This is our Deep Dive Into Local from April 2nd, 2018. In our Deep Dive series, we take a closer look at one thing in local that caught our attention and deserves a longer discussion.
If you have a special topic you would like us to discuss for the Deep Dive in Local, please reach out to us. If you would like to be on one or the other of our segments, reach out and send us the topic and your availability.
If you are interested in sponsoring this weekly show also please let us know.
Our weekly discussions are also now available as a podcast as well. SUBSCRIBE TO THE PODCAST HERE.
Mike: Hi. Welcome to the Deep Dive in Local. This week, we have special guest star, Cindy Krum — one of my all-time favorite most energetic local SEOs but who understands and appreciates local more than most SEOs, so welcome to the show, Cindy. You just finished a series of four very in-depth comprehensive articles that have a high degree of relevance to local search. Maybe, you could give folks just a brief overview of those articles?
Cindy: Sure. Like you said, they’re very long and in-depth and they’re all about mobile-first indexing and they break mobile-first indexing into a couple different topics. They talk about shopping, and they talk about media, and payments. And the very last one is actually my favorite one because it talks about maps and language and leads into entity understanding and how entities are being into maps? It’s most relevant for, I think, your audience. I mean it’s just super cool and I’m sure you’ve seen in Google Maps, the buttons with the grocery store, and pharmacy, and mail. And that seems like a no-brainer but I think it’s the beginning of entities getting into maps. And it’s especially relevant when you’re traveling.
Because it used to be if you’re traveling in a foreign country and you didn’t know how to speak the local language? With a keyword search, you would have to go to Google Translate to search for, “How do you say grocery store in German?” And then, you would have to put that in, and that’s so error-prone to get the right search result. But, now, they’re understanding that a grocery store is a grocery store — regardless of what language you speak. It’s an entity. It’s an idea — regardless of where you are or what language you speak so they’re putting buttons for grocery store and that’s the synthesis of entity understanding and how it’s useful for people. In Maps, they’ll get you to a grocery store even if you don’t speak the language, you don’t have to know the keyword for it. It’s great.
Mike: I’ve been rewatching Star Trek Deep Space Nine, where they have, obviously, the universal translator which seems to understand, both object/subject, verb and action in every language and manages to translate it in real-time between dissimilar species so this is something very similar, Same thing. So I’ve been looking…obviously, I spend a lot of time looking at entities and I have this…local was the basis for Google’s entity creation. It’s , they were dealing with local before they had the knowledge graph but the core element to the knowledge graph, when they first got going, before they went out to people, and things, and all of these relationships was local. And then, over this weekend, I was doing research as to how does the knowledge panel present on mobile? And a couple of trends that reflect this idea of yours that Google is looking to become the presentation layer of the internet. And what I saw was that the knowledge panel in mobile was multiple scrolls of the screen at the top. And the primary calls-to-action were things like overview, reviews, posts. And then I saw a new one mentioned on Twitter — service items. Where, literally, Google was creating a mini-website for these businesses where the main call-to-action was down the rabbit hole/into the rabbit hole.
Cindy: Absolutely.
Mike: So and then, obviously, important parts of these were reviews, and Q&A, and various other text elements. So perhaps you could give us some sense of how you see Google dealing with Q&A, and reviews, language processing, and machine learning, and blah-blah-blah-blah-blah?
Cindy: Well, and think about Google Post, they’re enabling people to have Google-hosted websites through Google Post and through, you know, Google My Business, first, but now Google Post. Which is like a mini WordPress/Google Post plug-in that…so the way I see it happening and I’m working on a whole new article series that I hope is going to be as successful and as useful as the mobile-first one, where I’m renaming mobile-first indexing to be entity-first indexing because I think that that’s what it is. I think that mobile-first indexing is a red herring. They’re crawling with a mobile user agent and that’s fine but that’s not revolutionary. They’ve been doing that, actually, for a while — that’s nothing new.
What’s new is the way that the web and the algorithm and the indexes all work together — and it’s based on entity hierarchies. When you talk about with entity results and the knowledge panel taking up so much room is a perfect example of entity-oriented search, and entity-oriented algorithms, and indexing. Because what’s happened is, now, you know…it used to be that Google’s index was based, primarily, on domains and you had to have a domain to be in the index. That’s not true, anymore. Now, entities are in the index and they have been for a while. Now, I believe the entities are primary in the index and domains can be entities. That’s considered larger than domains that can also be…or can also be index…in the hierarchy.
Mike: That relates to some research I was doing…I’m getting some feedback from somebody, here. Does somebody…I don’t know where but anyways. It relates to some research I was doing about how entities are ranked and how a domain is, essentially, an entity ranking signal. That they’re no longer ranking a webpage — they’re ranking the entity, And when they rank the entity, everybody needs to rethink about, what does it mean to rank the entity? And so a domain or a page on a domain is no more than one ranking signal, like reviews are a ranking signal of that entity, or reviews on a prominent Yelp page are a ranking signal of that entity, or a Wikipedia page is a ranking signal of that entity. And so we have to shift our thinking from the idea of domain prominence and page prominence to entity prominence. And that domain is just one more ranking element in a range of elements that they can now use to rank entities — in local and every place else.
