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: The thing that everybody has been buzzing about is the new service area business tool that has started to appear in Google. It would be great if you could talk a little bit about that.
Carter Maslan: We’ve known for a long time that our listings were constrained to those that had retail storefronts. There were a lot of cases where plumbers and other service people like this were only listed by their home address. It was kind of a dilemma, because they liked to be easily found but they didn’t want to disclose their home address.
There were also people looking in the area where they served, but not where the physical storefront was. We’ve been looking for ways to handle this, and we are experimenting with ways for people to describe both a retail storefront and an area in which they serve, either in combination or separately.
Eric Enge: Are you testing this in certain markets or just with certain users?
Carter Maslan: What we are trying to do is get a sense for how people are using it. It’s pretty broadly accessible, and we were checking to see how people use it before officially announcing it.
Eric Enge: So you are basically saying that most everybody can get it at this point off launch?
Carter Maslan: Right.
Eric Enge: Do you have any idea how many businesses could potentially add to the overall maps index?
Carter Maslan: Even though we don’t have specific numbers to share, we know that it’s a huge portion, as there are a ton of home-based and service businesses in this country. I don’t have a specific number, but I think it is at least a third of the overall total, and I am assuming there is going to be a big group of new businesses that can now list themselves legitimately. I think in the past there were people trying to shoehorn themselves into the orientation towards retail store fronts, and this ended up creating problems because we viewed what they were doing as spammy behavior, so I think this is going to help with data quality.
People and businesses with the intent to give us accurate information now have a way to do that without pushing the boundaries of what we originally intended the product to do. I think it’s going to cleanup data quality and I think it’s going to appeal to a large chunk of businesses that we couldn’t serve before.
Eric Enge: Businesses used to go get post office boxes and things like that to get themselves listed in the Maps index, correct?
Carter Maslan: Right, and giving the end user a post office box as a pin on the map is really not helpful. I am happy that we are making progress with this so that will no longer be the case. If a business really cares about its customers knowing where its PO Box is, I think it’ll be clear that it should be a service area business.
Eric Enge: What is it that defines a business as a service area business? What kinds of things are and are not included?
Carter Maslan: It’s primarily defined by whether or not the business brings its services to the customer. Whether that’s someone who takes people on boat trips in Miami, delivers pizza from their pizza storefront, or a plumber that comes to people’s homes; any kind of service that comes to people as opposed to the customer going to a physical place.
The example I use is my son’s dog-care service. I don’t want our home address listed as where he is working from on the map because no one is going to drop his or her dog off at our house. My son goes and takes people’s dogs on daily adventures and things like that. In that scenario, it’s great for him to be able to have a listing but still not display his address, as that is really irrelevant to the people who are asking him to take their dogs out.
Eric Enge: You mentioned something about what Google has done for people who just simply don’t want to list their business address.
Carter Maslan: An example of this is a consultant working from home who doesn’t necessarily go to the customer, but the customer can still call them in order to obtain its services. In my view, a consultant or firm like this is still service-oriented in the sense that the service may be delivered over the phone or by computer, but it’s still a company that people will look up and call when they need help with a particular project. So a company can still disclose its phone number without having an address.
Eric Enge: What happens when businesses try to abuse this? The example I have in my mind is a guy running a wine store who tries to get listed in cities where he is not physically located.
Carter Maslan: This is where we could get the question of servicing the entire country or just the area where the particular company is located. Our overall point of view is that we want people to be accurate in regards to how their business operates so people can find it accurately. If a company doesn’t provide anything special in a particular area, it’s probably not that useful as a local search result. If a business has nothing to do with its location, then it’s not really that relevant to have a local listing.
Eric Enge: What are you doing to manage that kind of problem?
Carter Maslan: It’s the same kind of thing we do with all kinds of spam or marginal SEO. We are looking to ensure listings are consistent with reality and people are finding them helpful to help us understand whether or not the listings are relevant and deserving of their ranking?
Eric Enge: But a plumber who clearly does visit client locations could potentially advertise that his service area is 1,000 miles, correct?
Carter Maslan: Yes, that’s right. There are a lot of signals regarding whether or not this is suitable. If you know that it’s highly unlikely that someone in the plumbing business is going to pack up their van and drive a 1,000 miles to do a sink repair, then there are things that matter categorically.
