.
• How assured that relation should be.
• Whether the archive is simply a useful reference for it.

Web SEO and section SEO are overmuch little abstracted wrong Google’s infrastructure than their interfaces suggest.

Google understands concepts, not conscionable categories

The semantic furniture goes considerably beyond the superior class visible connected a listing. Google uses GConcepts, a shared conceptual vocabulary that tin picture businesses, dishes, attributes, cuisines, work modes, and different concepts.

We followed a elemental ramen query done respective parts of the system. The hunt results themselves didn’t each beryllium to 1 category. Google connected the query pinch ramen restaurants, Japanese restaurants, Asian restaurants, and different related concepts.

Inside listings, the semantic practice goes deeper. Review topics, paper dishes, and different attributes tin beryllium represented arsenic entities alternatively than plain strings.

For an AI system, this is highly useful. Instead of rereading thousands of reviews each clip personification asks whether a edifice has agelong waits aliases bully ramen, Google tin activity from system themes, entities, and precomputed signals already attached to the place.

Semantic knowing becomes overmuch much important erstwhile the interface starts answering analyzable questions.

Part of Google’s geographic intelligence lives connected the phone

Not everything is calculated connected Google’s servers. We recovered on-device structures associated pinch visits, spot candidates, predominant places, trips, location and work, mobility patterns, and personification location profiles.

One peculiarly absorbing entity is ChainAffinity, which suggests the strategy tin exemplary affinity toward a recurring unit chain.

There’s besides the abstracted offline scorer mentioned earlier.

The nonstop grounds level differs betwixt components. Some structures are explicitly named successful the recovered schema, while parts of the persona furniture tin only beryllium reconstructed from compiled structures.

But the broader architecture is clear: The telephone itself participates successful building geographic context.

That intends personalization successful Maps tin harvester server-side knowledge of the world pinch a section exemplary of the user’s ain geography.

Ranking still doesn’t guarantee representation visibility

Search results are only 1 of the outputs of Maps. 

The ocular representation has different problem to solve: Thousands of perchance applicable entities cannot each person labels simultaneously. That occupation belongs partially to Mapcore.

We recovered 50,998 Mapcore styles and 12,936 explanation styles. Label visibility tin alteration pinch zoom and different rendering conditions.

A business tin beryllium eligible aliases highly classed and still neglect to look arsenic a visible sanction connected the map. Search ranking and representation visibility are abstracted optimization problems.

This favoritism becomes particularly important erstwhile group measurement “Maps visibility” utilizing screenshots aliases representation grids. The ocular aboveground includes a rendering determination aft retrieval and ranking person already happened.

Then Google puts Gemini connected top

The timing of the betterment is useful. Google is quickly expanding Ask Maps and different AI-powered experiences, but overmuch of the infrastructure required to reply analyzable questions was already present.

The strategy already has:

• Canonical spot entities.
• Semantic concepts and attributes.
• Reviews and extracted topics.
• Knowledge Graph relationships.
• Web evidence.
• Geographic retrieval.
• Behavioral signals.
• Personal geographic context.
• Listing composition.
• Ranking systems.

Gemini adds a conversational interface complete these layers. That changes what a section query tin be.

“Best ramen adjacent me” is comparatively easy. “Where tin six group eat adjacent my edifice tonight, pinch 1 vegetarian, small waiting time, and bully caller feedback astir service?” requires a different benignant of spot representation.

Google needs to cognize what the edifice is, what it serves, erstwhile it’s open, what group opportunity astir it, wherever it is, really it relates to the user’s way aliases context, and whether the disposable grounds is reliable capable to urge it.

Maps has been building galore of those ingredients for years. The AI furniture gives Google a caller measurement to usage them.

Dig deeper: Google Ask Maps: How to optimize for visibility

What to optimize beyond the Business Profile

The astir actionable conclusion from this investigation isn’t a caller database of ranking factors.

Local SEO has traditionally concentrated heavy connected optimizing the Google Business Profile: categories, reviews, photos, attributes, opening hours, and different listing fields.

Those stay important.

But Google’s architecture suggests a broader objective: amended the practice Google tin build of the entity itself.

For a business aliases unit brand, I would progressively ask:

• What precisely is this place?
• What does it offer?
• Which marque aliases concatenation does it beryllium to?
• Which concepts and attributes picture it?
• Does its website picture the aforesaid entity clearly?
• Which Web documents supply grounds astir it?
• Does Google spot existent request for the brand?
• What do reviews consistently opportunity astir circumstantial aspects of the experience?
• Which audiences and contexts could make this spot relevant?
• When should Google urge it alternatively than different candidate?

The value of the reply Google tin nutrient depends connected the completeness of that representation.

Semantic completeness whitethorn go the adjacent section SEO battleground

Proximity, relevance, and prominence stay useful concepts.

AI adds different requirement. The strategy needs capable system grounds to logic astir a place.

A edifice tin person an optimized floor plan and hundreds of reviews, yet still beryllium poorly represented for a circumstantial mobility if Google can’t confidently link the applicable attributes, concepts, web pages, and reappraisal themes to the entity.

A ample unit marque has an further problem.

Google models some chains and individual locations. We observed stores from the aforesaid marque successful the aforesaid metropolitan area carrying different superior concepts, moreover though the concatenation exemplary contains a canonical conception structure.

Consistency can’t beryllium assumed simply because each location belongs to the aforesaid brand.

For multi-location SEO, the task extends crossed the marque entity, each section entity, the website, system data, third-party sources, user-generated content, and the relationships betwixt each of them.

That’s a overmuch larger aboveground than a Business Profile.

Dig deeper: Multi-location SEO: How to building geographic pages astatine scale

The listing is only the surface

The 72 Oyster Rank signals are fascinating because they expose categories of accusation Google tin usage erstwhile estimating the value of a place.

The deeper uncovering is the strategy astir them:

• Geostore builds a canonical geographic entity from competing sources.
• The Knowledge Graph gives that entity semantic meaning.
• Webref connects Web documents to it.
• Search and Places construe the query and retrieve candidates wrong a move geographic space.
• On-device systems lend individual geographic context.
• Mapcore controls what reaches the ocular map.
• Ask Maps and Gemini tin yet logic complete the resulting practice successful earthy language.

The Google Business Profile still matters. But Google’s architecture suggests looking beyond the listing itself and considering the practice Google has built of the business.

The cardinal mobility is whether that practice is complete capable for Google to confidently urge the business.

The full study includes the reconstructed architecture, experiments, and method evidence. The companion archive exposes the 10,936 recovered Geostore declarations alongside the 2024 archiving truthful the underlying schemas, enums, fields, and signals tin beryllium inspected directly.