Bummer! This is just a preview. You need to be signed in with a Pro account to view the entire video.
Strings to Things: Entities and SEO31:14 with Matthew Brown
In the last year, Google and Bing have both indicated a shift to entity-based search results as part of their evolution. Google has unscored this point with rich snippets and Knowledge Graph, and Bing has now upped the ante on personal search results with Bing Snapshots. Find out how you can adopt strategies to stay ahead of the curve in the new world of semantic search results.
[MozCon - Matthew Brown - Strings to Things: Entities and SEO] 0:00 [Matthew Brown] All right. I was joking with Cyrus that after Dr. Matt, I could be like Certificate of Completion Matt or GED Matt. [laughter] 0:03 So I'm here to talk a little bit about the futuristic stuff. 0:12 That was really good stuff from Dr. Matt where it was what's happening in SEO now and what we're doing. 0:14 But first let's talk about me. I only want to talk about me right now. 0:19 Back when I was a wee ginger, hanging out with other gingers—this was 2011. 0:22 And you look at this guy who's got this confident business lean, sleeves rolled up. 0:28 What could strike fear in his heart? What could make him do a MozCon presentation that looked like this? 0:33 So this was 2011, and this was right before Google was like, 0:40 "Okay, we're releasing Panda and if you have thin content pages or page with lots of ads or pages that just suck in general, 0:43 we're coming for you." 0:51 That was the very beginning of it. So my whole presentation was , uh-oh, if you're a site like Mahalo, 0:53 which took us to the logical extreme where you're just spinning keyword derivations in your URLs or your title tags 0:57 or your keywords. 1:04 Or take it to an even more logical extreme, you don't have 1 page on Heartgard dog medicine, 1:06 you have 500,000 on Heartgard dog medicine. 1:11 This was industrial-strength SEO that worked pretty well from like 2003 all the way to 2011. 1:16 If you had a site with enough domain authority or a site with enough links, you could make this work. 1:23 If you get enough pages in Google's index, they're going to rank. 1:27 They're going to rank for every keyword variation, every keyword query that you can possibly imagine. 1:30 Panda kind of changed the game. 1:37 And I had this, I wouldn't say it was a great suggestion, but I think it would work which was, 1:39 okay, if this is happening, if Mahalo is going to get killed by Alexa—and a lot of the comparison shopping sites took a hit. 1:44 And a lot of really big sites with a lot of brand signals with a lot of domain authority got hit, too. 1:50 So the whole thing was, wow, they're coming for everybody with Panda. 1:54 They're coming for you, me, it doesn't matter who you are. 1:57 This was my solution—build pages like this. This is the BBC page for Eric B. & Rakim. 2:00 They don't really have content for this particular rap group, but what they do have is the ability to build pages that look like this, 2:04 that are really good content and they aren't bare and they aren't thin and they aren't placeholders and they've got keywords 2:12 and it's really good stuff. 2:19 But they built it from a page that doesn't look too pretty. This page, which looks awesome, was built from this. 2:21 And what this is, is this is actually the database version of the Wikipedia page that they built this from. 2:26 DBpedia is nothing but Wikipedia in structured data form. 2:32 So they can just call pages like this and build something that looks like that awesome page on BBC. 2:36 So my solution was we should do this, too. 2:41 If you have pages that are thin content, use the power of the semantic web, and use structured data and databases 2:44 and build awesome pages that not only Google likes, but users will like. 2:49 And then you solve you're whole problem with being hit by Panda because you have too many URLs 2:53 and too much thin content and too much crap. 2:56 One company in particular did an awesome job at this, at was Google. 2:59 Basically they said, "Oh, we have all this structured data in Freebase, 3:05 "and we can scrape it out of DBpedia, and we can use all of the sites that we're looking at, too, 3:08 "and we can build Knowledge Graph carousels and Knowledge Graph results. 3:13 "And we can tell you the weather; we can tell you data facts; 3:16 we can tell you all of these semantic results that you can get from the semantic web, and we can put them in the search results." 3:18 So that's where they were headed, and that's what scared me. 3:24 Now fast-forward to 2013—2013 Google I/O hits and Amit Singhal, who's the the director of all things search for Google, 3:29 he says, "Search is dramatically changing before our eyes." 3:35 What he means by that—you've got to take Amit with a little grain of salt because Amit's on some Star Trek shit, right? 