Remarks by Yusuf Mehdi, Senior Vice President, Online Audience Business
New York City, New York
March 25, 2010
MIKE GREHAN: I'm very pleased, as you're obviously aware, to be joined by Yusuf Mehdi from Microsoft.
So, Yusuf, welcome. Thank you very much.
YUSUF MEHDI: Thank you. (Applause.)
MIKE GREHAN: I could probably wax lyrical for about an hour on the background that you've had with Microsoft, and probably some of it would come out like a shopping list. Obviously, I'm aware of some of the fine work that you've been doing there, but maybe it might be a good idea just to give everybody an indication of this illustrious career, in your own words, and then we can get around to what's happening with Bing.
YUSUF MEHDI: How far back should I go? All the way back to childhood, about five years old?
MIKE GREHAN: I want to take you way back to your childhood now, and do a little bit of therapy.
YUSUF MEHDI: Well, good morning everybody. It's a real honor and pleasure to be here with you all to talk about this great industry of search. I've been at the company for about 19 years at Microsoft. So, when it was about 3,000-4,000 people, and I got there through a roundabout way. I was working for Reuters, the trading terminals that they do. And they had a real-time data feed into an Excel spreadsheet on the Windows 3.0, a beta of Windows 3.0. And all of the delight that you could use the power of computing, and the live data stream to manage your portfolio was just a fascinating thing to me.
And I was right out of college, and I was at Reuters, and they said, hey, you're from college, you must know about these things called computers, figure out what this thing is supposed to do. And so I did that for a couple of years and enjoyed it. And I said, I've got to go work at this company that's done all this great software.
And at Microsoft I've been fortuitous to be in a lot of the great places, I mean, there are so many great pieces of work going on. But I was there when Windows started, so DOS was king, and OS2 was the future, and Windows was kind of a toy application, and it becoming the platform for the company. And I was fortunate to be there for versions 3.1 through Windows 95. And then the next day, I went to go work on Internet Explorer.
Again, I was staying close with the developers. And I remember a funny story, I'm in the office with one of our star developers, and when you're in product management the key thing is, you've got to be tight with the developers. And so he said, Yusuf, come in, you've got to see this thing. And so you go in, and it's the classic developers office, all dark, all these monitors up, he goes, watch this.
He types in ping, some numbers that you all recall, we sit there, and then it comes back, one thing, one thing, one thing, one thing. And I said, what happened? He said, I just sent this data packet from here to Zurich to Italy to Brazil, and back to Seattle in less than a second. I said, that's amazing, and what's it mean? He said, I have no idea. But, he said, it's going to be hot.
MIKE GREHAN: It's a pretty cool trick.
YUSUF MEHDI: I'm like, OK, I'm going to go work on that. So, I worked on Internet Explorer Versions 1 through 5, and then came to the Online Group, came to MSN. I worked for a while in MSN. I actually ran engineering for a period of time when we started our search service. So, I was there when we started search. And that was also a fun story.
We were in the office, and while we were debating entering search as a company, because it's now a multi-billion dollar investment, we asked one of the developers, hey, see if you can write a crawler, something that can go crawl the Web. Let's see if we can even do it. And so one of the guys came over, and he wrote this crawler, and he came back and said, OK, I crawled 48 documents and then the server kicked over. I said, all right, that's a start, you know. And we started that with literally the server and the plug that some one tripped over 48 documents, and now we're off to many more.
Then most recently on Bing, on Bing and MSN and it's been a fantastic opportunity to be a part of the momentum, the early start we have with Bing. Then you blink your eyes and 19 years goes by.
MIKE GREHAN: For a lot of people 19 years is a prison sentence. So, it's been interesting for you.
YUSUF MEHDI: It's been great.
MIKE GREHAN: But, it is interesting, you mentioned just about Internet Explorer, Microsoft was pretty late coming into the Internet anyway, it was a 30 so you guys probably were a bit rushed to get something out there, and the same thing with search. I mean, it seems to me that folks' attention was elsewhere, and now you seem to be speeding ahead, but were you kind of under pressure with Internet Explorer to get out there and make it known and then the search thing?
YUSUF MEHDI: Yes, absolutely. I mean, I think that in both cases I think we had a view, but the market moves around you quickly. And you want to be responsive to the market and you also have to be on your strategy. And I think in search we definitely had missed the boat early on that the Web was about the long tail. We had actually focused a lot on the head of the queries and said, hey, the long tail maybe is not that important, and it turned out that it was much more important. Although things change and so one of the things we'll even talk about is that actually the head of the Web is in many respects, again, more popular.
MIKE GREHAN: Yes, so yes, we're going to get a demonstration later on about the latest and greatest technology that is happening over at Bing, and I would love to describe more of that. But, I was told yesterday if they explained to me what they were going to do today, they would have to shoot me immediately afterwards.
YUSUF MEHDI: We'd never do that to you.
MIKE GREHAN: So, let's just talk about the changes in search. I mean, you guys came in at a later stage, but exactly the same kind of thing. You know, we've go this protocol HTTP HTML documents, you go out and you crawl the Web and you come back with your 48 documents and think, try harder kind of thing, so you have to build a crawler. It is a very, very difficult task, isn't it, crawling the Web and trying to pull all of this data together, and try and make some sense of it, because it's completely unstructured data, yes?
