Carter Maslan is the Director of Product Management for Google Local. He also spent 6 yearsas a Director of Product Management & Marketing at Microsoft, and 2 years at Inktomi prior to joining Google.
Eric Enge: One of the big changes in the past year was that Google Universal Search came out and we started to get local results inside of web results, starting with three results, if I remember correctly. But recently, it’s expanded up to ten results and I am curious as to what went into deciding to do that and what it says about how effective these results are for you?
Carter Maslan: In general, we started out with either one result or three results. If it was a navigational query or a unique place that we knew definitively, then we’d get a single result with a map, and if it was something broader we’d get the top three results.
What we realized is this that there was a whole set of broad categorical queries that really didn’t make sense to answer with one or three results. We also realized the users liked something that made the same compact screen region much more useful.
They favored something that conveyed the sense that there’s actually a pretty broad query behind the way we interpreted their search, and gave them the sense that there are more precise results that may be filtered geo-spatially. So it conveyed that there’s a bigger list of results, as opposed to picking what could be viewed as an arbitrary list out of a very broad pool.
Eric Enge: Right. I actually have an example about that. If you could type into Google the phrase “San Jose storage” it brings up ten different listings. See how 8 of them are from one vendor, and the information presented for each of their entries is identical?
Carter Maslan: Yes, in that case, even though there is nothing inherently wrong with a single business ranking prominently in the top results, we are working to modify the way we handle cases like that. The challenge spans multiple areas around the quality of the data, the way we index it, the way we score it, and the way we present the UI in showing it.
So we are working on variations on ways to handle those cases better, but that’s not to say that there’s something inherently wrong with one business having a prominent presence in the top ten.
Eric Enge: Right, I understand. In the example I sent you, that one company has eight of the top ten results and the information presented is identical for each one, until you click the “more” link.
Carter Maslan: Yes, and that’s not ideal. I haven’t looked at the case to know whether they are just naturally ranking very well or if there is just a redundant link. And I think that’s an area that we need to improve.
Eric Enge: Right. And it’s a very complicated problem, because you are dealing with data from many sources and the problem of verifying the data is very tough, isn’t it?
Carter Maslan: Yes. I mean we have an explosion of different content from all different websites, aggregators, directory providers, and our own Local Business Centers so there is constantly a plurality of opinions on facts you can search. You wouldn’t think opinion is related to facts, but that’s frequently the case.
So a big part of what we do in search quality is try to figure out what the right representation of a business is from a lot of different sources of good information.
Eric Enge: Can you talk about the Google Local Business Center?
Carter Maslan: This is one of many sources we use. With this, business owners can provide their information to us directly. Let us say they have the same address but they change their phone, and it hasn’t rippled through all of the other references on web pages that were posted from a year ago.
By going into Local Business Center they can specify their preferred phone number and we show that whenever we are presenting basic facts of their business.
Eric Enge: Right. And, I understand that when you are submitting a small number of locations manually, you actually go back and verify that the person submitting it gave correct information and these sorts of things.
Carter Maslan: It is always a balancing act on the level of verification versus the simplicity of getting through the process, and we are experimenting with that. But yes, we do include verification so we can be confident in the facts we have.
Eric Enge: Right. There is also a method for providing a feed as well, but you are not verifying the locations within the feed, which would make it less trusted data.
Carter Maslan: That is also changing. We do have different parameters on how much verification is needed for different feeds. And we are adjusting that, because the ability to have bulk uploads is an extremely useful feature. But at the same time it’s also subject to abuse.
Eric Enge: Right.
Carter Maslan: And so, we are combating that with a combination of techniques, both from looking at the patterns of abuse and trying to figure that out regardless of more owner verifications. We will also include additional verification in cases where we really think that’s required.
Eric Enge: So far you are doing verification by checking other data sources, right?
Carter Maslan: If you go into Local Business Center and add your business, you can get a mailer sent to your address or you can verify your phone number via a telephone directory.
Eric Enge: But, what happens when you have six hundred locations?
