hen you perform some searches, Google might include Maps results within the web search results for those queries, or it might include some local results that change when you change your location in Google. Those queries are ones that don’t include geographic information within them, yet Google somehow decides that there’s some geographic relevance to the terms being searched for.
Some query terms likely have no geographic relevance to them, such as a query like [linux], which pretty much has a meaning unrelated to any specific location. Other queries may evidence an intent to find a location near a searcher, such as [restaurant]. A patent granted to Google this past week describes an approach that Google may use to assign an implicit local relevance to a query term or phrase when that query doesn’t contain any explicit references to a location.
A friend asked a few months ago why Google might decide that a particular phrase might be seen to have a geographical relevance in his region, but not show localized or Google Maps search results in other locations. My answer was that Google likely had developed a statistical geographical model that would trigger localized results based upon a combination of a query used and the location of the person searching. I’ve written a few posts in the past about a Yahoo! paper on geographic intentions, as well as a Yahoo! patent covering similar territory.
The patent is:
Identification of implicitly local queries
Invented by Michelangelo Diligenti, Wenxin Li, Fabio Lopiano, and Trystan G. Upstill
Assigned to Google
US Patent 8,200,694
Granted June 12, 2012
Filed: November 8, 2010
Abstract
Methods, systems, and apparatus, including computer program products, for identifying implicitly local queries. A query having one or more terms is received. The query is associated with a user locale. A degree of implicit local relevance for the query is determined. One or more search results for the query are received.
Each received search result has a respective score and a respective result locale. The score of a respective search result is modified using the degree of implicit local relevance for the query, the user locale, and the respective result locale of the respective search result.
Web Sites Associated with Locations
Websites might be associated with specific locations through many means. One might be that a location may be contained within the content of the site. The site might be specifically about a location, or contain an address and other contact information. Or a site might contain pages that are specifically about business listings that include location information for businesses listed.
A site might also contain metadata of some sort that identifies a location for the site.
Other indications that might be used to associate a web page or site with specific locations can include a country code top-level domain being used, such as “http://www.example.co.uk,” or an IP address that might indicate the site being located in a specific area. The user traffic to a site, such as the click records, might also be used to associate one or more locations with its pages.
Some sites might be determined to be associated with more than one location, all locations, or none in particular. Sites that are associated with all locales or none, in particular, might be said to be “globally relevant.”
Queries Associated with Location
Queries can also be associated with user locations. They can be done so explicitly, like a search for [restaurant chicago], which explicitly specifies the locale of Chicago. Such a search might best be served by Websites that are somehow associated with the location “Chicago.”
A similar locale restriction might also be included for a query that doesn’t specify a locale, but where it might be appropriate, such as a search for [restaurant].
Here are some signals that a search engine might look at when trying to determine if there is an implicit location within a query:
Location of Search Engine – A search via searchengine.co.uk might indicate an intent to restrict a search to the UK.
Implicit Local Relevance for All Terms in Query – Whether one or more terms in a query have been determined in the past to be relevant to a specific locale.
Language – Some queries might have significance in some languages, but not others. For example, the word “tax” has meaning in English speaking countries, but not in Spanish speaking countries, so it wouldn’t carry an “implicit preference for local results” in a country where people predominantly speak Spanish
Country – The word “freedom” has a fairly generic meaning in most English speaking countries and doesn’t have an implied preference for local results. In Australian though, it’s the name of a prominent home furnishings business and would have local significance there.
Aggregated User Behavior – If a statistically significant number of searchers who use a particular query tend to select search results associated with their location, then it’s likely to have an implicit preference for local results.
Again, if a statistically significant number of people tend to combine a term from a query with an explicit specification of a location, then it also could be seen as being an Implicitly Local query.
Also, if a statistically significant number of people who search for the term or phrase tend to add an explicit location to it in a followup query (during the same query session), then once again it could be seen as a term having an implicit preference for local results.
If people from a particular country code top-level domain tend to search for that term at a statistically significant rate, that could be another indication of a desire for a local result.
