Ziya Genceren
Ziya Genceren is a Group Product Manager at Microsoft, responsible for running the product management team for local, mobile and mapping search technology. He has been at Microsoft for three years, initially joining the product management team for Microsoft communication services products like Windows Live Messenger, Spaces, and Hotmail. Genceren holds a degree from Carnegie Mellon University in MS Economics and MBA.
Interview Transcript
Eric Enge: Can you describe the types of different data sources that you use?
Ziya Genceren: It takes a lot of different data sources in order to bring the local search and mapping experience alive. It includes Yellow Pages data from local data aggregators, like InfoUSA. But also, traffic data from Traffic.com, and user-generated content. But also, traffic data from NAVTECH and others.
Imagery is another component, and many other sources like Urban Mapping for neighborhood data are also valuable. We also get content from people like Yelp, Citysearch, Restaurants.com, MenuPages, Dine.com, Frommers, Zagat, and Hotels.com, in order to deepen the service in terms of ratings and reviews.
It just makes it richer and more useful for the user. So, there are a lot of data sources that we are bringing together to make the service light up.
Eric Enge: Right. Is there a way that a business owner can submit their address location, address and phone number, that type of data directly to Microsoft?
Ziya Genceren: Yes, absolutely. There are a couple of ways of doing this. If somebody searches for the listings and doesn’t find it, you can just scroll down and there’s a link at the bottom of the page, the local listing center. We can handle one listing or multiple listings at a time; it’s pretty easy to do. You can have your listing inserted, or if you have your listing already listed but want to claim it or edit it, you can also do that.
In the details page, there is a link that puts you into the right experience in the local listing center, so you can edit your listing in the way that you want. The other side of this is a user coming in can also report an incorrect listing and tell us why it’s incorrect. So, the answer to your question is yes, absolutely. You can insert your listing and edit your listing; it’s free, it’s pretty fast, and it’s a great way to reach customers.
Eric Enge: And, if you do that, do you actually validate the content?
Ziya Genceren: Yes, there is a validation process of course.
Eric Enge: And, how does that work?
Ziya Genceren: Verification is done by phone or mail. In both cases, the business owner is given a PIN to confirm. If you are a new business, we only allow mail verification. If you are an already existing business, we allow either. However, if the business already exists and you never verified before and change the phone number, we will require mail verification before accepting your changes.
Eric Enge: Now, do you do the verification for all locations submitted by a feed, or only just for the individual ones?
Ziya Genceren: The feed ones that we get from our data vendors don’t go through a validation process.
Eric Enge: Right. But, what if there was a bulk upload to Microsoft as opposed to entering one location at a time?
Ziya Genceren: We don’t currently allow uploads of multiple listings. We can take a KML file (not feed) and have that applied over our map as a layer of content. But there is no ability yet to take a KML file or feed with information on one business or all locations in a chain and have it integrated within our local business corpus. We’re of course, always looking for ways to improve and this is one of the things we’re focusing on.
Eric Enge: Right, yes. I mean it’s an interesting problem, because it’s obviously very complicated, especially for businesses on a lot of locations, because locations tend to change.
What about the role of KML files, because I saw that you have support from KML?
Ziya Genceren: We do have support for KML files. We first started this support last year with the introduction of the ability to import KML files into collections to view KML hosted on the web. KML doesn’t really play a role in listings that we find. Let me elaborate on that a little bit. We were talking about local listings, finding businesses on the web. We don’t currently crawl the web, find KML listings and augment our existing listing data with KML files.
What we do is if you search for something and our current index doesn’t return a result, we let you search or look into the user-generated content which includes KML. So, that’s basically the situation with respect to local listings. We do have the ability to export collections that users generate into KML, GPX, and GeoRSS, which you can load directly into any portable navigation system.
You can set up an RSS feed for KML files hosted on the web, and once a KML file is added to your newsreader you will be alerted whenever the author updates the KML file, which is a great feature. Our user content is made up of our collections; KML files from the web, and geocoded Wikipedia articles, so it’s pretty rich user-generated content.
Eric Enge: So, it would seem then that the gold standard, if you will, or the highest authority would be those that were submitted directly to Microsoft and validated by Microsoft.
Ziya Genceren: Well, I think there are several dimensions to it; the recency of that data, for example, is also important. I mean, if you have some data that has been submitted yesterday and validated by Microsoft, of course, the probability of that being accurate is higher than something that has been updated last year.
But, something that was updated last year through a data vendor has a higher probability of being more accurate than a user-generated content, which was submitted and verified by Microsoft two years ago.
Eric Enge: Right, now that makes sense of course. So, just for theoretical purposes, if they were submitted on the same date and they differed, and you validated one, and the other was validated by a third party, you would know your internal data would be considered better, right?
Ziya Genceren: That’s right. You are right.
Eric Enge: Right. So, I mean this underscores exactly one of the tricky issues that you may be getting data from InfoUSA, locally from yellowpages.com, or wherever it may be. And, the data may differ in each of these sources.