3:39 Google could be talking about a nuclear space station or self-propelled car or dark fiber—like you never really know 3:44 what he's talking about when he says it, but he says it's fundamentally changing. 3:50 But I do listen when Jason Douglas, who is the Knowledge Graph Product Manager for Google, when he stands up and says, 3:54 "This is our flagship product"—search. This is what makes Google, Google. 4:01 This is why Google has a $300 billion market cap is because of our core search product, and we're fundamentally changing it. 4:05 Billions of people use it, and we're changing it before your very eyes. 4:11 He's deadly serious about it. 4:14 And you have to ask yourself, why would Google change something that's fundamentally been working for them for so long? 4:16 Here's a hint. This just agrees a little bit with what Rand said earlier about desktop search queries. 4:23 On the left, healthy growth from desktop PC sales in Q1 of 2012. On the right, freefall. 4:27 So Q1 2013, all of a sudden we're not buying desktop PCs anymore. Windows 8 isn't selling how it was supposed to. 4:33 It's all tablets and notebooks. So Jason Douglas of Google Knowledge Graph, he was pretty clear. He's like, "It's mobile." 4:39 Everything that we're doing with Knowledge Graph and everything that we're doing with Semantic Web 4:46 and all of this search stuff that we're going for, this is all happening because of mobile. 4:50 So this slide is from Mary Meeker's presentation. Mary Meeker is at Kleiner Perkins Caufield & Byers. 4:53 Her famous State of Union—it's basically the State of the Union for the Internet. 4:59 She does this twice a year, and every year she puts graphs up that are like, 5:03 "C'mon, it's all about mobile. It's happening everywhere else. It's happening here." 5:06 Point nine (0.9) percent of all Internet traffic worldwide in 2009 went through mobile. 5:10 That trend line projects out that by 2013, 30% of all Internet traffic will be on a mobile device. 5:15 So this is what' Google's going for. 5:21 They're saying, "People aren't going to click the back button. They're not going to type in long keyword queries. 5:23 "They're certainly not going to click the back button on their mobile phone and type in keyword query after keyword query. 5:28 We have to find a way to get them the information that we're looking for faster than we've been able to in the past." 5:32 So this is another scary slide. If you really want to tie it into a scary slide, this is Facebook ad revenue per user. 5:38 Blue bars—desktop ad revenues per user. Yellow bars—mobile revenue per user. 5:43 They're making up the difference in what they're losing on the desktop, and they're losing it via mobile. 5:50 So if there's any slide that scares Google, it's this one. 5:54 If they lose the battle for mobile ad revenue per user, that's the one thing that could challenge Google's domination of search. 5:57 It's pretty clear what they're going for. They're telegraphing it. 6:05 They're saying it in their Google I/O presentations that it's going to be a mobile search world, 6:07 and we need to get good at doing mobile semantically-related results 6:11 that don't require us to pull all of this intent out of keyword queries. 6:15 So the first bullet point of their presentation on structured data and Knowledge Graphs said, "This is the end of search as we know it." 6:20 And you know when Google says something like that, SEOs, we're going to go crazy about it, right? 6:25 We're going to argue about it; we're going to blog post about it; there's going to be comment threads. 6:29 We're going to spend hundreds of hours debating on whether or not this is actually the end of search. And it's kind of silly, right? 6:32 SEO won't die. As long as there's a user intent and we can optimize just our websites, but mobile and content and images 6:38 and video and slideshows and whatever other content forms there are—that's going to be SEO, no matter how you slice it. 6:44 Links aren't going anywhere either. 6:50 So if anybody tells you this is the end of links, that's ridiculous, too, because all of these semantically-related data sources, 6:52 whether it's about a musician, whether it's about an artist, a person, a place, a thing, an object, or a concept, 6:58 they live in a structured database somewhere linked to with a unique identifier. 7:03 There's links to all of the structured data in DBpedia and IMDB and all of the CIA World Factbook data. 7:08 All of these sources are linked together with URIs, just like URLs link web documents together. 