When I search for Pizza in New York in Google Local, I’m told that there are about 79,400 results. The top result I see is Lombardi’s Pizza, which is “0.8 mi NE” of the green arrow on the map that points to New York.
Exactly what is that green arrow pointing to, and why does Lombardi’s Pizza show up number one?
To rephrase that question, how does a site become the “authority” for a region, for a business type, rather than authority for a specific location and business name?
In Authority Documents for Google’s Local Search, I wrote about a patent application that described how a specific site would show up first when searching for a specific business name and location. The author of that patent filing, Daniel Egnor, was the named inventor on several other patents on local search that came out the same week. This post is about one of those other ones, which is the only one I can recall from Google that talks about identifying specific regions and tying them to queries about business types:
Indexing documents according to geographical relevance
US Patent Application 20060149774
A local search engine may generate results for a local search query that are limited to desired geographic regions. The geographic region may be defined, for example, by a certain distance (e.g., 20 miles) from a specified point or region. The search results are efficiently generated by indexing geographically relevant documents based on the contents of the documents and also based on multiple location identifiers. In one implementation, the location identifiers define regularly spaced geographic areas and the documents are indexed such that the multiple location identifiers indexed for each document are selected to define a predetermined range around the region with which the document is associated. This document indexing technique allows for efficient searching by geographical region.
The patent application states that one way they could identify regions would be to use something called the Hierarchical Triangular Mesh, which partitions a sphere (or a roughly spherical object like the Earth) into “spherical triangles.”
Some other regional grid systems (such as the United Kingdom’s National Grid) are also mentioned.
Generic Queries and Regions
The section of the patent application that discusses searching for generic queries (such as “coffee shop”) within a region is titled “Operation of Search Engine.” It describes several different alternatives to handle such searches. After performing many searches in several different regions, I think that something like what is described in this patent application is being used, and more than one of the following alternatives may be in effect.
One alternative version of the process starts with a smaller radius around a center point within a region, finds relevant results for the query, and serves those, then looks at results from a larger radius and finds and serves relevant results, and so on.
Another alternative looks at all of the results within the region, and bases results on relevance for the query term without regard to the distance to the center of that region.
As an alternate possible variation of the techniques shown in FIG. 9, instead of pre-indexing documents with many location identifiers, each document may be indexed with only the location identifier associated with the document. Search queries may then be formulated as performed in act 903, where the range is the whole search region. That is, the query may include a logical OR concatenation of all the location identifiers within the search range. This variation reduces index size and may allow for more flexibility in selecting the region size and shape at query time but increases query complexity.
The Importance of Relevance
There are a few different methods listed in that section. All of them rely to some degree on the relevance of the document to the term. All of them rely upon the relevant pages having business information tied to a location. Synonyms may be used in determining the relevance of a query to a page. The main difference seems to be how results are returned based upon location.
- Within a five mile (or other distance) radius of the centerpoint of a region, with relevant pages ranked higher.
- Within expanding radii from the CenterPoint of a region, with closer businesses being ranked higher, and relevant pages within each of those radii ranked higher.
- Within a region itself, regardless of a radius, with relevant pages ranked higher.
That’s a rough reading of the patent application, but I think that it shows some alternatives that may work well.
I would think that relevance for the query would be determined in the same manner that it might be for web search results – looking at such things as the use of the terms in different parts of pages, anchor text pointing to the pages, PageRank, etc.
Local search has grown in importance since Google started serving local results above organic results in regular web searches (see the Google search for Pizza New York). So having an understanding of how local search may work could be helpful even if you are only really interested in organic results for searches.
We’ve been discussing the Google Local Business Center at Cre8asite Forums and descriptions that can be added to local results through that system. It’s difficult to tell what influence those may have. You may have also noticed reviews of businesses in searches. Again, it’s not clear what role those may play in local search results, though a large number of reviews that point to a business site may help that business become the “authority” site for a local search result.
What are you seeing with local search results that you find interesting?
5 thoughts on “Pizza at the Center of New York: Relevancy, Google Local, and Local Results in Organic Searches”
I have found several relevance enhancers in my experience.
1)selecting appropriate categories when you update your listing obviously improves results
2)Title of Listing ie Lombardi’s Pizza does better than just Lombardi’s would in the title of the business record
3)Number of external web references to the location and business. These seem to be calculated similarly to but different than the link command .
I was just getting around to review your article and the patent by Egnor…and lo and behold you’ve addressed this again. Thanks Bill!!
With regard to my business and competitors in the region, I’ve noticed some that “outrank” their relative distance from a point of origon and some that “underrank” the distance issue.
I’m going to go read your entries in depth and try and “valuate” the references. Hope to get back to you with some findings.
On a totally different point want to find a simple way to rank highest on a version of Local? Take a look at MSN. I haven’t reviewed it in depth yet, but I believe they utilize one of the Phone directories for placment of information. Of interest it appears you can give your brick and mortar business more than one location!
A competitor (with only one location) has seemingly located its business in about 5 critical areas. No matter where the starting point is…the competitor almost always turns up first.
I’m going to look into that one this weekend also. May have to duplicate what the competitor did!!! LOL
How’s that for a simple algo!!!!
Establishing the center seems to be a function of establishing the authority and relating the center to that.
Look at Pizzeria in NYC for example
Thanks for the observations. I think that your first set are spot on from the experimenting that I’ve been doing also.
For establishing which location is an appropriate one for a broad query like “New York” which could be the City or the State or different parts of the State, there seems to be a slightly different process, which appears to use the logic that you suggest, but is described more fully in one of the other recent patent applications. I’ll try to write on that today.
The distance vs. number of references, and the phone listing references are pretty interesting. Microsoft has some papers out on some of the research they’ve conducted on geographic location information. It’s worth exploring more.
Thanks for the input. Hopefully the other patent applications give us some more answers that add to a better understanding of what is going on.
You know you think that you can get this figured out and then I run across this example:
Insurance Olean NY
Note that the client I am working with is Hedley Brook and are listed H (8th). Above them are listings that have no references, and are further from the center. For example the 3rd position is occupied by Bantelman which is .9 miles from the center, seems to have no more relevant title, no categories selected and no significant web references.
PS is some categories (restaurants) reviews from cityscape also seem to carry significant weighting
Comments are closed.