YUSUF MEHDI: Yes. It's a super-hard problem. I mean, for a period of time I know a year ago every I think it was every second of every day in that year there were four new URLs created. So, you think about the explosion of the tail, and the fact that you have to crawl, index, and make sense of that. And that's just even on the old metaphor of search and then the other interesting thing is that on any given month about a third of our queries that show up is the first time we've ever seen that query. And then a huge chunk of those we'll never see again, they're gone into the wind, poof.
And so, the challenge of basically being able to be up to speed to understand that in-flow of data, and then to be able to index the right thing so that you can respond in sub-second time, the right answer, very, very hard problem. And the problem then gets harder as you move into like these new forms you just mentioned, Mike, unstructured data. So now you're saying, well, how do I get to things that are closed off, how do I get to Twitter feeds, or YouTube videos, or things that you can't actually crawl easily. And all of these are coming in new formats, bit URLs, for example, you can’t really crack those. So the whole page link architecture that everyone in this room understands quite well, what happens when you can't understand the URL to improve your ranking?
MIKE GREHAN: That's really interesting, because obviously we have a very hard core of people who come to the conference who are mainly interested in the organic side of the business, the SEO side of the business. And fundamentally we still talk about the on-page factors that are important and we'll move onto linkage in a moment. But, when you bring al of this data together and you're looking at documents, obviously way back in the day the only documents we've had to look at were basic HTML and it was all about text. There would be the odd graphic there. Things have changed now, particularly with the introduction of these blended results, where we have videos and we have PDF documents, and we have images, all of those kind of things.
In terms of the job that these guys have to do, when you're looking at an HTML page and we'll look at other file types in a moment, how important is that on-page stuff, I mean, making sure that you've got the title tags, and you've got the keywords in the right place, and the document, and are H1 tags important? I mean, I'm always curious to know in terms of the way that we have to do this job, which way do you do it? Do you start at that level, or is it the top down thing, should you be looking at links first of all? I mean, for Microsoft, when you take a look at an HTML document, what's important about the page?
YUSUF MEHDI: I mean, there isn't anything I would say I'll answer in a couple of ways. One is, there isn't anything out of the ordinary that people aren't usually doing today that you would expect. I mean, we have a whole site up, Web Master Center on our Bing site, that explains it. But, it is the classic things, in terms of making your HTML pages well designed and well formed, so that they're easily crawlable and the relevancy is on there. That's the case. But, I want to go to a different thing that you brought up, the question you had, which is how are things changed. And this is a big part I think for here as you plan the future, the thing that's the biggest change for the consumer is that search, which was invented however many years ago, 12, 20, whatever you want to count, was really initially very good at navigation.
So, the whole point was, I need to go find a Web site, what's the UPS Web site, what's the United Airlines Web site, and then PageRank, which was a great genius invention. It was all about let me understand the anchor text of the URLs, and the linking, because that is how that's probably the best way to find relevancy. And that was the case for navigation, and today it's still the case for navigation. What's changed now is that consumers are using search in a different way, in particular they're using search now to conduct transactions, they're using search for advanced research. And they're trying to do things with search that search was not designed to do.
And what happens then is the way I'll give an example. If the old query was, what's the American Airlines Web site, the new query is, where does President Obama stand on healthcare, what are the five causes of a certain type of pancreatic cancer, and trying to think about how you formulate that as a query, very hard, how you answer that in terms of search results also changes. And what happens then is we've gone on to say, let's design the next generation of search that can handle those types of tasks, where the answer is less fuzzy and in a way it's more of a dialogue with the consumer. So, it isn't give me 2.8 keywords, and that will return back for you the 10 best links on the Web. It's more tell me what you're looking for and then in a way I'll help guide you back to what you want. So, when you think about then, for folks in this room, how do you take advantage of that, the thing you have to have in your mind now is consumers are they are looking for something other than links. They're looking for complete answers, and they're looking for in a way almost decision support tools, because they're coming to search engines to do that. It's a mind-boggling thing.
So, in many ways when I think about the SEO community, I think very much of it as in a way developers for search. And we'll talk a bit more about that, but I think of them as developers, where you're actually helping publishers and advertisers design a different way to help the user accomplish these tasks.
MIKE GREHAN: So, I kind of think in a session that I was doing yesterday just talking about information retrieval on the Web, and it was fairly kind of layman's kind of look at the way that the science develops, it wasn't the science lesson, or anything like that, but just how we can relate that back to search engine marketing. And we went through all of the on-page factors, why that was important, and then it was very difficult, because you would have this abundance problem in trying to rank documents. Then you mentioned page rank there, obviously. That was the main focus of attention. For the past 10 years people have been trying to figure out all of these things about page rank and how important it is.
Linkage data is very, very important, and I guess we still need to look at that, but if you think about it, if all you ever do is use links to rank a document, then you're excluding like millions of end users who can't really vote for what they think should be number one, because they don't have a Web page to do it. But, I guess, now when you start to look at tracking the end user, and looking at these kind of query chains that we've talked about and user trails, and seeing where they there's a lot more that you can learn about the end user and what they prefer, beyond just looking at mapping keywords and looking at linkage data.
So, I think probably the end user is taking the lead now, and the more that you understand about the end user the easier it is for you to provide them with to connect them with the right kind of content, yes?