Cindy: Yeah. You and I are totally on the same page. So I think that Google has almost been not telling 100% of the truth. They’ve been telling…I know — controversial. I know.
Mike: Yeah. Google, often, lies by omission. They don’t lie, overtly. They just don’t tell you the whole truth. That’s standard operating position…
Cindy: …they tell you, “It’s just about, we’re changing the crawler that’s crawling things.” But what’s happening with mobile-first indexing is a reindexing of the entire web. And that’s why they’re crawling all their websites with the new crawler, they’re reindexing things by entity — with an entity understanding. And that’s why mobile-first indexing happened at one-domain-at-a-time system. Where domains got added one-domain-at-a-time — not one-page-at-a-time — one-domain-at-a-time. Because the whole domain had to be evaluated for its entity understanding. And I think that the domain, when it was added, it had to have entity understanding and it was translated…or, at least, the entity understanding was translated before it could be added. And the translation is part of entity understanding because entities are language agnostic. Entities exist outside of language. Keywords describe entities but entities don’t have to have a language. So a blue parrot is a blue parrot in any language or with a picture, We can have a picture of a blue parrot and it’s still a blue parrot even if we don’t have the keyword blue parrot to describe it, It’s hard and it’s a weird concept for some SEOs to get but concepts like inflation, This is a concept that exists…
Mike: …it’s purely a mental construct.
Cindy: Right. And so there are fewer keywords to describe it but it exists outside the keywords, And so that’s really important for entity understanding to happen. And so I think part of mobile-first indexing or what we’ve called mobile-first indexing was entity-first understanding. So they crawled it with the mobile user agent, evaluated it from entity understanding, and evaluated in all the different languages for entity understanding. So that it could say, “Is this also true in this language? Is this entity understanding also true in this language?” Yup. Yup. Yup. And then it was officially added. What I’m working on, right now, is an article that explains all that.
Mike: So let’s take local as an example because that’s what I’m familiar with. We have this entity in a database, which is an abstraction in itself — an entity of a local business. The real local business exists in reality but we have this thing in a database that we need to…that Google needs to understand what it is? They need to understand how it ranks, compared to other businesses? But, then, other businesses in a geography? Other businesses in its category in that geography? So is geography an entity?
Cindy: Yes, absolutely.
Mike: So we have a geography that’s Williamsville, which is the city in which this entity exists. And then, we have jewels, which is the category in which it exists. And we have jewelry appraisal and so where does…in the end…so we do a search for jewelry appraisal, Williamsville. Google knows all those concepts and delivers up local packs as the answer to that,
Cindy: Yes. So we have entities and we have keyword modifiers to entities. But then we have businesses, right, and businesses are also entities. So when you have Williamsville and then you have jewelryy and then Google can tell intent from things like where you’re searching? Like, if you’re searching for Maps, you’re probably…if you’re searching for jewelry in Maps, you’re looking for a business. You’re not a treasure hunt,
Mike: Right. At least, I’m not.
Cindy: You’re probably not, so you’re looking for a business so it’s going to go to business entities. If you’re searching for jewelry on your TV, then, it’s using the context of what device you’re using to search to determine what an entity is the most appropriate.
Mike: I see, so that’ll lead to…and when there’s ambiguity, that’s why you might get…
Cindy: …a knowledge graph…
Mike: …you might get a knowledge graph but you also might get pictures, instead/or in addition to, So you might get photos of engagement rings, or shopping items engagement rings, or local pack engagement rings — all mixed together because there’s some ambiguity in that, in their understanding of what entity you’re actually looking for?
Cindy: Yeah, I think we’re going to get…we’re going to see knowledge graph when there’s ambiguity about the context or when multiple kinds of behaviors would be appropriate for your context — knowledge graph is going to become disambiguation portal. So like on a TV if I’m searching for jewelry, I might be looking for a video or I might be looking for just to swipe through pictures with like my whole family around. Like, do we like this engagement ring or that engagement ring, as a group? So they’ll use that as a disambiguation for intent. But in a car, A, it’s probably going to be illegal to watch videos or swipe through images so we’re just going to be looking for Maps or businesses.
Mike: Got it. That’s interesting. What do you see is the role of machine learning in all of this? And, in terms of language understanding and…etc.?
Cindy: So I think so machine learning is…you asked about reviews and the Q&A? And I’m seeing Q&A multiple/choice questions and any opportunity for feedback pop-up absolutely everywhere. Google’s added apps in PWAs and ways to pull in any feedback or help in categorization everywhere they can.
Mike: So some of that’s freeform, like reviews and Q&A?
Cindy: …but a lot of it is multiple choice and multiple choice is the best way to train an AI, or to train machine learning. Because it’s taking hints from the freeform stuff and putting it into…narrowing it down into something that it can use as a recommendation. So for instance, if it sees freeform feedback, like there’s no handicap ramp, here, If someone’s angry and says, “There’s no handicap ramp, here.” When you put that into a multiple choice question about the handicap accessibility of the venue or of the restaurant. And say, “Would you say this is a handicap accessible place? Yes or no? Or, would you say were the bathrooms handicap accessible? Was there a ramp somewhere?”