Use of Term in Content – If the term frequently appears in content related to particular locales, it might also be determined to have some implicit preference for local results. For instance, “if a term frequently appears in content associated with the United Kingdom, the term may be determined to carry some implicit preference for local results, specifically results from local to the United Kingdom.”
Takeaways
This patent appears to focus primarily upon organic web results rather than including Google Maps results within those web results, but it could be one way of identifying whether or not to include maps results within a set of Web search results. It does appear to be the kind of analysis that Google might follow when they insert more localized results into a set of search results so that people interested in seeing local results for a query might see them mixed in with more “globally relevant” results.
In March, I wrote about a Google patent application that described How Google Data Centers may be Split between Regional and Global Data. In the patent filing I covered, we were told that Google might use a prediction algorithm that may decide which data center to send us to based upon whether the query pointed to a result that might be geographically relevant for a certain location. The algorithm may be related in several ways to the one described in this patent on implicit local queries.
Here we see that some queries may implicitly call for local results in one area, and not in others. I’ve seen that happen for some fairly small regions before, where it seemed likely that people searching for a particular query likely tended to choose a local result in response to that query a high degree of the time, and where similar searches in other regions weren’t being shown localized type results.
I decided that it might be a good idea to identify and link to some interesting posts about local search, and came up with the following list:
- Assigning Geographical Locations to Web Pages
- Was Google Maps a Proof of Concept for Google’s Knowledge Base Efforts?
- Location Prominence at Google in Ranking Businesses at a Location
- Location Sensitivity in Google Local Search
- Authority Pages for Businesses in Google’s Local Search
- 10 Most Important SEO Patents: Part 8 – Assigning Geographic Relevance to Web Pages
- How Google May Identify Implicitly Local Queries
Last Updated June 26, 2019.
And still, seems that google is continuously modifying the way it Ranks local keywords, as some keywords are moving form global to local and other form local to global.
Here is a nice example about how “Limo” ( which has implicit local relevance ) Ranks in different US cities:
http://goo.gl/UO32w
They don’t mention it in this patent, but I wonder also if a high interaction rate on localised results for one query could trigger localised results for a similar query. For example, high interaction on a search for [blinds] could trigger local results for [roller blinds].
Very relavent topic. Great help to pull back the curtain of the inner workings by analyzing the patents.
Thanks
Keep writing such good reviews.
Ben presents an interesting contention. I’m curious as to whether or not semantics plays a role in the triggering of local results. In some contexts, semantic inferences might be effective. In others, though, they may not. Much of semantics is reliant on the factors discussed.
Great review.
I’ve recently (just in the past few days) started looking into the use of place names as keywords that don’t necessarily indicate locality and how Google responds to that. For example, try a search on [new york cheesecake]. I mostly see recipes in the results, but if I switch my location to New York City, I get local results.
Great post! Respect to Google for the clever way they are using data mining to improve the user experience.
Bill: That is another patent with real application to local smb’s and local websites. I noticed the phenomena it describes over a year ago, by chance, and discussed it with you. Even back then your hypotheses suggested some of the elements that the recently released patent describes.
BINGO!! BOOM BOOM BOOM You hit the nail on the head!!!! Ha Ha. CONGRATS!!! I had a different main hypothesis and it’s not referenced in the patent. LOL We also discussed other possibilities…still not referenced in the patent.
Frankly it worked for the site for the Bartender School outside of DC and we saw it applied for different websites in different regions and different verticals.
Its a topic that people focusing on local seo should pay attention to.
With these segmentations of queries, how do you find that it affects rankings on SERPs for companies trying to target specific terms that is both geography targeted and broad terms?
It’s also interesting how the local results can change based on the conjugation of the words themselves. For example, check out “auto finance” vs “auto financing” in a de-personalized search, I’m seeing the Places results in different places on the organic results. I wonder if the localization algo you mention here is somehow being combined with a stemming algo to formulate not only which words trigger local results, but which forms of the same words trigger them. Interesting…
I always felt that if a keyword was competitive its less likely to pull a Map result but will pull a local web result. I see this often in local searches and general searches, less competitive terms pull maps more often but more competitive seem to only pull web results. So it seems some level of competition is being used as a factor.
