Local prominence in local SEO: how Google ranks local businesses beyond distance

Location prominence is a score that ranks local businesses by signals of how well known they are on the web (mentions, reviews, links, directory listings), instead of, or in addition to, their distance from a point in the searched area. Patent US 8,046,371 B2, "Scoring local search results based on location prominence," describes it. Google filed the original application on May 27, 2005. This continuation was filed on September 29, 2010 and granted on October 25, 2011.

What does patent US 8,046,371 describe?

Patent US 8,046,371 describes a local search system that scores documents inside the broad area of the search with location prominence, and documents outside it (but still inside the search circle) with distance. Its 3 inventors are Brian O'Clair, Daniel Egnor and Lawrence E. Greenfield. The patent holds 26 claims and 12 drawing sheets, and it continues application 11/138,670, filed on May 27, 2005 and granted as US 7,822,751.

The patent defines location prominence as a score "based on one or more factors unrelated to the geographical area with which the document is associated, the searches performed by users, and/or the search queries provided by the users." In local search, "a common document is a business listing or a web page."

Which problem does location prominence solve?

Location prominence solves the weakness of ranking local results by distance alone. Older local search engines picked a point in the area, such as "city hall, downtown, or a geographic center," then ranked businesses by their distance from it, or alphabetically. "If the user does not live near the identified location," the results "are not meaningful to the user."

The patent's example is a search for pizza in Washington, D.C. Ranked by distance, the result closest to the chosen point scores highest. Ranked by location prominence, the factors are "intended to convey the 'best' documents for the geographical area rather than documents based solely on their distance from a particular location."

How does Google define the searched area?

Google defines the searched area in 2 layers:

1. The broad area: "any geographic location that is specified as an incomplete postal address," such as a city, a zip code, a street, a county, a metropolitan area, a large area such as the Lake Tahoe area, or "a combination of areas (e.g., Sunnyvale and Mountain View)." It comes from the query, or from "the entire visible map area within the map window" when the user searches on a map.
2. The search area: a circle "with a predetermined radius (e.g., 30 miles, 45 miles, 90 miles, etc.)" drawn around a location inside the broad area, such as city hall, a downtown area or a geographic center.

A broad area can be a set of zip codes. The patent's example for Mountain View is the range 94039 to 94043, stored as "94039:5," and a zip code that is not allocated to any other broad area, such as 94044, can join the range of an adjacent area.

How are businesses inside and outside the area scored?

Businesses are scored with 2 different functions, after a relevance filter. First, the system keeps only the documents of the search area that contain the query terms "in their title, content, and/or category string" (as a phrase, all the terms, some of them, or synonyms). Then a business inside the broad area receives a location prominence score. A business inside the search circle but outside the broad area receives a distance score, which "may have a value between zero and one depending on the distance from the location."

In the claims, the location prominence factor "does not include a measure of distance or a measure of relevancy of the first document to the search query," and the outside document is scored by distance "and not using the location prominence factor."

The location prominence score can optionally be combined with the distance score, to present businesses "that are closer together." It can be combined with "a minimal value (e.g., one)," which "may ensure that documents that fall within the broad area are scored higher than documents that do not." The scores can be computed in advance and stored "as metadata associated with the documents."

Which factors build location prominence?

5 factors build location prominence in the main implementation:

1. The score of the authoritative document: the link-based score of the page that is authoritative for the business. "A document corresponding to the home page of a restaurant may be considered more authoritative for that restaurant than a document corresponding to a review of the restaurant."
2. The total number of referring documents: documents that refer to the business "by mentioning the name of the business, the address of the business, and/or a telephone number associated with the business."
3. The highest score of referring documents: the highest link-based score among the pages that refer to the business.
4. The number of documents with reviews: reviews in "newspapers, magazines, web pages, and blogs."
5. The number of information documents: pages that give "the address, telephone number, and/or hours of operation of the business," its reviews and atmosphere, or whether it "accepts credit cards," such as Dine.com, Citysearch and Zagat.com.

The factors "may be combined and/or weighted in any manner." For example, they "may be 'squashed' and linearly combined." Squashing is "a function that prevents one large signal from dominating the others," and some factors "may also be normalized, if necessary." The claims describe giving each factor a weight and combining the weighted factors. The patent gives no formula, no weight and no threshold.

Which other factors can count?

The patent lists 7 other possible factors:

1. Review ratings: "how many stars or thumbs up/down."
2. A function of all review scores, for example their average.
3. The type of document containing the review: "a restaurant blog, Zagat.com, Citysearch, or Michelin."
4. The language of reviews: words such as "noisy, friendly, dirty, best."
5. User logs: "what businesses users frequently click on to get detailed information and/or for what businesses they obtain driving directions."
6. Financial data: "the annual revenue associated with the business and/or how many employees the business has."
7. Age: "the number of years the business has been around or how long the business has been in the various listings."

The patent adds that the factors could "train a model using machine learning techniques," for example to determine "the probability that a user might select a particular document in the search results."

Can the user turn location prominence off or choose its factors?

Yes, in the patent's interface. A link labeled "Turn Off Location Prominence" switches the feature off: then "all of the documents in the relevant set of documents are scored based on distance." A more detailed interface can let users "set his own function of various factors or to exclude certain factors," for example not using "information from Zagat.com." Claims 12, 13, 19, 20, 25 and 26 cover these user-selected factors and the off option.

How does this patent relate to Google's local ranking guidance?

Google's Business Profile help page on local ranking names 3 main factors: relevance, distance and prominence. It defines prominence as "how well-known a business is," based on information such as how many websites link to the business and how many reviews it has. The patent, filed in 2005, shows how such a prominence score can be built and combined with distance.

The local experts patent adds a layer on reviews: it identifies reviewers with expertise in a category and an area, whose reviews weigh more.

What does location prominence change for your local SEO?

Location prominence means that a business can rank across its whole area when the web talks about it, not only when it sits close to the center of the search. 7 consequences follow:

1. Make your website the authoritative page for your business. The link-based score of the business's authoritative page is the first listed factor: earn links to it.
2. Get mentioned with consistent name, address and phone number. Referring documents are identified by name, address or phone number, and their total counts.
3. Earn a mention from a strong page. The highest score among referring documents is a factor: one article on a strong local news site can matter more than many weak mentions.
4. Collect reviews on several types of sites. The number of documents with reviews counts, and the type of site containing them can count as well.
5. Be listed in the information directories of your sector. Pages that give your address, hours and payment options feed the information documents factor.
6. Keep your listings stable over time. The years in business and how long the business has been in the listings are possible factors.
7. Do not rely on one signal. Squashing limits the weight of any single factor: balance links, mentions, reviews and listings.

In local SEO, we audit these 7 factors together, from links to the authoritative page to the stability of listings, since squashing caps the weight of any single one.

The patent describes what Google's system can do. It does not confirm which prominence factors Google uses today, or their weight.