YUSUF MEHDI: Yes, I think that's right. We've done part of what I think has been the success of Bing is, we've done a huge amount of data mining, and data understanding. And you talked about search sessions and chains, and that has been a big breakthrough for us. We watch on an anonymous basis, and you can see how someone will actually conduct a task. So, we used to think that, OK, when someone sits down and types in those 2.8 keywords, that the result that they get back, that's going to be what they need at that moment. But it turns out that for things like commerce, it's mind-boggling. Like when someone types in, let's say, a travel query. What percent of the time in what time do you think someone will make a purchase? Any guesses?
MIKE GREHAN: Anybody want to guess?
YUSUF MEHDI: More than 60 percent of the time, the actual purchase on that query, the initial query, won't happen for two to four weeks. So, think about, at the time, whether you're buying a keyword ad, or you're trying to provide a response back, in that first query, they don't do that. In fact, as we started to watch, we see how people kind of navigate slowly to say, well, first, I'll do some research, and I'll do the following, and then, of course, you know, oh, they have to run to dinner, or they have to go do some piece of work. Then they come back later that night. Then they compare it with their spouse or their friends. Then they come back. And we've done this for so many things, in particular people, places and things. And you see how the chain and the usage of things change. And that has been eye-opening for us.
Again, even, let's take for example on a car. You do a query for a car. Usually what happens is, again, today, page rank, let's get you the link. But if you think about it here, if you and I are having a conversation and I say, I need a new car. How would you respond back? You don't say, oh, here's this Web site. You'd say, hey, what kind of car you looking for Yusuf? A new car? What do you want? Do you want a sports car? What's your budget? This is my budget.
OK, well, then you can come back, now you can come back and say, hey, here's a couple of cars in your range that have these trim packages, and et cetera. And that's the dialogue model I'm talking about. That's how people are using search engines. And so the way to help them is not necessary to say, OK, OK, let me get that link up. It's in a way how do we come back and help you with that task.
MIKE GREHAN: With that task, yes.
YUSUF MEHDI: And that task is unique and different for all sorts of queries.
MIKE GREHAN: And I kind of use this analogy that the more that you know about the end user, like we've been looking at search engines for such a long time, and you look at the search engine, and you think it's a black box that you invented, and there's voodoo and nobody knows what goes on inside it without realizing that the search engine is looking back at you, and you're a black box, and they have no idea what goes on inside of you. So, the bigger that information exchange is, then the more relevant the result is. And the easier the task becomes, yes?
YUSUF MEHDI: That's right on the money. That is the way I think about our vision. So, if you compare and contrast our vision with other players, everyone has like a unique approach, and no approach is necessarily right or wrong, they're just different. One of the other visions is digitize the world's information, and make it useful. That's a noble goal, that's a decade hard problem.
MIKE GREHAN: It's not going to happen.
YUSUF MEHDI: Well, but it's a great goal. The way I contrast what we're trying to do, and how I think search will be different from our point of view is, we are about just what you said, understanding user intent, and then mapping that intent into task and into action. And that's also a very hard problem. In some ways, I think it's almost a harder problem understanding user intent. It's almost like, can you build the ultimate mind reader so that when someone pulls out their phone, or sits at their computer, before they start typing, we have a decent sense of, OK, well, hey, where is this person, lat/long? What time of day? How fast are they moving? Is it 60 miles an hour or two miles an hour? Are they stationary? What's on their upcoming appointments? So that when they start to type, we say, OK, now we understand most likely the intent in that moment.
MIKE GREHAN: And basically, I think, what's been happening with search, the way that users have approached it up until now, and I think the end user is more sophisticated now, it's kind of been getting an answer to a short-term information need. You put your 2.8 keywords in, there some links, and you have a look, and then you come back and you do it again, and then you come back and you do it again. And I sometimes wonder with an information provider, because I think, you know, a search engine is one thing, but an information provider is going to be multi-modal sources anyway. If you know what it is that I like, you know what I want, why do I have to keep asking? Why don't you just give me it, you know? So, satisfying a long-term information need, I think, is probably something that we'll see in the future.
YUSUF MEHDI: Yes.
MIKE GREHAN: I think right now before we go into our demo one of the things that we need to do from a very practical point of view is that SEO is still around. There are all of those best practices, just for a couple of minutes let's talk about what are the best things that we need to do to be able to get found at Bing? The keyword research obviously is extremely important, understanding the user intent, on page, linkage. Just for a minute or so, what do you think is best practice to be found at Bing?
YUSUF MEHDI: You talked about a bunch of those. I think making sure again, on our site we have a place where you make sure your URL is crawled properly. I mean, the way we we don't get into the details of how we do a bunch of the ranking. But, again, making sure that the proper keywords are there, we're very focused on matching the relevancy of the keywords that we read off your page with the page itself and the landing that the descriptor of the anchor text, et cetera, is accurate, because that's a lot of how we think about, OK, is this page actually relevant t the content that's on there. Those are some of the primary things that we'll do.
MIKE GREHAN: So, link anchor text has been the workhorse really of search for a very long time. I know you alluded to that before when we talk about PageRank, I think it probably applies more to link anchor text than it does to a specific page. So, I think that's a very important thing to take away that at the end of the day having a great title tag is very, very cool and obviously it says a lot about the page, but the more that you can have in the link anchor text that describes the target page, and the closer that is to the query, I guess.
YUSUF MEHDI: And then some of the things I can show you in some of the product demos is where things will go, and how I think start to think about, OK, in addition to the linking and the keywords, and the HTML on the page, what can you do with applications, what can you do with richer answers on the page, and that's where I think in the big part of as you think about the SEO industry, I think that's where we'll head. It's baby steps and we have a ways to go, but we can kind of maybe show a little bit of where things will head.