7:15 So links are going to evolve a little bit, too. 7:19 And we'll be talking about things like data rank and link trust and who's got a trusted link sources and who's structured data is trusted. 7:22 But building links—maybe will be building links to data, but we'll still be building links, 7:29 and links are still going to be an important ranking factor going forward. 7:33 That doesn't change fundamentally with the semantic web. 7:36 Somebody told me one time that good SEOs make a lot of small bets, and this is true. 7:40 If you make bets on what Google is going to do or what Google can't ignore or where Google is going, those usually pay off. 7:43 In 2002 before the launch of Google, you kind of bet on keywords. Then Google launched and you kind of bet on links. 7:50 Fast-forward to 2006, you saw that Dig was getting a lot of traffic and Twitter was coming on board 7:55 and there was real-time search and all these things were happening. 8:00 Then it was like if you bet on 2006 and 2007, not only did you clean up on social traffic, you probably cleaned up on SEO, too, 8:03 because you built a lot of links and a lot of visibility. 8:10 So this is a really good bet to make right now. You saw Matt's slide that looked like this, right? 8:12 Less than 10% of people are using schema.org on that data set. 8:17 If you look at sites like Blekko's grep the web or you look at BuiltWidth and see how much schema markup is out there, 8:21 it's a ridiculously low percentage of domains across the web that have structured-data markup 8:27 and have schema.org in use right now. 8:32 So this is a great time to make this bet. 8:34 This isn't something that's been beaten to death yet; 8:36 this is something the engines are telegraphing they want to do, and so far they don't have most of the web marked up 8:38 in a form of how they can do it. 8:43 So talking about leveling up a little bit in semantic SEO—it's a little bit different than some of the other presentations here, 8:47 because as SEOs or as search marketers, we all know links counts, we all know social media counts, 8:51 we all know keywords and title tags count and h1s count, and all the things that Matt presented in the correlation data. 8:56 This is futuristic stuff. 9:02 This is not stuff that somebody's going to be able to write a post tomorrow and say, "Oh, I can tell you how to do semantic SEO. 9:04 I can tell you how to rank for structured data and entity base search." 9:09 It's not there, yet. So a lot of this is something that—and I think this is exciting. 9:13 It's forward looking, and we haven't figured it out completely yet. 9:17 But there's one good thing that's working in our favor. 9:20 Unlike when you sit down and see Matt Cutts talk to somebody or you see a Google blog post and it's really not transparent— 9:23 it's like, "We're releasing updates," or, "We're releasing penalties." 9:29 And Matt Cutts sits down with Danny Sullivan and you're like, "What did that mean?" 9:31 Panda's rolling out over another update over 14 days; you can't tell what ranking factors are. 9:34 It's all confusing. You don't know what it is. 9:39 The research side of all 3 search engines, plus Facebook, they're being very transparent 9:42 about what they're doing and what they're solving. 9:47 We're just looking in the wrong place. 9:49 We're paying attention to the wrong people at Google, Yahoo, and Bing, 9:51 because all of the research out there shows exactly what they're trying to do and what problem they're solving. 9:53 They need to disambiguate entities—people, places, things, objects, concepts. 9:58 Anything that they can get out of a keyword string—and there's some ridiculous names for Snoop right here. 10:02 Does anybody call him Snoop Scorsese? [laughter] It doesn't even make sense. 10:08 So they need to be able to know that that is the artist known as Snoop Dogg, and that's a harder problem than it sounds like. 10:12 Obviously if you see a picture of him, everybody in this room knows who I'm talking about, 10:17 but given the ridiculous amount of keyword strings Google sees across the web, they don't really know. 10:20 So they need to solve this problem at scale to figure out from web documents, from rich media types, from video, 10:25 what you're talking about on an entity level. 10:31 This is really easy when you do it with Wikipedia. 10:35 If Google sees—and they did this with the WikiLinks Corpus—if they see 100,000 pages all linked to artist, Bansy, 10:37 and it all points to a Wikipedia page for Banksy, they're like, "Bam! We have a disambiguated entity. 