Two Different Restaurants
In my college days, I cooked at some local restaurants (free meals made it an attractive option for a starving college student). One of the restaurants was in the center of town, at one end of Main Street, and it was a popular place for residents who returned repeatedly. It had a great reputation, and word-of-mouth propelled advertising for the place. Another dining venue I worked at was outside of the center of town, nearby an interstate highway. It didn’t have a great reputation and very few regular customers, except for people who would stop during mealtime from the busy interstate. The “food” sign from the highway attracted most of the traffic to its dining room.
Funny thing is that most of the regulars that frequented the first restaurant rarely had to look up its location because it was so well known. Most of the second restaurant people had never been there before and relied upon the highway sign. There’s another restaurant in that location now, and I do not doubt that many people find it via maps or navigation on their mobile devices or in-car navigation. I mention this because I have some issues with a recently granted Google patent.
The patent, which got granted to Google this week, describes how the search engine may look at “popularity” signals, such as how often people look up driving directions for businesses with locations that they can visit in person. It also tells us that in some cases, such as where driving directions lookups are sparse, Google might look at some alternative signals, such as reviews of those businesses, to use a popularity signal to rank pages.
The 4 inventors listed on the patent are also the 4 authors of a Google white paper that I cited in a blog post a couple of years ago, in GPS to Correct Google Maps and Driving Directions as a Local Search Ranking Factor?. The paper is Hyper-Local, Directions-Based Ranking of Places (pdf).
Correlation between Driving Directions Requests and Number of Reviews
While I cited the white paper for its description of the use of driving directions from Google as a potential ranking signal, what I didn’t include in that post was a section in the paper that considered the use of those driving directions as compared to the number of reviews for businesses. That conclusion was pretty interesting.
It seems that there’s a pretty close correlation between the number of reviews that businesses receive and the number of times that people look up driving directions to those places. I left out the details of the investigation that Google conducted (the link is above if you want more details), but here are conclusions from the comparison:
In this experiment, we test the feasibility of using driving directions logs to proxy for the popularity/importance of a place. We compare the correlation of the driving directions-based signal with the number of reviews for a place, which is a commonly accepted measure of popularity. The number of reviews was extracted via Google’s business directory, and it is the total number of reviews found in various data sources on the Web…
…There is a clear correlation between the rankings of the results and the number of reviews that a place has, which shows that the driving directions logs indicate the popularity of a place.
Why A Correlation Between The Rankings of The Results And The Number of Reviews Is Important
The paper also explains why this conclusion is important:
- Driving Directions logs are cheap to collect
- Driving Directions can get requested for specific places much more frequently than user reviews can get left for places
- Driving Directions logs provide “near real-time” indications of changing sentiment for places that is often hard to capture with other types of signals, such as reviews and even PageRank
- Driving Directions logs are usually available for broader types of locations than reviews are
Besides the patent and the paper sharing the same authors, the subject matter significantly affects one, and the other overlaps significantly.
The patent is:
Directions-based ranking of places returned by local search queries
Invented by Hector Gonzalez, Petros Venetis, Christian S. Jensen, and Alon Y. Halevy
Assigned to Google
US Patent 8,538,973
Granted September 17, 2013
Filed: June 4, 2010
Abstract
A system and a method for ranking search results of local search queries. A local search query and a current location of a user can get received. Next, two or more places that meet the local search query get identified. For each respective place, a corresponding distance from the user’s current location to the respective place is also identified.
The two or more places are then ranked following scores that can get based, at least in part, on the popularity of the two or more places and the corresponding distances from the current location of the user to produce a set of ranked places. The ranked set of places is then provided to the user.
The patent provides some other reasons why Driving Direction requests are useful as popularity ranking signals for local results and can provide more desirable results to searchers:
- Driving direction requests can get used only to help rank pages that are within a certain distance – people may not drive more than a certain distance for a meal, though they may drive further for an antique shop or to visit a landmark
- If the directions request is for pedestrian travel, the distances may be even shorter
- Rankings based upon things such as the numbers of links to web pages may not accurately indicate the actual popularity of a business that people can visit in person
Driving Directions Requests Search Features
The patent does tell us that there is a difference between stationary directions requests and mobile requests. The distances collected for mobile requests tend to pinpoint the actual locations of searchers much better than for people making requests from desktops since mobile locations can get traced back to Wifi locations or cellphone tower locations (via triangulation).