MIKE GREHAN: That's probably a good idea, actually. Why don't we just do that, who don't we just take us into the future and let us know what's happening over at Bing.
YUSUF MEHDI: All right. So, what I want to do is take about 10-12 minutes and show you a few things that are coming, so you get a view of some of the things that Mike and I were just talking about, how will search change.
So, I'll start here on our home page and you guys know a little bit about the page if you've used Bing, and you can see that we start out with just a new visual approach, a new kind of graphical approach that lets you discover information. And this has been something that has been very hot with kids in schools. In fact, in the elementary schools this is the dominant homepage, people switch over just because of the beauty of discovery.
I want to start with, first, with a query, and a query on, for example, Taylor Swift, a person. So, I want to talk about people, places, and things. Now, one of the things you'll see is in the past you'd come up and you'd get, for example, some links, and then little by little people sprinkle in some content. One of the things you'll see is in the past you'd come up and you'd get, for example, some links, and then little by little people sprinkle in some content. One of the things that we're doing as were adding a bunch of information here in this box that effectively becomes the answer and we'll start by, for example, calling out this is the official site, and we link to it. We'll have images. But, one of the things that we start to do is we start to understand a little bit about, this is not just a page with keywords, this is an entity, this is a person.
So, here on the left rail at the top I want to call your attention to how we have made, yet again, another big bet with the user experience in search, taking what is probably one of the most valuable real estate, the left pane and we start to basically bring topics to bear. For example, Taylor Swift we know is an artist, and so we find songs, lyrics, her ring tones. Of course, we can find related searches, because we start to draw inferences about what are other musical artists? What are the things that are related here? And you can come and find that. And then along the top, this is a new thing that we're just introducing and this will also start to show up in our user experience in flights, 1 to 5 percent flights starting today, if you get lucky you'll see the flight, because we have the ability now to come in and then look at different topics on that.
So, for example, Taylor Swift people do a lot of searches for images, likewise for news, or even for events. So, I can come in so, for example in this case we know she is a performer, and so here are the list of events that are coming up. And we can pull that out and I can go ahead and click on events and I can then refine my query. So, for example, I want to see where she's playing this weekend. OK. She's playing in Auburn Hills. I can pick that particular one, and I can come in and actually buy tickets. And so what you can see is now a very easy way for the consumer to say, I want to discover that artist, let me go find that.
Another example is, for example, let's say you want to do travel. So, I'm looking to go its spring break. Let's go to Miami Beach, and I'll type in Miami Beach and, again, pull up a whole entity card and we know this is an entity, and on the left we know, OK, related to this entity is probably hotels, tourism. We know that there are different parts of Miami that are probably hot like South Beach. We know about spas, resorts, et cetera. And so we're pulling this intelligence out. You can come in and look at things like the weather, or flight details, or we even aggregate a bunch of information.
So, for example, you can come in and slideshow and in one click you have a full screen experience. So, you can actually discover and maybe take a look at the place that you're thinking of traveling very, very easy. And, again, this really resonates with consumes. They really are they get tired of the links and the text and they want to come in and see the images, see the visuals as a way to discover information.
So, I'll come back out here and what we'll do is, I come back and I can even now they'll say, OK, I need to book a flight. So, we talked about, do they need to go right to the particular airline site, or do they need some help in finding that? We've now introduced a number of tools.
So, one of the great tools we have here with Bing Travel is the ability to see the prices, and we use algorithms to define whether the prices are going to go up or down. So, in this case, going from San Francisco to Miami, it says prices are going to rise in the future. So, you might want to buy your ticket now. If you haven't you should try it. You'll save a decent amount of money with the predictions. And we give you the confidence rate of how likely we are to be accurate. And then once you decide you can come into the calendar, see what flights are available, and then we'll take you all the way through this decision support system to actually help answer it.
So, for example, here are all the prices, from lowest to highest. Then you can come in and refine your query nonstop for a particular airline. So, again, what happens is we're taking this in a direction where it is helping you get the task done, not just giving you the links.
Now, finally I want to show you something here. Here is the New York Times. We understand this concept of the times that it is, in fact, an entity. And what we'll do is, in addition to the official site and deep links, we have the customer service number, one of the hardest problems to find is the customer service number for any company. We know how to go and find that deep on their page, and I challenge anyone to find a customer service number for a classic retailer. It's very hard. We'll pull it right to the top of the page. And one of the things that we've just introduced now with the real-time information is we can pull the most popular shared links on that topic.
So, from the New York Times we can see what links are most shared about the Times, and by clicking on it. These end up being the stories. So, now you get great insight. Not only do you see the news that's going on, you can actually come in and see the links that are being shared and scroll through here. So, it's a very powerful feature, and just in general we have the full fire hose of Twitter, and you can see what topics are trending, et cetera.
So, that gives you a little bit of a sense of that. I want to finish that section by talking to you a little bit about the car example we talked about. So, let's say that I want to find a new car. There's now a whole different way to start your search. Here's a visual way. These are all the cars that are new and I can just scroll through and find the one I want, because the studies show that the human eye can detect at 40 times better something that they're looking for versus looking at it for text.
So, I can scroll that way, or I can say, for example, all right. I think Michael, he's telling me he's looking maybe for a sports car. So, I can say show me all the sports cars. And then let's figure out your price tag. Let's keep it down to $25,000 and then up come the cars that fit that price range, and I could say, OK, Ford Mustang. And I come in and I can see the car. We'll have specs, safety, et cetera on that car. And if I want to, I can come in and see all of the information about that, and I can read all the reviews. So, an incredibly rich way to discover and explore information.