If a company offers a special travel guide service where they really do take people to a whole bunch of coastal areas to snorkel and scuba dive, that’s more correlated than the plumber example. I think there is enough information about businesses and the way people use them to determine if the advertisement or claim about the business is correct, but that’s part of what we are experimenting with. We want to see how this goes in terms of usage, and it’s looking really positive that people are using it for what it’s intended for.
Eric Enge: So you can setup metrics by business type?
Carter Maslan: Yes, it’s one signal of many. Because there are people trying to find weaknesses in the way that we score and rank in order to exploit those rankings, we go back and try to cure them. The reason I am not being that specific is just because of the cycle, but there are a lot of different signals we have that will help us manage abuse.
Eric Enge: Another related question is does this reduce the importance of proximity in local search rankings?
Carter Maslan: We think we have done a pretty good job mitigating that already. In the early days of local search, we had an incident where people were trying to put their locations and virtual post office boxes at the center of New York City. I think we’ve mitigated that already, however, so this shouldn’t help or hurt that.
Eric Enge: What about the enhanced local listing stuff that you had running for a while in San Jose and Houston. Is that something that’s in the process of being extended?
Carter Maslan: I am not the primary person in charge of this at Google, but I can say that they are experimenting with it to see how it does. I don’t know of any specific rollout schedule, but I know that they want to make that work out, and they are running trials in a few cities.
Eric Enge: The next line of questioning relates to the relationship between businesses that have a website and how those are used together with local search as opposed to businesses without a website. For businesses that don’t have a website, you have something called Place Pages, correct?
Carter Maslan: Yes, Place Pages. What we are trying to do is present the best of the sources about that place above the full search result. We view Place Pages as search results that are all about that place, and the overall goal is to find the best of the geo-annotated web and get users to the best sources of information about that place.
Eric Enge: Basically, you aggregate data from other locations to create a profile page for that business?
Carter Maslan: I think there is an element of it that can be considered a profile, with information like a basic address, phone, contact, and category information. It’s a nugget of information about that business that helps form the seed for finding all the other references around the web. Right now it may seem like a profile, but it really is meant to ultimately produce the best searches that we can for that place.
I kind of view it as a hybrid, where there is that nugget of core information that the business helps provide, and then there is all of the information that we see referencing that business from around the web.
Eric Enge: So you are picking up data from a variety of web references, and you are appropriately integrating some of that data into the place?
Carter Maslan: Right.
Eric Enge: Let’s say a business does have a main website with links to the website for one individual location and the Google local business listing. Does that business also potentially get a Place Page?
Carter Maslan: Yes. You can basically think of the Place Pages as a collection of all the general sources of information about that place from the Internet. Then the homepage of the business is its authority page if the listing has been claimed with a local business center. So Place Pages show both search result and the company’s own homepage. A lot of people find that a business’ homepage can be a great target, but restaurant websites can often seem more like brochures. If people want to get information about what the New York Times said about that restaurant, along with other reviews and things like that, then Place Page helps with that.
Eric Enge: Let’s say you have more than one location, 100 for example. In your view, is it helpful to have individual pages on the website for all of the locations? Also, is it helpful to have the Google local business center linked to each of those individual pages rather than having 100 locations that point to a single web address?
Carter Maslan: I can tell you what I think the ideal end state is, and there are various levels of getting there. Ultimately, we would like to have the store-specific page known so that people can just click through and see today’s specials and any kind of adjustments for that particular day. We would love to have all of that information on a direct click to the most specific page for that location.
That’s what we encourage, but there are still a lot of chains and things that just link to their top-level domain. I guess it’s a split answer. We want to get to a store specific page, but we are not uniformly there across all of the businesses.
Eric Enge: Could that potentially be encouraged by making it a ranking factor, for example?
Carter Maslan: Yes. I guess there are two sides to it. If you create a store-specific page that really just has an address, it wouldn’t be as helpful as having some genuinely good content on the page that the user would really appreciate having as the first click-through experience. That’s what I think we need to work through.