Carter Maslan: That’s something that we are taking a look at, because the bulk upload has been the primary source of map spam, so we are reexamining that. We are looking at a combination of factors to address it either through added verification or through, for example, consistency with the listings that have been manually verified.
So, we could take a case where somebody has legitimately got 600 store locations, but there are techniques that detect the abuse by including information that’s not similar to the data that they have manually verified. We may go to a more stringent verification. What we are looking at right now is how much can we address algorithmically and how much needs to be done via a more stringent verification.
Eric Enge: So that is definitely a challenging problem overall, but I think it’s interesting one. There are other data sources, like Yellow Pages and SuperPages.com. There are also syndicators like Localeze and infoUSA, and also Local review sites. And I understand that these are all sources of data that you use as well?
Carter Maslan: Right. We try to find as many references to the businesses as possible, and then determine if they are referencing the same business, and then present that as related information to that business. So we try to look as comprehensively as possible for sources of information on a business.
Eric Enge: Right. So if someone owns a business with a lot of locations, what’s the best way for them to ensure you have the right data, and be confident of it as well?
Carter Maslan: Well, the Local Business Center is designed to help businesses if they have conflicting sources of data on the internet.
The most common frustration is that they will see extraneous references to old locations or incorrect information that was transcribed incorrectly by somebody who entered it. And so, they will ask how they can vacuum up the poor references?
That’s what we are trying to enable with Local Business Center, by providing a way that help them clean up their data. That can sometimes be difficult, especially if you are unable to verify your old location. That’s one of the challenges that we are trying to address right now, strategies for cases where it’s difficult to verify something that was either completely incorrect or unable to be verified, because of the fact that it was just so wrong. For the mainstream cases, we want the Local Business Center to help be the vacuum cleaner for addressing data quality problems.
Eric Enge: Right. So, if you have data in a Local Business Center that you have uploaded, and there is information on the web that contradicts it, what does Google do, and what is the best way for a business to address that?
Carter Maslan: Let’s say that you were in suite 400 at 100 Main Street in a particular town, and then you changed to be in suite 201. And, you gave us a feed in Local Business Center that said “supply that new suite number,” even though there were all these references to your old suite number on the web.
In that case, because the address goes to the same location and you haven’t changed your business name and your phone number, there is enough confidence that we would replace the information that was stale with your current up-to-date snapshot. It only gets messier and messier as the data is further and further away from any known prior representation of your business.
Eric Enge: Right. Say you move your office from San Jose to Seattle?
Carter Maslan: Exactly. That is where the Local Business Center helps in making that explicit link. The ideal case is that you can still verify your business at the old location, because then you can say okay, I am in San Francisco; I am at my old office. I am about to move to Seattle and I will verify both.
I am the owner of this business in San Francisco, and I am telling you that the new location is in Seattle. Then, we can continue to identify prior references as being about the same business even if it’s now in Seattle. It gets complicated if we don’t have that link, because we are trying to help people present their information correctly when there are not reliable ways to ensure that they are not abusing the submission of business information. So, that’s an ongoing challenge.
Eric Enge: Right. And of course, another challenge is when you have a business with a large number of locations that are in places like Yellow Pages and SuperPages. Let’s say they have a 1,000 locations. So they upload data to the Local Business Center. There are differences between that data and what Google finds on the Internet, and they have just begun working directly with you. That’s another scenario which I imagine is pretty challenging.
Carter Maslan: Well, that’s the same as if there were web pages or any other source referring to their listing. As long as it is close enough where their preferred representation of their basic business facts can be matched with other references, we can help them. Whenever we really don’t have the confidence to know that they are talking about the same thing, we refer them back to wherever the original source of that information was from.
Eric Enge: Right. So it does make sense for businesses, small and large, to try to get the data correctly implemented in the other important local reference sources out there?
Carter Maslan: Yes. We live in a world of many different sources of information, and as much as we try to make it convenient, so you can control the way that your basic information is presented, there are always going to be many, many, more sources of information. So unfortunately, keeping the original sources from all the different places accurate is a still an active task people have to be mindful of. .