The second thing I wanted to show you was how that will go forward into the mobile area. So, what I have here is a very hot and rare device. Don't take this when we're done with the demo. This is one of the new Windows Phone 7 series, incredibly hard to find. I had to arm wrestle to get one. It's very much in prototype mode. And what I want to do is I want to show you a little bit about how search will change on the mobile device.
One of the things that's fantastic about these Windows Mobile phones is that they are context sensitive and aware. So, as we talked about, they know your location, they know how fast you're moving if you're in a car. So, if you're looking for directions and you're moving 60 miles an hour we'll give you driving directions. If you're moving 2 miles an hour we'll give you walking directions. If you're stationary maybe we'll send them to you.
So, here is the user experience. Go ahead and pull that up. So, you get this nice home screen, and this again will be a lot like you see on the Bing home page. What I'm able to do is I'm able to scroll up and down, a very nice touch screen. And there's a button on here, the Bing button. I'll press the Bing button. And up will come our homepage, like you've seen before with the hot links. And I can come in and actually just do a query. So, for example, this film I just took from Vegas. So, we'll assume we're here in Vegas and I'll type in sushi, and automatically up will come all of the latest restaurants that are right near me, and we can pick, for example, the Buddha Cafe. And what you'll see is we have the ratings for it. We have the directions. We even have useful information. So, for example, on ours it says no lunch. And I can actually read all the reviews. We'll go out to the Web, extract all the reviews. And if afterwards I want to do something I can come in and click nearby, and see what other events are nearby and do that on the map. So, it gives you some sense of how that will change on the phone.
The final thing I wanted to the final section I wanted to show you was around the area of mapping, and how things are changing in the geo-spatial, and this is in particular a big area in terms of how search is going forward. Local and commerce, in particular, will be a very big thing, I think, as we see search evolving. And I want to show you a couple of thoughts here. First off, for those of you who haven't seen it, I just want to show you Bing Maps.
We've done a huge amount of work to get a very fast, and highly performing map application that lets me just let the scroll button scroll all the way down from up in the air satellite to get all the way down into the city, and you could see how performant this is. I can scroll around, this is New York City here. And you can see how easy, you can just zoom in and out to any place and go all the way back up. See the map, then get down into New York, you can see how fast the performance is.
And one of the things you can do is, I can come in, for example, and say, let's go to Hilton, New York City, and it will automatically find you, and zoom you right down to the map. That's in New York, and there are a number of them. And we'll say this is the right one here. And it finds you here. And I can say with one click, I can say, tell me what's nearby? We will automatically find for you, what are the hotels, what are the restaurants, what are the bars nearby, the museums.
So, again, as a consumer, finding information now becomes much more powerful, because you don't have to go and tell it all these things. You just say, hey, I'm at the Hilton, New York, what should I do? And Bing will automatically bring up a whole number of things for you in this area, and then you can explore. So, again, it's more of a dialogue model, the consumer is dialoguing with the system, if you will, to find information.
Another example is a great little application called the Newseum, which basically shows you the front page of major newspapers anywhere around the globe. So, for example, here we're up in Seattle, so we have the Seattle Times, the Pondexter family's MVP. I can come down the page there. I can go down to California, and see what's happening at the San Jose Mercury News, the San Francisco Chronicle, there it is. And then, I can scroll back, if I want, and I can go across the pond, as they say, and I can see the Guardian in the UK, and I can click on that and I can see, for example, or the Journal, what are the front pages of those papers. So, for example, Irish Examiner, but they'll come up, and you'll actually see the front page of that page as well.
And then one of the things I want to show you is that we have a number of new applications as well that basically are new applications. And so we talked about how the future of search will change, and how you can write applications to help get answers. And some of those were those instant answers you saw on search. Some of them could be map applications. And so one of the ones, for example, I want to show you is Twitter Maps. And what we've done is, we take the live feed of Twitter, and we can put it right on the page here, and take a look.
So, for example, let me do a couple of things. Let me filter, so we'll see tweets just that have photos in them. So, we'll show these. And as things start to tweet in real-time, what we'll do is, we bring together the power of real-time tweeting right onto the page. And so you can see here, I can scroll back over the period of time, and see previous tweets with photos. I was looking at this before we started, and here, for example, is a picture of the Hilton taken about 13 hours ago. So, this one was someone probably coming in a little late, late at night.
MIKE GREHAN: Wearing a hat, apparently, yes.
YUSUF MEHDI: You can see what's happening there. So, you get a sense, again of how that real-time information comes in. And these are apps you can do today.
Now, I want to show you something that is prototype, and coming soon, that's not yet here. And this is how we take now the power of both all of our imagery, plus the ability of consumers to interact, as we were talking with the engine, and then the live feeds.
And one example, so here is Pike Place Market. This is taken now at street-level view. We basically drive cars around the globe, and we take photos, you know, two feet off the ground on these cars with cameras, and you can scroll around, and see the actual location that you're at. So, if you're looking for a business, you can come in and look at that.
One of the things that is pretty interesting is, you can come down, and we can say, for example, show me pictures. When people take user-generated photos, we can take those photos and we can overlay them on this map. So, for example, let me find one here, this is now what I'll do is I'll overlay this is a photo someone took, and we overlaid it on the map. We understand the 3D world, and it's an accurate representation, and then we geo-code it, so that when you take a photo, we know where to put it.