When information gets collected about directions, the time and day of the request and the distance indicated can be saved. This can show that one restaurant might be more popular for breakfast, and a nightclub that serves food might be more popular around or after dinner. It can help the search engine know how far people might be willing to travel for different types of businesses, too.
When looking at features such as time, other popularity signals might get looked at as well, such as:
- User ratings of the place
- User reviews of the place
- A query independent page rank of a web page associated with the place
If people are more willing to travel further to one place over another, that’s also a signal that can get considered in local search rankings for places.
Driving Directions Conclusion
In the example I started this post with, I write about two different restaurants. The first was very easy to find and very popular based upon word-of-mouth advertising. It was much more likely to earn glowing reviews. It was also straightforward to find.
The second had a much less robust reputation but benefited from being near a busy highway delivering new diners to the place who had never heard of it before and might never return. The second restaurant was less than half a mile from that interstate and was one that people requested driving directions to make it more likely that they would arrive at their destination.
People visiting the second restaurant were unlikely to leave a review. The food wasn’t exceptional, but it was adequate to fill the bellies of travelers in search of a quick meal. It was popular based on its proximity to the heavily traveled interstate.
A couple of days ago, I listened as a friend gave driving directions to a hardware store about 20 miles away to someone interested in finding the best hardware store that they could. There are other hardware stores with a couple of miles that is nowhere as stocked with high-quality choices.
A place that gets great reviews may be a better choice than one that is popular based upon the distance people might travel to get to it. However, chances are that for most of my personal needs, I’ll never visit that much more distant hardware store.
We don’t know for certain if Google is using popularity signals from places such as their driving direction logs. We do know that they like to collect and show reviews for places since many people like to look for reviews.
Google may be using driving directions information alone, or in conjunction with review information to rank local search results. Chances are that given a chance to spend time searching for businesses, people might look at both types of information. If someone is searching via a mobile device, they might not have the time to look at the review information.
Thanks for a very interesting read, as always, Bill.
I hope driving directions doesn’t become a noticeably important ranking factor. Intuitively, I can’t see how it would. But, like you, I think it’s problematic. I can just picture the Fiverr gigs now: “130 driving directions from different IPs – only $5!†Which makes me wonder whether Google might give extra weight to driving directions requested by a user who’s signed into his/her Google account. Still easy to game, but a little harder. It just seems that to the extent directions becomes a ranking factor, there may be abuse.
Hi Phil,
Than you very much.
I could see why Google would experiment with Driving Direction information as a potential ranking signal, especially since it does evidence a pretty strong interest in the company being looked up. Chances are that many people who might do that, at least on mobile devices, are likely to be logged into their Google Accounts. A mobile lookup can provide more information than a desktop one, since Google likely has a better idea of where the searcher was at the time of the driving directions request, by knowing Wifi or cell tower triangulation information for the search.
A desktop search for a local business, and one where the request indicates a different starting point, such as a city and state name that the searcher might be planning on visiting, might not be considered as strong a signal and might be weighted differently.
If searcher is logged in, having the Google Account information as well might make it easier for Google to know more about the trustworthiness of the account and the search. If that account originates in another country, or a stream of requests for driving directions comes from it for diverse locations around the country, that’s very different than a search for driving directions to a point where the searcher might then be seen to connect to the Web for navigation to that point, or connect to receive Google Now or Google FieldTrip results near the final destination of those driving directions.
It’s possible that attempts to game a signal like this through fiver might run into problems that would cause those specific signals to be ignored.
What bothers me more is that some businesses might be extremely busy and popular without much of a need for people to look up directions for them. If everyone knows that there’s a great and very well known restaurant down at the end of Main Street, and road into town goes right into Main Street, they might not bother at all looking up its location on Google Maps.
That is an interesting thought. It seems that google would take into account amount of driving directions requested but would at the same time, take into account reviews. Also, it seems that reviews would be held at a higher weight than just driving directions. Sometimes when I don’t know an exact address, I just type in something I know is close. Getting directions isn’t a sure was to know that a user actually went to the restaurant but a review is.
Interesting stuff Bill, thanks for reviewing all that information.
I’m assuming that this is relevant to directions in general on Google. For example, I live in London and frequently look up a place on Maps not because I’m specifically looking for directions but to see how long it’s going to take me to get there on the Underground. And most of those lookups are done at home, not on mobile.
This sentence doesn’t make any sense:
…There is a clear correlation between the rankings of the results and the number of reviews that a place has, which shows that the directions logs are indicative of the popularity of a place.
if rankings are correlated with number of reviews, why would that show that directions logs are indicative of anything…?