10:44 We know you're talking about Banksy. We know who Banksy is. We know all of the related associated facts about it. 10:49 We can do this on a really finite data set. If we can do it on Wikipedia, that's easy. And they've already done it. 10:54 They've already done it on DBpedia; they've already done it on Knowledge Graph. You see it. 11:00 Every time you see a Knowledge Graph, that's Google saying, "I've got an entity here. I know exactly what it is. 11:03 I can serve results for it." 11:08 It looks like this. This is the Tron: Legacy Daft Punk soundtrack, DBpedia page. 11:10 When Google gets a link to that or Bing gets a link to that or Yahoo does, really easy. They know exactly what's in there. 11:15 What if it looks like this? This could be about Tron. It's about Tron, the movie. It's a game. 11:20 Obviously it's not about the superior in 1982 version of the movie, but it's really good. 11:27 But they don't know what you're talking about here. 11:32 Even on a title tag or an h1 tag, people are linking to this, and maybe they're using anchor text or not. 11:34 They don't know exactly what to make of this, and so it's a lot harder for them to do this at web scale. 11:39 So all of these connected on the Google Knowledge Graph and all these connected things in all these databases, 11:44 Google, Bing, and Yahoo are trying to solve the same problem which is, how do we do what we do with DBpedia 11:49 and structured data and Knowledge Graph, how do we do this across the other 90% of the web and all of the content types 11:54 and all of this HTML that we know exists? 12:00 How do we get there? 12:03 We can see this in their presentations. 12:05 If you go and watch anything from Google I/O, watch this structured data and Knowledge Graph presentation, 12:07 because they're very transparent that we need structured data across the web to do what we've done with Knowledge Graph 12:10 across the rest of the web. 12:16 So we need to promote schema.org. 12:18 And they're saying, "We need to promote it with web publishers and small business owners and people that have small sites 12:20 "and people that have big sites. 12:24 "If we want to be able to serve these semantically-related results, which are really good for mobile— 12:26 and we know by 2015, 2016, mobile's going to be hugely important; we need people to do this." 12:30 And that's really their challenge. That's what they're trying solve. 12:35 So they're releasing things that are like this. This is the schema sameAs tag. 12:39 If you mark up items with schema.org, you've probably done something that looks like the top right here. 12:42 You've marked up something that's a schema.org actor type; it's Tom Hanks. That's pretty easy, right? 12:46 Now they've released the sameAs tag which says, "Okay, we don't really know what you mean, necessarily, 12:50 or you could have screwed this up, or we don't trust you." 12:55 If you put a sameAs tag that says it's the same thing as the Tom Hanks Wikipedia edition, that's essentially telling them 12:57 you can trust me. 13:04 You can trust that I've disambiguated this entity before you, and I'm talking about Tom Hanks, the actor. 13:06 So you're going to see more and more this out of Schema which is like, please link to a trusted data source 13:10 because we don't know that you're a spammer or we don't know you know how to do this, but if you link it to IMDB 13:14 and you link it to DBpedia and you link it to WIkipedia, then maybe we'll trust you 13:18 and maybe we'll give you a rich snippet or rank you higher or something like that. 13:22 So that's recently released. We're going to see a lot of stuff that looks like that. 13:26 And we see it in conversational search, too. 13:30 If you go to your Google Chrome browser on a microphone and search for weather or search for Portland or search for Who am I, 13:31 Google doesn't just give you a string of 10 links based on relevancy. 13:37 They know who you are, if you've logged into Chrome. They know where you are by location. 13:39 If you're connected on Google Plus, they know what you're connected to, on personalization. 13:43 They're starting to determine user intent and you can see it in conversational search 13:47 based on way more than just keywords or strings like that. 13:51 So if I type in whether I'm going to get like a Wikipedia or a Dictionary.com definition, it's going to be ridiculous. 13:54 But if I say it into conversational search they're like, I know this dude is in Portland. I know what he's looking for. 13:59 I know what he searched on in his last session. I know what he searched on the last hour. 14:04 It can go on and on and on. 14:08 This is going to move out of conversational search into this experimental thing and be more and more 14:10 of what we're going to see out of desktop search. 