Ziya Genceren: Yes. Typically, we have rules that provide the priority to content from certain sources. For example, let’s say we have a priority list of content; let’s say vendor A has top priority, vendor B has second priority. If we get different data from both, we know where to go and look at how the data is presented. It depends on the priority we give to the content source.
Eric Enge: So there is this temporal aspect which affects which source you give the most weight to, but your data may be a year old. If a dataset comes in that’s a month old, that just sort of complicates the algorithm a bit, but only because you have more information to work with.
Ziya Genceren: That’s correct.
Eric Enge: Right. This is a very confusing situation for many businesses. What would be your recommendation as to how they should deal with it?
Ziya Genceren: Sure. I would say ensuring that the information is accurate is the only way to deal with it. If you go through your listing and see that it is not accurate, of course, there needs to be a process to update it. And we also try to make it easy for people by refreshing our data frequently, and also merging and de-duping data from different data sources.
So, what you put on the web without any user action is our best understanding of reality. And, if there is some problem with the data or some gaps in the data, users, businesses have a way of correcting it. I think the best way to deal with it is to check if your listing is correct, if it is not, submit a correction to it.
Eric Enge: Right. And, of course, given what you said, it also makes sense to go out and make sure your data is correct at the other sources out there.
Ziya Genceren: That’s also correct, yes.
Eric Enge: So, it’s a little bit painful potentially for a business because they may have to go to half a dozen different places.
Ziya Genceren: Yes, that’s also correct, but it depends on the priority of this action for the business. If you rely on web search engines to bring you business, to be visible, to find your business, I would say it’s a critical part.
Therefore, if I was a small business owner for a large business who was dealing with multiple locations, I’d do everything in my part to make sure that the listings were accurate and people knew about it. You have to make it easy for people to inject this data, use this data, and work with this data.
Eric Enge: When I go into Live Search and search on “Seattle Pizza” I see 3 local results, but I don’t see a map. Is there a plan to integrate the map for the visual benefits that brings?
Ziya Genceren: Yes. We are always working on different ways to improve the results presented, and the map is definitely something we are considering doing.
Eric Enge: Can you talk about the key things that go into deciding which pizza place in my example here shows up first?
Ziya Genceren: Yes. There are a hundred plus different features that determine where the business ranks. Words and phrase matching on the business names is, of course, one of them. Category membership based on the metadata that we have on the business generated is another one. Our data vendors provide some specific features that specific business might have, we take those into consideration. Data vendor provided descriptive text goes into it, matches against unique identifiers such as phone numbers, domain names, and URLs.
Of course, distance from the query like location is something that goes into it as well. Currently, we don’t take into account business popularity. We have very deep data from people like Yelp, Zagat, from our own user-generated content. We don’t currently use that as a factor to rank category based results, but that’s something that we are definitely working on.
Eric Enge: Right. So, you did mention the geographic component, and so people refer to that in the industry as proximity to the centroid. Is there, in fact, an advantage to being right at or very near to the centroid as opposed to say several miles away in the same city?
Ziya Genceren: Distance from query center is definitely something that we take into account, but that’s not the only thing.
Eric Enge: Right. And, you also mentioned the category. Do you go through some process to try to validate the category? In other words, you get user submitted data and it’s miscategorized.
Ziya Genceren: We limit the number of categories per listing to six. However, there is still an opportunity to SPAM, which we are aware of. We do have a category scrubber tool that we eliminate potentially bad categories based on review of the listing data. However, this is still very imperfect and we will need to do some data mining and data quality work to identify and clean up problem listings. So far, it has not been a pervasive issue.
Eric Enge: Right. What about the role of reviews?
Ziya Genceren: Currently, we don’t take into account how popular the business is or how highly reviewed it is in determining how we rank different businesses in our results. But, that’s something that we are working on.
Eric Enge: Do you make use of web references? This is as opposed to links where someone might mention somebody’s business name without actually making a hyperlink out of it?
Ziya Genceren: In determining the rank order of the business, we don’t look at web references, but we do provide web references within the results. I understand that’s a bit different from what you were saying.
Eric Enge: Can you talk a little bit about 3D Cities?
Ziya Genceren: Yes. So, the 3D effort provides accurate, high-resolution, detailed city models; realistic models to provide an immersive search experience. Actually, users will see that we have more immersive experience in 3D now as compared to before where there would be more detail at the high-resolution level, terrain features, 3D threes even in our virtual environment.
250 cities are currently available and continuously being enhanced, and more cities are being added. 3D has been very well received in the market place. It’s a very promising area, a big area of investment for us, and we continue to deepen that.
Eric Enge: Yes, indeed. It must be a very significant investment of resources.
Ziya Genceren: Yes. The way we do this is through high power software; of course, it is a scalable way of doing it., but it is a high area of investment.