We don’t want to arbitrarily tell people that they must create a store-specific page, because we are really just trying to find the most useful page for that business. That’s why I am not so definitive on the store-specific page or not. I really just want what’s best for the retailer, store or businesses, first and foremost giving the user what he would want to see when he clicks on that business.
Eric Enge: Say you have a store-specific page that lists specific and individual things about just one store location. Depending on the kind of business that could be an inventory list that shows you’ve got extra stock?
Carter Maslan: There is a chain of stores that carries yoga equipment that my wife really likes. They have special yoga instruction, carry special brands, and host lectures on some special days. There are all kinds of things that the retailer does that relate to that specific store location, and there is also a general corporate catalogue page. So this is not black and white, and even though we want to encourage it, it’s not that there is a definitive guidance saying companies need to have that page.
Eric Enge: Obviously it’s good if there is a quality page with information unique and specific to each location.
Carter Maslan: Yes, that’s great. If we know that there’s good information about that page, then that helps on search and the snippets that we can show on the search results, because we know that the page is referencing that place. It does help even if it ends up not being the page that you list as your primary homepage. If there is good content that we know is content about that place, then it helps us do a better job with query results.
If a company has a page that’s store-specific and talks about its class schedule, and there is one that says its holding Tai Chi class tonight and someone is searching for places to do Tai Chi, then that helps us to score it. If a lot of people have found that page helpful about the Tai Chi class, then when people search for Tai Chi we would know that that location has something to do with Tai Chi.
Eric Enge: So let’s talk a little bit about error reporting, which is something that I think there has been some change in lately allowing businesses to report errors. Can you talk a little bit about that process and what the intent is?
Carter Maslan: The intent was to make it easier for people. Say they had an experience where they found something incorrect, and even though we would like the edit function to be used if they know the answer, a lot of people just know that it’s wrong and don’t necessarily know how to fix it. So we are also providing that option for people to tell us if they know there is a problem, even if they don’t know much more beyond that.
That is helpful to us in terms of identifying errors we’ve made algorithmically or the quality of data sources that we are trusting. It’s very helpful for improving the overall accuracy of the data, and that’s why that Report a Problem launch was geared towards increasing the volume and ease of people being able to give us a heads-up.
Eric Enge: One example scenario is if a business is listed at 123 Main Street and for some reason your database might say that 123 Main Street isn’t a valid address.
Carter Maslan: Yes. It is also possible for an address to get entered that throws off our geo-coding of the address, and a business owner may not notice it at the time they enter their listing. Once someone realizes it is a wrong geo-code, there has to be a way to fix that. Once we make an improvement and learn more about the street segment, there is a lightweight way to just say the marker is in the wrong place. This helps us get better at geo-coding and helps us improve the accuracy of the listing.
Another problem business owners sometimes encountered was when they had outdated information from some prior business location. They were always stuck in this dilemma of cleaning up old, outdated listings because they couldn’t receive any verification mail there. For some reason, it was never apparent to people that they could go and edit or remove a listing. We wanted to make it clear to owners that they could do that.
Eric Enge: How do you determine that a listing has been reviewed? There has to be some sort of error checking, correct?
Carter Maslan: It’s a combination of clues where you could picture how corroborated some piece of information is with other pieces of information. We are looking for ways to efficiently identify when something is accurate or not, and that could be through other users corroborating or other sources of data either corroborating or refuting it. We are just looking at a variety of techniques to help understand the trust level behind any particular report.
Eric Enge: So one thing you can do is you can look for multiple reports of the same issue?
Carter Maslan: Right, exactly.
Eric Enge: Or maybe you drew one conclusion, but you had to make a decision because there were two conflicting pieces of information on the web and you made the wrong decision. Now you can go back and do it algorithmically for a lot of situations, which is great. But presumably there is a layer that falls through where a human would eventually have to look at it?
Carter Maslan: Right. We do have people in our group that really serve to almost be the aids in training the algorithm decisions. Anytime you’re training an algorithm, you need to have a set that is the right answer. You at least need to have enough of the right answers to be able to say you changed the algorithm and it was very closely aligned with all the right answers that a human helped with. So yes, we do have that sampling to be able to improve over time.
Eric Enge: So you can manually build a good dataset, and then you can see how close the algorithm approximates it, because obviously you don’t want to build a dataset of 10,000,000 businesses by hand, but you could do 10,000 and see how that works.