One of the things we can do, for example, is, I can take an old photo. So, for example, let's say I want to understand what Pike Place Market looked like many years ago. So, here's a picture of it back in 1919, and it's overlaid on the exact location. So, for example, if you said, what did this look like back then, if I was standing here about almost 100 years ago, what would that look like? I can do that picture. And I can say, for example, what if it's nighttime, what does that thing look like at nighttime? I can click on that. And you can say, this is the look of it in the evening.
So, you start to get the power of that, of people overlaying, and then being able to go in there.
Now, what I'll show you is. Let's say I can actually drive down, as if I'm actually headed down the street. And so, as I'm looking for a business, I can look around. Let's say that you wanted to come see this notion of people, and I don't know if you guys have been to Seattle and the Pike Market, but what they do is, they throw these fish around, and you can actually see them throwing the fish. Let's say we want to see that. Now, imagine that thing I showed you with Twitter, where you have a real-time data stream, let's take a real-time data stream that is a video feed, overlay that, geo-code it in exactly the right place, which is one of the things that we've done. So, here is a video, live, or it was live, we shot it live, of in the market of a guy throwing fish in that exact area. And what you can imagine is now very quickly in the future, you'll be able to say, oh, before I go out to get my coffee, let me just zoom into the maps, and see how long the line is at Starbucks. Let me see, in fact, if that bus is coming down the street. So, before I run out my door, I know how fast I have to run. And here you have amazing power of geo-spatial data, we've rendered a 3D accurate picture, and then you can overlay. And as other people overlay videos, or content, you can get this richness here. So, super, super powerful.
And the last thing I want to close on is, I want to show you a new application that is going to come soon. In fact, we're going to roll it out in the next couple of days, and it's with another very hot effort that is emerging on the Web through a little partnership we have with a company called Foursquare.
And to do that, what I want to do is, I ran into him here, he is the CEO of Foursquare, Dennis Crowley, I want to invite him up. We were having coffee yesterday, and I said, hey, why don't you join me onstage, and let's share the app that we're going to collectively beta in the next couple of days. And so, he said, sure, why not, we'll come up and talk a little bit about Foursquare. So, please welcome Dennis.
And so, Dennis, let's just show them a little bit about the Foursquare app, and maybe you can talk a little bit about what's going on.
DENNIS CROWLEY: Yes, sure thing.
YUSUF MEHDI: I'll pull it up here, and what you see is a little bit like we showed before with Twitter, I can come in now and I can zoom down into New York City. And what will happen is, once we get down there, we can start to see the activity that people have. And people post information, they check into places. And what you'll see is, I'll show you some of them here, they start to show up on the site, and you can see them scrolling in real-time.
And, Dennis, why don't you tell them a little bit about what Foursquare is, and what people are doing here live on the site?
DENNIS CROWLEY: I'm assuming we have a bunch of you who have used Foursquare before? I saw about two dozen people are checked in at the conference, and specifically in this room, which is kind of great. Foursquare lets you check into places, and share the places that you go with friends. And we also have a layer of game mechanics on things that reward people for doing interesting things. So, you can earn badges for going to multiple karaoke places, or exploring different barbeque joints. Then you can leave different tips all around the city, all around the world for your friends to discover. So, it's not uncommon to sit down at a restaurant, check into a place, and a message will pop up from a friend of yours telling you what to order, or what to try, or what to talk to the bartender about.
And so, I mean, this is one of the first visualizations that we've seen of Foursquare data showing up in real-time. So, you can imagine the next time that you come to New York, or whenever you go back home, you can use these maps to really explore your city in a different way, and unearth some of this content in a different manner. It's pretty amazing stuff.
YUSUF MEHDI: And so, I'll just zoom out, completely, I'll come back to the U.S., and you can see what's pretty fascinating here is that picture of the U.S., and as people start to update, you can see people checking in live. And so one of the things that's powerful here is, when people post infrastructure, it's getting disaggregated from the local context. And what you can do now in the maps is, you can reaggregate. So you can now bring, again, that locality of what is going on when people post information, reassemble it here in real-time, and get a sense. So, as Dennis was saying, if suddenly there's a big music conference that he was at recently, South by Southwest, you could go down to Austin, Texas, and see all of the activity that's going on, and quickly get a sense of what's the pulse? What are people saying? What are the bars to go to? What bands are hot? And that you can get now. Imagine doing that in a visual way on the map. And that's really the big opportunity.
DENNIS CROWLEY: Yes. You can see here there's a whole bunch of people there unlocking badges for being the mayor of a particular venue, for being there more than anyone else. The red and yellow one is for being the local, for that being your local spot. I saw one all the way on the West Coast which was a camera icon, which means someone found a place with a photo booth. So, these are all things that you can unlock with Foursquare, and this is one of the first times we've really been able to visualize it on a map like this, which is great.
YUSUF MEHDI: Anyway, so that was all I wanted to do. Thanks, Dennis.
DENNIS CROWLEY: Great. Thanks for having me up, appreciate it.
MIKE GREHAN: You should hang around, Dennis, we're going to do some crashes. Fantastic demonstration, guys, thanks very much. (Applause.)
