Have you ever wondered how Google might be coming up with the geo-local elements of a SERP? Ok, well, I probably haven't, lol, but I came across a recent patent award that was actually quite interesting. Sure, I can hear YOU yawning already, but I actually do get a little piqued over these things. Anyway, the patent in question is;
Enhanced Search Results - Filed; November 23, 2009 -
Awarded; June 3, 2010
Inventors; Walton Lin; Dorward; Sean; (more here) ; Castro; Luis; (Piscataway, NJ)
Interestingly this is a re-release of a patent published in 2007 (covered here by Bill) and as he noted at the time;
“Two of the listed inventors are also named inventors on Google’s patent application for Site Links, which are the lists of additional pages from the same domain that sometimes show up under the first link in search results, in response to a query. ”
Which implies there are some query constructors at play here that work on other enhancements to the regular search results presentations. In many ways location information is also something we can consider to be 'quick links' to specific entity location data.
They start off by discussing local search tasks as a expanded navigational search where the user is looking for the contact information which in some cases may not be found via the initial link for a given query. The goal is to save the user from having to click 2-3 pages deep into the site to get what their after. It is reminiscent of Ori Allon's 'no click searching. But that's another story.
Moving towards the no-click SERP
Let's get an example; Staples business centre, Peterborough Ontario
If we were presented with just the website result, with site links in this case, we would be unsure if we should use the 'Store Locator' link or the 'Contact' link. In our scenario, we simply are looking for the phone number to contact them. This is the expanded navigational task at hand. Fortunately, we get not only the site listing, but also the universal element with the phone numbers of the two locations for this area;
This method can be used to serve location and phone number information in a text listing, map listing if available or (in my testing) both. In the case of there being multiple locations, they will bring up as many as are deemed relevant based on the query analysis. Which is what we've seen in our example above.
Much of what the system is seeking to achieve is through query analysis. By that I mean looking at the user query, how granular it is, to establish not only the source for the navigational query, but also which and how many results to bring back. This affects search quality as well as performance, (processing).
The data is stored in what they term a 'local search engine' which provides information relating to geographical elements of a given site. This can include;
- Government offices
Just to name a few. To learn more about this and navigational search, you can look more into what are generally known as 'named entities'. This is a person, place or thing (events for example). For those in the local SEO game, I am sure you're familiar with the power of non link citations. That is an important element for geo-local ( as well as branding).
Furthermore, the local search engine may be populated with data from thirds party sources such as online yellow pages or local directories, (and of course, Google maps business listings). This does also bear out in what many local search gurus have encouraged for some time now.
It is interesting to note that varied data, for the same entity, can be mined from multiple locations on the web into a cluster. For example, let's say the website owner hadn't put the actual location information on the site, but had put it into other online databases, then this data may be displayed in the search result for the named entity, although it doesn't appear on the site. Somewhat unlikely a scenario, but an interesting concept/approach none the less.
Other information that can be stored, beyond location and phone number, might be elements such as prices on rooms for a hotel, ratings for the hotel and so forth. These can also be served to the user when making a geo-graphic named entity search. What is in the cluster for the site depends on the query type and data available.
They also suggest that the mapping component could take the location information found and dynamically map to the cluster ID. This means that even entities that haven't add themselves to Google maps for example, can still have a map returned. The system would take the location information found elsewhere and create the listing.
And even look at going beyond citations on the web and using registrar data to add to the cluster;
“In still further alternatives, mapping component 410 may perform a "WHOIS" query with a domain name registry entity, such as Network Solutions. The WHOIS query will provide a business name, address and telephone number associated with a domain name. If the information provided in response to the WHOIS query matches the address and telephone number stored in local search data storage 230, mapping component may select that URL as being the correct URL to map to the cluster ID. ”
In essence, they are looking for supporting evidence to map to a given URL/site. If the data from a given entity is conflicting, certain citations may be disregarded. This can happen in the case of situations such as a company listing on a directory site. The directory site would be dropped at the URL source for a given entity listing.
And of course, there is an inferred personalization based on the users location. If they search for say, a more generalized query like 'Wendy's restaurant' in an attempt to serve relevant results, the search engine might take the users current location into account when constructing the results. This can be based from IP or cookie which looks at other local searches conducted by the user in the past.
Universal Search Matters
If there is one area that a lot of SEOs aren't really getting the most from, is the Universal Search strategy. This means working from the SERP on out and understanding the make up of the space you're working in. That is why it seemed a good idea to further look into some of the elements of what goes into the geo-local aspects of how search engines might deal with this kind of information.
This particular offering, does build upon what Matt (Cutts) was saying about proper treatment of multiple locations in a recent blog post. It would seem to support strongly identifying locations to ensure that the search engine can easily digest and deliver the right answer. The main thing is that I hope this motivates you to not only dig deeper into localized technology, but as always, to get more of an interest in the workings of search engines in general.
... cya in the SERPs
Some of Bill's stuff;
Get your Geek on!