Eric Enge: Localeze and InfoUSA seem to imply that they have a trusted feed that goes directly into Google. The implication is that they have done the manual verification and you trust them as a data source. Is that a fair characterization or is it a little different than that?
Carter Maslan: We don’t talk publicly about any of the relationships that we have. But regardless of individual sources of information, I think the challenge is even the best sources are often incorrect 10%- 15 % of time and that is just nature of businesses moving their locations. There is constantly a lag between what happens in the physical world and when it’s digitally recorded and propagated through all web pages and other databases, so it is an ongoing issue that exists regardless of how many sources there are supplying information.
Eric Enge: Right. The process is still prone to error because of the reasons you outlined, but it does seem to offer something that’s closer to what you might call a gold record.
Carter Maslan: Well, I guess that is the same kind of concept behind submitting a self-verified listing in the Local Business Center. If you were to add each individual location and verify it individually, then we would individually verify accuracy by calling the phone number or visiting the address and so on. I think the challenge is even when you are doing a re-verification on a bulk basis, you still have a lag time.
Eric Enge: Yes, There are really two problems- one is getting the current state of your business right and the other is maintaining that state.
Carter Maslan: Right.
Eric Enge: So the best thing to do at this moment is to do multiple individual uploads to Local Business Center because of the independent verification, right?
Carter Maslan: Yes, that is the highest form of verification and priority of the data, and it gives you the granular control to supplement with photos and videos and things like that. We would still work with feeds and other sources too, though.
Eric Enge: So in the future the bulk upload in the Local Business Center will have the same level of verification?
Carter Maslan: Well that is something we are looking at. We are experimenting to find out how many of the locations in a chain need to be individually verified. Or if you’ve verified ten or twenty, whether that is enough. We are just trying to find the sufficient amount to know and trust.
We also work to detect abuse algorithmically to support that.
Eric Enge: You are trying to make it bit more scalable.
Carter Maslan: Exactly.
Eric Enge: Is there a way a business can determine the source of information that you have used when they see their Google Maps results’ page?
Carter Maslan: I think most of the time you can do it just by searching. So for example, there is a site operator that applies equally on local searches like site: operator. If you are wondering if your listing came from a specific review site, you could just use the site: operator and see what information came from that domain. That is one technique.
Eric Enge: If there are four places on which your information shows up, you need to try out four places and see what one matches up?
Carter Maslan: Right. So if I am Joe’s Barbeque and I want to see, where the information came from about my listing, that comes back from Google Maps or in Google web search.
Eric Enge: So for example, if you see the wrong phone number listed you can just type in the phone number in web search and find the source of the error. It happens for businesses a lot of time, they see the listing and they realize it’s an error, and then they are try to figure out what they need to do to fix it.
Carter Maslan: Right. I think search is the probably one of the best tools to try to find the original source.
Eric Enge: David Mihm recently published a study on what the various local search experts thought were the most important factors in local search ranking. The top few were claiming your listing, proximity to centroid, picking a proper categorization product key words in the title records like in a feed for example, and then reviews. Can you discuss these things a little bit, and what you advise people to think about when they are trying to address these things?
Carter Maslan: Sure. So claiming your listing by itself is not so much the issue but what is important is having good control over the way that your business is described. I think the value of claiming your listing is that you can assign it a category that is of your choosing that is likely to be the way that somebody may be searching for your business. Also, a month or two ago we extended the category system to be open to users contribution of categories, which was kind of big move for us.
In that sense we were struggling with how important it is to encourage some level of convergence, just so that information is discoverable and in a group. We wanted to make sure that if people do a search for Thai restaurants they get a good search on them. We wanted people to converge on terms. For example, if you are a Thai restaurant, you are easy to categorize, but if you are an Asian fusion with Cuban influence restaurant, it’s much harder.
We wanted to enable the long tail of truly descriptive and accurate categories as well. So I think one of the key advantages in claiming a listing as a business owner is that you can accurately categorize something that could have been poorly categorized listings without your direct editing.
Eric Enge: Right. So the act of claiming your listing does not really matter that much, but the things that it enables you to do does matter.