So, basically, this is kind of the game is changing. I mean, we started off talking about SEO, and 10 blue links, and what we're looking at here is completely multi-modal, it's pulling in data from so many different kinds of sources, and understanding the task that the end user is involved in, satisfying that longer-term information need where it was necessary before for me to look at a few links, and click on seven to see if they were relevant, and find out that they weren't, that you understand a lot more, and that you are actually kind of like what I was saying before, why do I have to ask you guys for this? Why don't you just give it to me? And it's all happening right there on the page, right in front of me.
YUSUF MEHDI: Yes, absolutely. The world is changing quite a lot, and I think you saw there some of the ways, I think, that the experience will change for consumers. And this is the first time we've showed it. Like I said, I think it's going to go live today. So, we wanted to wait to show it to you first here.
MIKE GREHAN: And we thank you for that. Fantastic, it was a brilliant demonstration. So, it is a game changer.
We have a few minutes left, I'm sure that you guys out there have questions to ask about this fabulous partnership to begin with, and some of the newer things that Yusuf has been talking about?
YUSUF MEHDI: There's a hand back there.
QUESTION: Hi. Congratulations on some great stuff here. One question is about, is this the end of shopping comparison sites, or do you work with shopping comparison sites, because we looked at some travel features, and some other like shopping for cars, and things like that? So, what does this mean for those sites? Are you working with them, or are you now basically their biggest competitor?
YUSUF MEHDI: That is a good question. The question is, what does this mean for shopping comparison sites, since we showed a couple of examples where you can do some shopping comparison. And I think that there are we still believe that there are many opportunities for comparison sites for shopping, for travel, et cetera. What we try and do there is provide one set of capabilities.
And I won't go back to it, but if you wanted to, for example, in travel, when we take you through the funnel, you can still ultimately before you buy go and say, hey, I want to go compare this with other sites, and we'll actually have some of those sites listed on the page. And so, in many cases, we are doing some of that functionality, but we still provide the link off to a number of different sites.
MIKE GREHAN: So, it is quite interesting when you think about it from the end user point of view that what's happened in search previously is that people come to Microsoft as it was before, or to Google, and they do a search, and you get 10 blue links. You click on one of those, and catapult off somewhere else. And now you're on a journey on your own. Whereas, in this type of approach, we can just sit here, and you'll just keep feeding us the information back. And we don't have to go tracking around looking for it.
YUSUF MEHDI: Yes, exactly. The key thing, whether you're an advertiser or a publisher is, I don't just need lots of traffic, give me the qualified traffic. And when someone comes, can you pass context for me, like are they ready to buy now, or are they still in research mode, so that when they come to you, you can better able deliver results for them.
MIKE GREHAN: I have a question down here.
QUESTION: Thank you. Great presentation. As Mike said, unlocking the black box that is the person actually interacting with the search engine is an important component for you to provide this richer and richer experience. So, what are some of the ways that you will be getting that information, or encouraging people to volunteer information which will help provide a better experience?
YUSUF MEHDI: Yes, I think it's a great question, and it's really right on the money. People, everyone has different levels of comfort with the privacy of who they are, and how you are known. And we respect that in a very deep way. So, we have, on Microsoft sites, the ability to go in and delete any information that we have on you at any time. So, if you want to delete all the cookies, you can go up and do that.
But, to your point, people will expect value in return. So, they'll say, hey, I'm happy to share some information about myself is you can return back for me either better products, or better answers, or value of different kinds. And our approach will be two efforts. One is, we'll show you some of these features. So, there is a feature up there on search history, in fact, I think it should be on that page, if not it will be up there soon, where you'll be able to go and look at the history results. But you can clear that, or you can say never show me history again. The benefit of history in this case is that we've studied – 50 percent of all searches for any given user are repeat searches. So, we're able to then provide that. When you do the auto-suggest, and I start typing in that box, and we do a drop down, we will bring your history in that dialogue box. And so, there will be different ways that we'll return value for you.
MIKE GREHAN: We have a question right down here.
QUESTION: Given Google's actions in China this week, I was just wondering how Microsoft plans to respond, whether they're going to try to move in more heavily to China, or is it going to respond in kind as to what Google is doing?
YUSUF MEHDI: And so the question is, what is Microsoft's position in China, and how does it compare relative to other people's efforts? You know, I think there really, everyone has their own point of view on how to approach things. And so, we don't judge on that. Microsoft's approach has been that we're in over 100 countries on a worldwide basis where they all have different laws, and how people are supposed to be behave, business is supposed to behave. And we respect those, and we'll follow those laws. Our view on the best way to help consumers around the globe is to be an active participant in helping shape that in the country. And I think you can have more impact being there in the country helping folks providing a service. And so that's the approach that we will take, and we will work with many parties on that, but that's our approach.
MIKE GREHAN: Another question right in the middle here.
QUESTION: All these features that you're showing are really, really cool, and really exciting, and look really fun. But, I'm wondering how the businesses can monetize these opportunities?
YUSUF MEHDI: Absolutely. So, I think there are a couple of different ways. Some of them, like I said, on let's say the very first part of the demonstration we showed, what happens there is, you're able to get better context for the consumer. And so what I think you'll start to see is the type of ads that you buy will change. And so, like I said to you before, if you know that this person is at point X in the purchase funnel, they're in the research phase, then the type of ad, and the type of ad text that you should show at the time should be more things about exploring. If you know they're at the point of purchase, then it's maybe more deal offer related, because you know now it's time to purchase. So, the first part will be the type of ads.