14:13 Bing research is actually super awesome about this. 14:17 On Bing research, if you go to Microsoft on their research site and search for entities or search for semantic web 14:20 or some of the things I'm talking about right now, they've got gobs of research that talk about exactly how they look 14:25 at structured data markup and how they rank it and what they show semantic results for. 14:30 So in this case they're saying, "There's certain things we trust schema.org markup for and there's certain categories we don't." 14:35 And they're explicit about, "This is a category we don't even look at structured data markup, and this is a category 14:40 where we can trust it because we've verified that they're good query results, and that's exactly what users want." 14:47 So you can see what—Bing is completely transparently telegraphing when they show entity search and when they don't. 14:52 Yahoo's doing some awesome work, too. Don't sleep on Yahoo on this. 14:58 This graph right here is before they deployed Spark, which is their semantic web, Yahoo Knowledge Base. 15:01 It's similar to Google Knowledge Graph—the click-through rate before and after. 15:06 And they saw massive increases in click-through rate as soon as they launched this semantic web service, 15:09 which is interesting because all 3 engines say, "We look at user data. We look at click-through rate. 15:14 "We look at bounce rate. We look at what we would call pogo-sticking. We look at query logs. 15:20 We look at what they searched for again." 15:24 They're taking user signals—all the stuff that we know is important for SEO and they're saying, 15:26 when semantic results are launched, they're better. 15:31 And when we don't have semantic web results in entity base results, they're not as good. 15:33 And if you actually have gotten a link from Yahoo on the homepage or one of the section fronts, 15:37 you probably know that drives more traffic than maybe all of your Google organic search traffic in a month. 15:42 Data analysts everywhere and web analysts are like, yeah, January was because we had a link on the Yahoo homepage 15:48 and then everything else was going out over here. 15:53 That's what Yahoo's doing with this. 15:56 They're going to serve related results on their sports pages, their entertainment home sections, their real estate sections. 15:57 And if you get one of those semantically related links, you might not care about your Google traffic that much. 16:02 So definitely don't sleep on Yahoo on this. 16:07 One of the things that's good about this whole research based part of Google, Bing, and Yahoo and what they've been doing 16:11 is the fact that a lot of these tools that Google has used are available to us, too, and the best one is Freebase. 16:17 Freebase is the structured database that Google used—they bought it in 2010— 16:22 to build Knowledge Graph and to build all of the semantic results. 16:27 What's awesome about Freebase is it's a creative comments license on everybody that contributed to it. 16:30 So we can use it just like Google used it. 16:35 They even gave us a How to Tap into the Knowledge Graph Freebase API tutorial in Google I/O. 16:37 So if you want to figure out how to have entities baked into your search process or baked into your content process, 16:42 they gave it away on how you can do it. 16:48 And it's awesome because you can get ranked lists of entities from the Freebase API just like Google ranks them. 16:50 So they actually use their algorithm Freebase to say, okay, if you search for Stanley and Stanley is a film director, 16:55 we're going to give you a score that says Stanley Kubrick is the most relevant entity 17:01 as a film director in the Freebase entity database. 17:05 So you can start developing models that are ranked, just like Google ranked them, if you want to use the Freebase API 17:09 to get ranked lists of what Google considers the most important people, places, and things around a specific topic. 17:13 If you want to go back and do what the BBC did and roll your own content, Freebase Topic API does the same thing. 17:19 You can query the Freebase Topic API and get back—if you put it in entity—you can get back their Wikipedia bio, 17:24 images, thumbnails, videos, other relevant entities, facts, figures—all of these things are available here, 17:29 just like roll your own Knowledge Graph. 17:35 If you want to do that, the Freebase Topic API let's you do that. 17:37 Google also has a tilde search operator which can give you semantically related results. 17:42 If you type in a keyword, you can get a result—and actually I'm not even going to—my PowerPoint's awesome. 17:45 They actually disabled this 2 weeks ago. 17:51 And the reason they disabled it is because they said, "We don'thave enough service space for it. We don't have enough processing for it." 17:53 Which you kind of think, really? 