Eric Enge: Yes, absolutely. And, you’ve got it set up so it’s pretty seamless to move into a 3D mode from just a regular map mode, right?
Ziya Genceren: Yes.
Eric Enge: So, that is part of what makes it an engaging experience; you can kind of flick back and forth between the two modes.
Ziya Genceren: That’s correct.
Eric Enge: Great. Can you talk about 1-Click Directions?
Ziya Genceren: Yes. We are always looking for ways to improve the user experience. And, 1-Click Directions is a fantastic experience, because you know how to get from your home to the major artery wherever you live. For me its I-5; I know how to get to I-5. Tell me how I get to the place I want to get to after I-5. I don’t need to know that I need to turn left on a one-way street after I get out of my driveway.
So, 1-Click Directions really makes it easy for you to get simple directions to the point that you want to get to from the common directions that you might approach it from. It essentially takes the step of having to input two addresses and shrinks it to one address. It’s pretty innovative, and it has been very well received in the market place.
Eric Enge: Right. Also, when you get your directions at the end of it all, you don’t have a bunch of steps you don’t need, such as how to get to I-5, making it harder for you to what you need to do once you get to I-5.
Ziya Genceren: Absolutely, yes.
Eric Enge: Right? So, I think that’s a pretty nifty feature. You have done another thing related to traffic which is the Clearflow feature which is oriented around getting around congestion that you help identify for them.
Ziya Genceren: That’s right. So, traffic is, of course, a key feature in mapping; it’s very dynamic. And, our research folks came up with this fantastic feature called Clearflow that expands the coverage that we have for traffic data.
It basically hides all the complexities from the user and gives it to them in their familiar interface that they were already using, while simply extending the coverage of the feature. Now, they get to see not only the major arteries but also what the other streets are looking like. So, it’s a fantastic feature and is really very useful..
Eric Enge: Where is that available?
Ziya Genceren: That’s available every place we have traffic data. It relies on traffic data and it is available in more than 77 cities.
Eric Enge: Can you also talk about the integration of all this into the mobile environment.
Ziya Genceren: Yes. It is a big area, but it’s a very promising area as well. I believe we have a great client that serves the user needs really well. I believe we have a very good mobile experience to present to our users that offers live traffic information. It offers directions, it offers a great local search and it offers a great web search. For Windows mobile users for example, if they download our client at wls.live.com, they can get our mobile client which features voice input as well as great maps and local search.
They get access to a really easy way of finding cheap gas. We get data on the current prices for different stations around your area. We give you a chance to look at what the prices are, so you can pick the one that fits your budget.
Eric Enge: Yes, that’s very cool. So, you can decide whether you want the closest gas station or the cheapest for example.
Ziya Genceren: Yes, that’s right.
Eric Enge: Can you do the 3D Cities on mobile?
Ziya Genceren: Mobile does not have 3D Cities currently, but what you can get is roadmaps and high-resolution satellite imagery on it.
Eric Enge: How’s the adoption been of the mobile product?
Ziya Genceren: We have been very happy with it. Users are engaged and they are using it. Windows mobile users have accepted it, so I have been very happy with the way both user adoption and user engagement is going with the product.
Eric Enge: Can you provide a high-level summary of your mobile product?
Ziya Genceren: I believe we have a great user experience that addresses the user needs on maps.live.com where people can find local businesses, look at ratings and reviews and make the smart choice for them. They can look at user-generated content and contribute their own content at maps.live.com, so I believe that’s a great thing.
The innovative, immersive 3D experience and high-resolution Bird’s Eye Imagery gives users more context and a great search experience on the PC side. On the mobile side, I believe our local search client and browser experience offer people a real solution to find the businesses that they are looking for or find the content that they are looking for. It features voice input, therefore even when you are in the car (although we don’t advise using the mobile phone while you are driving) you can easily input your query by speaking, which is a very natural and easy way of doing it.
We also provide another solution which is 1-800 call 411, which is a fee-free 411 service. If you don’t have a data plan, and if you are on the go and you need to find the address or phone number of a local business, you can just call 1-800 call 411 and speak your query, and we will provide you your results and even connect you with the business so you can speak to them. So, these services cover the need for people to find the local business, look at ratings and reviews, and be able to do this on the PC or on the go.
Eric Enge: Right. Now, do you have anything linking the directions to GPS?
Ziya Genceren: Actually, the client works with GPS enabled phones, so if you have a GPS enabled phone, you are able to use our client and the location features of your phone.
Eric Enge: Right. So, you can actually track your progress along your route.
Ziya Genceren: That’s right.
Eric Enge: And then, will it potentially speak the directions to you?
Ziya Genceren: Well, that’s a feature that we currently do not have available, but that’s a very interesting one of course. It’s a very high-value feature for the users.
Eric Enge: Yes, I know, absolutely. It all sounds very cool; there are a lot of good and unique things that you have done with the product.
Ziya Genceren: Thank you very much, I really appreciate it.
Eric Enge: Thank you Ziya!