Carter Maslan: Exactly, as for proximity to centroid that is an old issue which we’ve addressed, so that is not too important any more. Early in the history of local search people would try to setup locations near the centroid, but that is just not that important any more.
In terms of verification, it is not that the categories themselves are verified but that we try to have a good carrot and a good stick. So the carrot is that you can categorize your business in way people might use to find you. The stick is that we do take action against people who try to spam the system.
What we are hoping is that the majority of the time people are honest with their data, and that we can verify it by one means or another. I believe that the great majority of people are well-intentioned, and they can improve the user experience by having their stuff discovered. Proper categorization is definitely key.
Products and keywords in the title record is much the same as it is with web search, where it is a double-edged issue. You don’t want to spam your title, in general, you just want to describe your business accurately.
Eric Enge: Sure, we are not necessarily talking about spamming your title here, but if you are in the blue widget business and your product is called the Alpha-136, you can put Alpha-136 in your title, or you can put blue widget. People are more likely to search on blue widget.
Carter Maslan: If you were coming up with a business from scratch, and there was one title that had more searches and on top of people’s mind , then that would have an advantage. But I think what we have seen is people literally just submit variations in their business title as their business name.
That is what we don’t want to have- a whole bunch of confusing listings. We had cosmetic surgeons doing that over and over again for every different alternative synonym for a particular procedure, and that is what I am sensitive to.
We do look at reviews. It is not so much something that will radically change where you are in the results, but we do consider it.
Eric Enge: Right. So that brings us to the concept of references. In the world of web search rankings are impacted by physical HTML anchor links from one page to another. Do references to your business (sites that list your phone number, address and business name for example) that don’t happen to include a link influence rankings in local search?
Carter Maslan: Just the fact that there is an implicit link in the geospatial world, is not as strong as the explicit anchor text that goes straight to an URL that we know is a definitive domain for a business. But yes, it does help to have your business well-described and geo-coded in references on different pages.
Eric Enge: Right, so a link may still count for more, but this is an additional fact that you can look at.
Carter Maslan: Right, exactly.
Eric Enge: This is a good thing for people to be aware of. It should be interesting for you to get your business talked about on other sites anyway, but it is useful to understand that it is a factor in local, which sounds completely appropriate.
Carter Maslan: Right, yes.
Eric Enge: I first heard about KML a while back and I have seen some stuff suggesting that KML would allow you to provide a deeper level of metadata about your business. I believe people have talked about using this within the context of the sitemap files, is that something that is useful for people to do?
Carter Maslan: Yes, it helps a lot in knowing the precise geographic location that is being described by a page so it is definitely good to have a sitemap that references a KML file with an accurate description of the entities referenced geographically.
Eric Enge: Right, the sitemap file is something that is authenticated, and therefore the KML file would be by implication authenticated.
Carter Maslan: Correct. It helps both from that perspective and simply just knowing that there is additional information about that place, and that it is also correlated with that web page, helps.
Eric Enge: Does it make sense for business to work with aggregators because they generate additional listings?
Carter Maslan: Yes, basically we are always looking for the best sources in information across the web, to the extent that the sites that reference your business have good, additional information. Whether it is features of your business, or reviews of your business, or additional context for the products that you sell, that is helpful.
Being online and having a good presence online will continue to be important. There will always be multiple perspectives on businesses and we are interested in being a great starting point to find the best information related to a place.
Eric Enge: Right, so you mentioned reviews of the businesses. I would imagine that there are other kinds of data that might be interesting to combine with information about a particular business. It could be locations or other services in the area that are closely related to whatever your business or service is, if there is a geographic climate dependency in your business then information about that would be valuable.
Carter Maslan: That’s right. So then there are the basic facts about your business and there are all kinds of enhanced content that is more descriptive about your business. So to the extent that enhanced content helps a user decipher what your business is about and what it provides is always going to be helpful.
Eric Enge: I really appreciate your taking the time to chat about this with me today.
Carter Maslan: Thanks for taking the time to speak with me.