17:58 You're building dark fibre and you're doing all this stuff, and you can't buy a couple of servers to continue to support the search operator? 18:00 So they're actually taking a little bit of data and a little bit of research possibilities 18:06 so you can't get entity and semantic-related search results directly from the search bar anymore. 18:09 However, Bing has something awesome now that my friend, David Mimm, showed me, which is at explicit.bing.net. 18:14 And if you put in a query string—if you actually just type in that domain, you'll get redirected to Bing.com, 18:22 but if you put in a query parameter into this particular Bing search URL, you get an entity disambiguated search result, 18:26 which might be different than the same search result on Bing.com. 18:34 So what I'm predicting this is, this is Bing saying this is explicit.bing.net with an entity is showing you what a search would look like 18:37 when we absolutely know there's a semantically-related result on that page. 18:44 And Bing.com is like, maybe we do know, maybe we don't know. 18:47 So we're going to show you whatever we have, which could be image thumbnails or pictures or local graphs or things like that. 18:53 So if you want to get a peek under the hood of what Bing's doing with entity search results, this is a good place to start. 18:58 Bottlenose is a social media semantically-related layer, so if you go to Bottlenose, you can look at a social entity-related graph 19:03 based on Google+ or Facebook data. 19:10 So if I type in SEO Moz, which I did in this example, that graph shows me entities that they're able to extract from the newsfeed 19:12 and social feed of Google+ and Facebook. 19:19 So you can see that there's MozCast and MozCon and there Q&A, and there's all of these different layers of social entities 19:20 that you can relate to the keyword string there. 19:26 So you can go from strings to things on the social graph by using a tool like Bottlenose. 19:29 Really useful to layer that over with data that you're extracting, your normal keyword research data, 19:33 and some of the other things that you're doing. 19:38 Yahoo is apparently naming some of their research tools after either Gentleman's Clubs or Mariah Carey movies now. 19:40 So what Yahoo Glimmer is, is it's a peek under the hood of what Yahoo's doing with entity search. 19:46 So just like Freebase API gives you ranked lists of entities based on order on what Google is seeing in Freebase, 19:51 Yahoo Glimmer is saying for Yahoo Knowledge Base, if you type in an entity, 19:58 here's a related entity for person, place, thing, or object from that string. 20:03 It's a little bit weird to use. They don't rank it quite right. 20:06 But if you really want to see how Yahoo Knowledge Base is giving you semantically-related results, that's one way to do it. 20:09 And finally, Fresh Web Explorer. We talked about brand a little bit. 20:15 I know that Rand and Matt just talked about it and how brand entity mentions and brand mentions correlated 20:18 pretty highly in that study to ranking or pages that tend to rank well. 20:23 One thing that I should mention about brand is Google says specifically, we don't really look at them as brand signals, 20:28 like we don't really look at them as brand mentions. 20:34 And what they mean by that, in my opinion, is those are disambiguated entities. 20:37 We call it a brand signal, and we know for sure it's about Apple computer, we know for sure it's about Moz, 20:42 or we know for sure it's about your company's neighbor, your website's name. 20:47 That's what they mean by a brand mention. 20:51 And what that is, is that's a linked, easy-to-follow mention that people have that they can find across the web. 20:53 A good place to find those is Fresh Web Explorer. 20:58 So you can think of if you go to Fresh Web Explorer and look for your brand or your company name and then you see 21:00 that there's an unlinked mention somewhere and you want to link there, that's building a brand signal. 21:05 You're building a disambiguated entity signal that Google's like and Bing's like and Yahoo's like, I know that's talking about this company. 21:10 It links to the right Wikipedia page. It's marked up and structured data that this is company type, SEOMoz or Moz 21:16 or Apple or any of those things in structured data. 21:22 I know what I can do there. That's a brand signal. 21:25 And if you look at it and think of tit in terms of how an SEO—we're talking about partial-match anchor text 21:27 or exact-match anchor text and a balance of those. 21:32 We're really looking for a balance of entity signals and keyword signals in the traditional SEO sense. 21:35 So Fresh Web Explorer is a great way to track both of those kind of entity mentions. 21:41 RelFinder is specifically for DBpedia. 