It’s an interesting concept that Google is trying to localize searches as it can be very frustrating at times. Our business benefits from local searches, but on the whole it can be a pain sometimes.
This actually explains a lot on why results are so wildly different when searched from different locations. It would be great if we could check cases where Google is mistakenly treating a website as local, although it is just as useful internationally (i.e. local business website with a useful article).
Its good to see that google localizing the search. Users will get the exact & relevant results on searching using some local queries. Local searchers will get their nod through this… 🙂
There are positives and negatives to the recent changes in Google’s apparent leap towards localised searches. For example our business IP runs through a city 400 miles away so searching I am faced with results from that city instead unless I physically change my location. As an SEO’er it can cause a headache as well as clients can see completly different SERP results based on lcoation.
Although with google+ local on the scene we might start seeing greater integration with local.
Interesting stuff, although I do wonder: when does Google realize that cross-location searches also happen? So what if people from China want to find the company with freedom in the name? And what if people from Australia are NOT looking for the company with freedom in its name? And what if what Google assumes to be the location of a query is actually not specific enough?
All I’m saying is: a lot of assumptions are being made that could hurt instead of improve the user experience.
Good example: since the Google Venice update I now see search results from Amsterdam while I do NOT live there. I live an hour away from Amsterdam so I have to turn localization OFF to get more relevant results for my location. Like I said: assumptions are being made that could hurt the user experience.
Hi Sergiu
Thanks for the example. I think it’s clear to see that some of the determination of whether or not a specific query is going to be implicitly calling for local results may depend upon searchers in different location. Just like the word “Freedom” might be global in most parts of the world, but often associated with a particular story in Australia, local intent for some queries is prone to change from place to place.
Hi Ben,
Actually that is mentioned within the patent’s description. There are a few other aspects that build upon the ones that I mentioned as well. If you have more than one word, and each of the words has some level of implied local relevance, the combination of those words might as well. Good call.
Hi Joe,
Thank you.
Hi David,
Having lots of data, and a chance to do testing by showing different people some different things at times really helps.
Hi Travis,
The patent is describing the meanings and relationships of words to searchers in different places and languages, which types of results people click upon when they see them, the use of query terms as they are expanded upon in query sessions, and more. There’s an element of semantics in all of that.
There’s a link to the patent in the post if you want to explore that more fully.
Hi Dave,
I really enjoyed our discussions on when the search engines might show some localized results for particular queries. Happy that I wasn’t leading you astray either with my educated guesses.
Of course, Google may be doing other things as well, or doing some things that are somewhat different. But it is always really nice to see a patent that presents a process confirming some of the guesses that you’ve been making. 🙂
Hi Bob,
I’ve done a lot of those types of searches, too. New York style pizza versus Chicago style Pizza, and what happens when you look for New York style pizza in Chicago, and Chicago style pizza in New York and so on. It is fun trying to guess beforehand what might happen when you do things like look for [new york pizza palo alto]. 🙂
Hi Larry,
Good question. Definitely something to experiment with.
Hi OptimizePrime,
The patent does talk about canonical queries, where they might reduce certain words down to stems. Definitely worth digging deeper into the patent to get a better sense of what they were trying to do, and what they might possibly do.
Hi Alan,
Not sure that I agree. I think that the geographic relevance of some query terms are independent of the competitiveness of those queries. For instance, [pizza] is a query term that is both very competitive, and very likely to trigger a maps results. Now when you add a geographic term to a query, it may be less competitive and more likely to pull a map. But remember as well, that this patent is focusing upon returning localized results and not always maps results.
Hi Keith,
If it were easy, everyone would be competing with you. 🙂
Hi Egor,
It does explain why Google will show some very different results, even in cases when the locations being search from might be fairly close as well.
I suspect that you could find a few examples of “Google is mistakenly treating a website as local, although it is just as useful internationally” without too much difficulty. I’ve always considered it to be a good approach to include both locally relevant content and globally relevant content within the same site when place matters but the potential audience (and consumers) might also come from a much broader area. For example, people might travel a good distance to come to a museum, and while it wants to attract a local audience, those more distant travelers are still important.