Hi Mark,
I cut out some of the middle of that passage, but the whole thing can be found on page 295 on the paper, Hyper-Local, Directions-Based Ranking of Places (pdf).
Good read, Bill.
I wonder if there is/will be any driving directions data gathering from third party websites/apps. Would be interesting to see if Google could pull that data as well.
Simply insofar as google might consider a variety of signals to which they have access that could provide signals of popularity is nice info to know…and interesting as to which of the many signals they have access to is additionally good information.
But…do the writers of the patent have access to dashboard statistic that google delivers to smb’s or to they have a more reliable and trustworthy data base? I would hope there is something dramatically more trustworthy than the dashboard stats google has delivered over time. I’ve found dashboard stats to the web page, the local page and driving directions vary widely by time periods…in fact I’ve seen trends reverse. Then I checked with other webmasters. The trends had reversed for those webmasters following different businesses. The only thing that was consistent was that the trends being delivered to webmasters were similar.
Uhhhh…..Is google telling us that from March to July the huge overwhelming volume of searchers go to websites for every type of business everywhere….and then from August to the following February searchers always go to the Local page and never go to websites????
Seriously what is delivered inside the dashboard has always been suspicious. Now does Google hoard accurate data and use it? Apparently the writers of this patent believe that is the case.
My question is ..if google can get accurate information it believes in to write a patent..why can’t or why won’t it deliver that information to the operators themselves that have google records?
Okay…so I’m off topic somewhat. Sorry.
Phil brings up an interesting point. If the application of this patent became observable in some way…ie for some reason spammers could get a feel for aggregate directions requested…would they use that to spam the Maps/Pac algo’s?
Probably….the spammers are on top of their games. In the last couple of weeks there were a rash of spammed maps results as noted in Linda Bouquet’s blog. Pure spam.
But back to topic. Google has the engineers. I only have intuition and some experiences as you described Bill as to how consumers act. I would suggest that asking for directions is certainly an expression of interest though and one that should probably merit consideration from an algo driven company.
Thanks for the enlightenment in how goog’s engineers are thinking these days.
Dave
Bill,
The implication that Google didn’t mention in the original paper is – if you can compare and correlate driving directions with reviews, then it’s potentially a great way to identify review spammers.
Actually you commented on my article on the paper some time back and had said you’d love to see a patent associated with the paper – you got your wish!
I had a few best practices I suggested people might consider, including, ironically, not making it too easy for people to get directions off of your website – better to encourage them look it up on their mobile device somehow:
http://www.coconutheadphones.com/how-to-use-driving-directions-in-local-search-seo-for-google-places/
I am curious about this statement:
A mobile lookup can provide more information than a desktop one, since Google likely has a better idea of where the searcher was at the time of the driving directions request, by knowing Wifi or cell tower triangulation information for the search.
If a search is done on a mobile device using a web browser and not an app, does exact location other than the IP (ie allowing them to do Wifi or cell tower triangulation) information get passed to the search engine? I can see it being passed perhaps if an app was used, but I have not heard of it being passed if someone was doing a web search on mobile. It peaked my curiosity. Thanks
I would suggest that the future is now and that Google is currently using some mobile signals in their current ranking algo. Certainly, as David points out, Google has invested in a number of technologies (Coupons, Wallet, Offers, PunchD, Talkbing, check-ins) that will give Google on the ground signals as to whether a consumer actually visited a location and consummated a sale. Most of these have not achieved any sort of scale and are forward facing investments that attempt to close the “search to sale†loop for analytics. All could also provide popularity signals to Google when they do achieve some scale.
But Google has two very widespread highly trusted technologies, Driving Directions and Android, that function at huge scale and could be providing signals now.
It is pretty incredible what is all being gathered nowadays. I sometimes yearn for the 1970/80’s again. Things were so much simpler then. Or so they seemed anyway.
Thanks for the post.
George
I’ll never eat at a place that is located in a good location these days unless the Yelp reviews are good. Still don’t 100% trust Google Reviews yet. In a way they are forcing businesses to use their reviews since it affects search rankings, and they have a monopoly on search
Very interesting Bill. Actually this happens to correlate with a recent drop in rankings for one of our clients that dropped simultaneously as driving direction requests stopped. Of course there are many factors, always.. nevertheless interesting.
Another good article. It makes sense that driving directions could be used in addition to reviews in google’s local algorithm. I would think reviews will still play some part in the ranking. I’ve also been taught that citations are critical for local SEO (i.e. yelp, manta, brownbook)… I wonder if this patent will discount those as well.