21:46 So DBpedia is the database for general Wikipedia, right? 21:48 RelFinder let's you put a graph together of what's in DBpedia and compare entity types and where they're linked to on a graph. 21:51 I'm really visual. If I want to put in a BMW, Mercedes, and Volkswagen in DBpedia into RelFinder, 21:57 I can see all the other entities that have co-references with those entities. 22:02 So maybe it's a racetrack or maybe it's a specific model type or maybe it's car manufacturers or some of the other things. 22:06 If I can pick out entities from that graph in DBpedia, I can start layering out content models on that and mixing it up 22:12 with the other data that I've accumulated. 22:18 Okay, so what do you do with all of this entity research from Freebase and Glimmer and Bing and all these things 22:21 that give you entities, which are things instead of keyword strings? 22:26 How do you actually work with them and build a content model? 22:29 Well I think we've seen that you want to avoid targeting entity-based SERPs like this. 22:33 This is where Google has eaten everybody's lunch. Things to do in Seattle. 22:37 There's a carousel.. People are going to click on a lot of the pictures. There's Knowledge Graph results. 22:41 There's facts, figures, events, weather's on the side. This is where it's really tough to rank now. 22:45 If you're not in the top 5, you might not get any clicks on a page like this because it's eating so much of the clicks on the page. 22:50 When you're doing this kind of research, you want to look for something like, okay, I put in Bottlenose Seattle, 22:56 and I put in beer and wine and all these things. 23:00 And then I can start backing that out into string searches like this. 23:03 Oh, well I found Seattle and beer—related entities that are heavily connected. But this SERP, this looks doable, right? 23:06 Beer advocate's pretty strong, but there's a lot sites in here that you could probably rank and you could probably be relevant 23:12 for entities that are both related to Seattle and beer. 23:16 So you really want to get away form searches that look like that. 23:20 Dr. Pete on Wednesday is going to show you a lot of variations where Google has really changed the game 23:23 on how search results look and find the longer-tail, entity-based queries. 23:27 This is what it looks like from Freebase. I took Freebase entities; I plugged them in. 23:31 Then I put some Seattle-based entities and I plug them into Bottlenose. 23:35 Then all of a sudden you can get an entity graph that looks like, wow, we could do a content model around that. 23:38 We could actually launch a travel section or a Seattle section that looks a little bit like what we would normally do 23:43 for SEO based query string SEO. 23:48 We can do that for entities and now build a content model. 23:51 So I had this great idea about how to build a semantic content model and maybe how to do it across one site or a bunch of sites. 23:54 And then I realized that it was going to end up like this. 24:00 If you stuff entities into a page or a set of pages, just like we've stuffed keywords on pages— 24:03 and I think every SEO has done something like this, at least a test over the years—that's how natural it's going to look, 24:10 like a dog with a shark on its head. 24:14 Google's not looking for you to have a page full of entity-related string keywords. 24:16 They're looking for you to have structured data that says, this page is definitively about these things, 24:20 and they're looking for other people to link with you that they can disambiguate the entities in those links to say, 24:26 hey that's what we're looking for, too. 24:31 They're not looking for entity-base keywords across the site. 24:33 Simon Penson of Zazzle Media UK, he did a really great semantic content model on marketing led. 24:38 I had this great idea on how to do it but he beat me to it, so he gets all the credit. 24:44 What it is, is it takes your keyword process on how you would normally do keyword-base string SEO, 24:48 and then it layers in semantic entity results in a microsite strategy. 24:52 Maybe I have a beer site that's related about beer. Maybe I review beers in other cities. 24:57 Maybe that links to my Seattle site about things to do in Seattle. 25:01 And all of a sudden you're mixing keyword research that's happening that we would normally be doing in SEO 25:04 with an entity content model, so that you're semantically in the right place. 25:09 For entities that are related to each other, you've got content that matches that and is relevant to that 25:13 and marked up with structured data; that's going to be really useful. 