And the second one, like more in the map and demo area, that's a whole different set of experiences of how you monetize it. One of the things that I think is interesting on Foursquare that Dennis and I were talking about is, there are different ways that merchants will provide customer loyalty. So, for example, they did a Starbucks badge where if you come in often, you can actually help provide some value in that respect.
Some of those will be maybe more commerce related, they won't be advertising related at all. They'll be helping make a purchase. So, for example, again, if you know the user's time and location, and there's a local business at the time that you're going buy, they might say, hey, let's offer a coupon to people who have checked in in this location, or have this intent, and so those will also change.
MIKE GREHAN: So, can I just add to that, just talking about user intent, because this is one of the most important areas of just trying to decode what the intent is behind a query, and then serve up the right kind of content. With this blended search, as I was talking about before, if you understand whether this is navigational, or it's informational, or transactional, instead of just throwing out a Web page now, you have an opportunity to think, well, maybe a video is a better answer for that one, or maybe it's a PDF, or a white paper is a better answer for that one. And start optimizing around file types that we didn't think about before.
The one area that seems to be becoming fairly apparent, if you understand user intent, we're so used to typing in a query and seeing something in what we call the organic listings here, and then surrounding that there are these paid search ads that we know that some people are sponsoring, some people are paying for.
But if you understand user intent, and you get to the point of transaction, is there any point in showing these organic listings? Why wouldn't you just show all paid ads, because you know for a fact I'm sitting with a credit card?
YUSUF MEHDI: Yes, again, it's a good question. You're saying at the point of purchase wouldn't you just move to ads. Theoretically that could be possible. I think one of the great things here is it's all driven by data. So, in fact, our decision of when we show ads or not ads is all based on click-through rate and relevancy, as well as a number of other factors. And so, I think that will be some of the things we play with. But, ultimately users really do like to it's an interesting balance. While they like to have the exact answer, they also like to know that they have choice.
So, a lot of times we'll show something, like some of the feedback we had on earlier designs people said, hey, awesome that showed me the lowest price, but I need to see all the prices, or I won't feel like I really did get the lowest price. So, I need to see the choice. How do I know it's the lowest price? And so it's interesting, sometimes ease of use and convenience, and trust are all interwoven together.
MIKE GREHAN: Interesting. We have a question, right over here.
QUESTION: Sorry. I was the first one with my hand up. So, I'm sneaking in here. It's similar to what you just said. How many of those results were SEO and how many of those were SEM? There is no way to tell and I think that's the way, or do we have to think differently about how to work with Bing? In other words, how did you serve up most of those results in the first half of that your talk?
YUSUF MEHDI: You're talking about the rich caption, the rich answer, or the links below it? The rich caption is something that we assemble together, where we pull data from different sites, and that's something that we that's something that we do, and again we measure whether people like that or not, based on a number of factors like click through rate. The rest of them are basically the links in the same way that we described before. And so, how we rank those, and how we do all that, we don't we probably don't get into all the details, but it should be consistent with how SEOs and SEMs participate in that user experience today.
MIKE GREHAN: So, again, I've used this term multimodal before. I tend to use it a lot now, because it seems to be that it's a lot more about partnerships. I mean, the partnership with Foursquare, the partnership that Microsoft has with Twitter, to tap into that feed to get more of the real time stuff that's going on. So, you're kind of optimizing around a number of different sources and what I call signals, a number of different signals.
YUSUF MEHDI: Yes, exactly. I think you hit a keyword. Data signals is a thing that I think we all think about. It has really opened our eyes, I would say, over the last year about how core that has become to us being able to do this thing, to be able to understand user intent. And so, signals of data come from all sorts of things, like we talked about. Whether it is your lat long, or the speed at which you're moving, or the time of day, or things on your calendar, it can also be just the signals on how you use the search engine, how fast did you dwell, did you hover, did you click and come back. All these signals are things that we put in, and you can't have too many signals, really. I mean, you can have a hard computing problem, but you can't have too much data.
MIKE GREHAN: A lot of data to deal with. I think we have one more question over here.
QUESTION: Yes, this was a great user-centric experience, very rich, and we actually can get everything at our fingertips without leaving Bing. I didn't see you going to an actual Web site, which is great for the user. So, I'm wondering am I missing these visitors on my Web site. I cannot track them, how do I see partnership with Bing?
YUSUF MEHDI: That's a great question. I think a couple of things. The reason you didn't see the click through the Web site is just because I just didn't demonstrate going to the Web site. But, there's absolutely a lot of a huge amount of click through. In fact, as I said before, what we found is more click-throughs occur when we put some of the richer captions on there than less. And so, in many cases they do drive more click through, but to your question of how will you work in the future, one of the things that we're really thinking a lot about is how do we open up so that some of those things that we do aren't all just done by Microsoft. How do we open it up so that maybe other third parties can come in and provide their own efforts, even in the core search results.
I showed you on the maps how the different application vendors can provide an app on the map, but even in the core of the search results we're playing with the notion of is there a rich caption, a richer caption that someone could come and provide and then people could interact on that page, before they, let's say, clicked through, and so, different thoughts in that area that were also spending a lot of time thinking about.
MIKE GREHAN: So, I think that's been certainly a great conversation and I'm really pleased that you were able to make it. And thanks for making it exclusive for us. It would absolutely be fascinating to look at how much the end user is beginning to drive and become in charge of the search results and become a bit more demanding. It was a great look into the future.
Please say thank you to Yusuf Mehdi. (Applause.)
YUSUF MEHDI: Thank you.