Hi Chris,
There still is some difference in regional language in different places in the US, and other countries that can trigger localized results in some places and not others. That does mak things interesting.
Hi Malcolm,
Yes, your local results do tend to be a lot better if you actually change your location than they are when you let Google try to guess. Google does try some ways to get people to change or set their location, like giving people a chance to do that when they check weather or movie times and so on.
Clients will potentially see different results than you if you’re not in the same location. I ran into that back in 2009 when I first started seeing very high rankings for a client on a very broad and competitive term, but only from locations near them. Having an idea that Google may do this is something that can really benefit you, by making it possible to rank for terms like that in an area that may best benefit your clients.
Hi Dennis,
When Google localizes results, it doesn’t usually localize all the results. It includes globally relevant results as well. So the person in Australia searching for [freedom] may see a few localized results, but will also see many results that aren’t the store.
People from China might have to refine their queries to include some other terms that may help call forth the store. If I search for [freedom home furnishings] here in Virginia, the second result is the home page for the Australian site. There’s no way that I would want to see the Australian store on a search for just [freedom]. I don’t see a problem with that.
A lot of assumptions are made in the creation of algorithms, in the writing of patents, in the shaping of how search engines might work. I’ve always told people looking at patents that one of the most important things they can look at are the assumptions behind the patents. Some of those assumptions are helpful, and some of them aren’t. I’m not an apologist for Google, and love pointing out when those assumptions are faulty.
Have you actually set your location in Google to be local to you, or are you letting Google set it for you?
If it worked the way it’s described to work, as opposed to the faulty way that it’s implemented for you presently, would you like seeing some localized results for queries that might not actually be shown otherwise because there are plenty of other sites that are better optimized for a query, with higher pageranks and more relevance for the query?
If the localized results shown for you were Holland-based, would you be calling those assumptions faulty? 🙂
I think that aggregated user behaviour is becoming one of the biggest factors in chosing and localising search results. That’s what I seem to find, purely from a consumer standpoint, as it seems to be the most accurate.
As Malcolm has found, I often have to physically change my location for Google to get the right results for my local area. No quite to the extent of 400 miles, but given how many bigger towns and cities are close to my local area it can be frustrating.
As a user and a SEO, I like the move toward inclusion of more local results using a predictive algorithm. There have been many times that I’ve been disappointed not to see more local or map results. This should also make it easier for a local restaurant to rank over sites like restaurant.com, which is good for the user.
Definitely happy to see as a user of Google, that they are looking to localize more of their searches. It can be frustrating when your looking for a local store or restaurant, and constantly flooded with the big national chains that may or may not be in the area you live. Great article, love reading the specifics of the patents too.
I’ve seen local results when I search on some of my client’s services but only in some areas. I was wondering why local results were showing even when I didn’t put in a local in my search phrase.
Hi Alan
Not sure that i agree. I think that the geographic relevance of some query terms are independent of the competitiveness of those queries. For instance, [pizza] is a query term that is both very competitive, and very likely to trigger a maps results. Now when you add a geographic term to a query, it may be less competitive and more likely to pull a map. But remember as well, that this patent is focusing upon returning localized results and not always maps results.
Great post! I would have to agree with Dennis that there are assumptions being made by Google that could leave a bad taste in the mouth of some users.
It will be interesting to see what kind of affect this will have on a smaller local business trying to push out products to a national level. You do have to give Google credit for they consistent effort in trying to increase the user experience.
Does localized search become effective if it just based on ip location. National brands and companies still dominate in local searches. Suppose you search for coffee, there is a big chance that you will get Starbucks as result. I think local search will get better and it something more for mobile devices than for desktop computing.
Your Blog is very very good and fantastic.I have glimpsed local outcomes when I seek on some of my purchasers services but only in some areas. I was wondering why local results were displaying even when I did not put in a localizedizedized in my search phrase.
I think that aggregated user behaviour is becoming one of the biggest factors in chosing and localising search results. That’s what I seem to find, purely from a consumer standpoint, as it seems to be the most accurate.
Bill, interesting post. Local search definitely seems to be a priority for many of the small business clients we deal with.