25:18 The last step is—a lot of SEOs ask me this, too—are we going to profit form this? 25:22 Is this going to be something where Google takes all of our structured data and all of our semantic mark up 25:26 and all of the entity-based work that we do, are we going to profit from it or are is it just going to get sucked into Knowledge Graph 25:30 or Bing Snapshots or Facebook or things like that? 25:35 I personally believe that based on what Google has said and all their research, 25:38 they say we need to get out of what we have in Knowledge Graph, which is just primarily Wikipedia and IMDB and a few other sources, 25:43 and they need to get to the rest of the web. 25:49 And to do that they need us to mark up their data, so from everything that they've said so far, 25:51 you should be able to see more sources be part of the Knowledge Graph and be part of structured data 25:56 and be part of the general semantic web so that it's not a closed system like it is now. 25:59 It actually makes publishers and website owners, it makes it worthwhile for us to mark up our data semantic. 26:04 So that was a lot. Thank you. I'll take questions. [applause] 26:12 [male speaker] This is why I have to download your slide deck afterwards. There's just a lot of information. 26:23 So using your crystal ball, we've seen Google incorporate more semantic data into the snippets, into the search results. 26:28 I think just last week they were blowing up pictures for news results. 26:36 Do you see more of that happening and influencing click-through rates at a greater effect? 26:40 [Matt] Yeah, and I think that is the fact that if they're going to mobile, they know that they might have 1 shot at a query. 26:45 If it's on a Smartphone or if it's on a tablet, you might not refine the query to get exactly what you're looking for. 26:52 So they want to get your eyes on a thumbnail or a snippet or something that stands out to get a click off of that 26:57 rather than maybe you bail out and don't search again. 27:02 [male speaker] Darin. 27:06 [Darin] Hey, Matt. So you talk about the benefit of doing this. 27:08 If you get in with the grand floor, you're placing a smart bet. What are the things we can do now? 27:14 What can we mark up now on our sites that Google will start eating up and contributing to our being on the floor like this? 27:20 [Matt] I think that they've said everything that's in rich snippets now—everything that's in Knowledge Graph 27:30 on the carousels and to the side, are the same things that we're putting in rich snippets— 27:35 events and things that are on Data Highlighter and news and thumbnails and video stuff. 27:40 So they're going to start moving stuff that we would normally see marked up in the 10 blue links into that side bar 27:44 and into those carousels. 27:50 So I'd start there and say, okay, I'm going to mark things up that are really consistent with what I'm seeing 27:52 Google generate rich snippets for and what Bing generates snapshots or rich snippets for. 27:56 And then slowly as they migrate more and more types into those graphs, you should see maybe, hey I've got a snippet for this 28:00 and I've got a link in the Knowledge Graph and I've got one of the pictures in the carousel. 28:07 That's the chain I would prioritize when you start using mark up like that. 28:11 [male speaker] Yes. 28:17 [male speaker] With Google's recent pushes into semantic image recognition, how do you see that affecting search? 28:19 [Matt] That's interesting. 28:25 So there was a panel at Sem-Tec Biz, early in San Fransisco this month where somebody from Google 28:27 who's in the Google research division did how to take events that are happening and look at images from those events, 28:33 and make semantic associations about what's happening and when it's happening and what it's about. 28:39 And the whole presentation was about social data. 28:44 So in other words when they see all of these images happen for maybe a specific news events, 28:48 they actually go extract the meaning from those images. 28:52 They go extract it from what's happening on Twitter if then can see it or what's happening on Google+ 28:55 or what social networks they have visibility to. 29:00 So it's interesting that they're taking social data and saying, well we don't know what these images are about 29:02 and there's not a lot of keyword data about it—maybe it just recently happened—so we're going to go mind social 29:06 to be able to tell you exactly what this is a picture of and what it's semantically related to. 29:10 So I think in a round about way the answer is, they have to get there with the most recent data they can, which is social. 29:14 [male speaker] Thank you. 29:21 [male speaker] Matt, one question I had— 29:23
You need to sign up for Treehouse in order to download course files.Sign up