Carter Maslan: Correct.
Eric Enge: And then this input can be fed into the manual review process because you’ll get some number of thousands of reports.
Carter Maslan: Right. The reports that we get are also part of helping with the training because every report that we resolve internally usually results in either some kind of tweak to the way that we are approaching that particular problem or it has an example of the fixed problem. It’s a continuing process. We solve problems, and they get amplified for everybody else that has a similar class of problems. You resolve one case where there was a particular problem where listings were combined because they shared the same phone number and address, for example.
If they were actually two different professionals, then the person that’s diagnosing and resolving that then determines to try something that looks at the parent organization or for other credentials in order to try and prevent that from happening in the future. Basically, there is an ongoing cycle of improvements that come from specific instances of isolated issues.
Eric Enge: This process should keep everything scalable.
Carter Maslan: It’s also the participation of the people that are in the local area. If you picture the literally millions and millions of global businesses, you really have to be somebody that is familiar with the area or familiar with the business to do a good job on that. A key part of it is the involvement of users not only in submitting accurate information but in telling us when it’s wrong, and their own interaction with it is also helpful. All of these things we hope will just result in a really good global comprehensive view of things in the real world.
Eric Enge: Let’s say you have a plumber who you are sure services ten different towns. Presumably, someone could type in that the plumber is located in Raleigh, North Carolina, when he actually lives in Durham. The person who typed in that listing declared that Raleigh is within the plumber’s service area and it’s probably a reasonable radius.
Carter Maslan: A good example of this is when you search, “Window Cleaning Le Roy, IL”. When you receive the result, click the Maps link up top, and then click on Cosmic Services & Window Cleaning, you can see an example, where this business just has its service area just as a radius. Then you can click on the show service area and it shows the designation of its service area, which is a simple radius in this case.
I think the visibility of what you’ve specified is also helpful because of some of the questions you asked about the way people will game at or try to game it. This is good because it’s kind of like transparency, and the user gets to see what the business has said about itself.
This is like your example from before. If a plumber said it served a radius of 1,000 miles, people can make their own judgment about how legitimate he really is. Maybe that works if he is in Northern Alaska, I don’t know, but definitely not in a dense urban area.
Eric Enge: What are some of the things that have been done in order to resolve some duplicate content problems?
Carter Maslan: This is a topic that I really want to help clarify, because there are two lines of complaints that people have. Sometime they’ll say that we incorrectly combined their listing with another, or that they can’t get rid of all these old stale listings that are duplicate references to their business. This is the whole balance that we are trying to find between duplicates and things that are known to be the same. The value of trying to get all of these references together is that it helps people know who is talking about them across the web.
It helps with our understanding of the terms that are used with business and with the scoring relevant matches because of all of those references and the way that people talk about their business. It’s hugely valuable to know all of those references and understand that they are all referencing the same thing.
One of the things that we’ve been continually working on is improving that algorithm, so we’ve had a variety of techniques to know that a listing was referring to the same place. We are expanding on that so it’s not just around geography or phone numbers, but looking at a whole bunch of other kinds of signals to help us know that they are the same business. That’s a big, big focus area, and I think it’s one of our top priorities. The work is ongoing and improving significantly.
Eric Enge: Generally speaking, what are some of the tweaks that you might use?
Carter Maslan: I already mentioned some of the things that we are working on in relation to the trustworthiness of various sources. It used to be that we didn’t have quite as much granularity in our understanding of trust. We would say internally how much confidence we had in the accuracy of something, but that was pretty much it.
Now, rather than making broad judgments, we are more granular in saying maybe it’s x confidence in this category and this geography. We are improving the granularity of our judgment of probable accuracy from a lot of different sources, and that helps a lot when you are trying to pick what the seed of a listing is.
Eric Enge: That seems like you see how many references there are or you examine all the various references for a particular business, and if they are all consistent then obviously you have a very high-level of trust.
Carter Maslan: If an outlier is subsequently confirmed to have been an exception, it may be because that person was fast in submitting an update for a new address or a new section of a business that is also helpful. Same with geography and category.
Eric Enge: Thanks a lot Carter!
Carter Maslan: Sure thing, thank you!