And so it’ll narrow that down so that it can give it a thumbs up or thumbs down in the filtering of — is this a handicap accessible place, Because someone angry about one aspect of it doesn’t necessarily mean it gets a full-thumbs-up or full-thumbs-down on handicap accessibility and they need to answer that in a better way.
Mike: Right. Right. So one of the things I always have been looking at a great deal is those Q&A quality and review quality? And you would think that with these freeform features that Google’s machine learning would be better able to ascertain inappropriate questions and inappropriate reviews and keep them out of the system. And, yet, they either…they haven’t devoted very much energy to this or the system isn’t smart enough. Or, because it isn’t structured like you said, with specific multiple choice questions, these terrible review qualities in Q&A seemed to be making it through. In the case of Q&A, I was seeing between 10% and 25% of the questions were gross violations, in terms of service. Then, when I saw their algorithm run, it got rid of about 7%, only, which left them somewhere between 50% and 500% shy of the mark.
Cindy: Yeah. But the filtering, I think the filtering is critical for where Google really wants to be, which Sundar Pichai is talking almost exclusively about AI and saying that AI is going to change the world, more than the wheel/more than fire, It’s big. He is betting on AI, so more than fire is a pretty big claim so if we take that into consideration? The filtering is going to go into the voice search and what Google calls eyes-free. Eyes-free, right now, to us, means, basically, voice — voice search.
So let’s say I’m interacting with an assistant and I say, “Okay, Assistant…” I’m going to not trigger it because it’s sitting right, here. “Okay, Assistant…I want a restaurant. I want to find a restaurant and I want to make a reservation.” And it says, “Okay.” And it answers all the things. It’s going to be able to execute this but it needs to be able to not do a search because the search…Google’s search algorithm would just give me a bunch of options that I would have to manually go through. It needs to be able to think like with a brain and through that eyes-free. So what I would be doing when I’m searching is saying, “Okay. It needs to be handicap accessible, and it needs to be open at 6:00, and it needs to have a table for two available at 6:00…” and all of the things. And that’s all done eyes-free with voice and so the filters, the AI, and the feedback that we’re giving it with the freeform…and, then, the freeform being funneled into multiple choice and the multiple choice being fed into filters is how we’re training…we are an army of people training Google’s AI.
Mike: Big surprise there,
Mary: We always have been.
Cindy: …and we always have been.
Mike: So if I say, “Okay, Google, who is Cindy Krum?” Am I going to get an answer right now or not, from your machine?
Cindy: I don’t know. I haven’t tried that one. I don’t even know if I want that…oh, she says she doesn’t know. You actually made it work and she said she didn’t know.
Mike: So how do you see this playing out…first, let me preface this by saying that Cindy’s going to be at LocalU Advanced, April 12 and speaking on how agencies and local SEOs can be approaching voice, now. So how do you see this playing out, in terms of marketing and marketing optimization?
Cindy: So I think the main thing is, obviously, being aware of it and monitoring it. Looking at how reviews are coming in for your business? And the other thing is understanding, reviews are being translated. So if you’re putting reviews in, in any language for your business, or people are searching for reviews in any language — Google’s translating those reviews to be appropriate for whatever language the person is searching in it. And most people in the U.S. are unaware of this but it’s much more common and people/and businesses are much more aware of it in other countries. But sometimes, those translations are bad and I think this is going to become more and more common, too. That Google owns the reviews, they’ll translate them in anyway they like. So giving feedback on the transactions/giving feedback on the reviews and making sure reviews meet terms of service. Making sure that the right kinds of reviews are being started and fed into the system. Making sure that you’re setting yourself up to be added or labeled or engaged within people’s phones, actively. Because I think private indexing on people’s individual devices is going to become much more important. Google really want…for AI to work, there has to be a high-level of personalization. Because, otherwise, weirdos throw off the AI.
Mike: So does this relate to Google’s recent announcement they’re going to be discontinuing their shortening service — goo.gl and replacing it with Firebase links?
Cindy: Yes, absolutely. Google wants to push everyone towards Firebase and I’ve been saying that for two years, now. Google’s looking at ways to get people into Firebase because Firebase…their goal with Firebase is to make app-indexing possible and more scalable. And Firebase pushes…pushes us towards Google hosting of app content. And the more Google can host — the more they can understand engagement, which means the better their algorithm is because they want to surface the most engaging content.
Mike: So I was thinking so you are of the opinion that engagement increases prominence or visibility of an entity — yes or no? How do you see that?
Cindy: It will. My opinion is that it absolutely will. Does it, now? At least, as a secondary metric — yes.
Mike: Yeah. Because I was thinking of this circle reinforcing process where you have a local entity’s knowledge panel and you enhance it with nice photos and good reviews, which increases user engagement, which increases conversions. You end up with a happy customer who writes you a good review, which increases engagement, again, with the knowledge panel, which increases visibility of the knowledge panel. So you have this virtuous cycle that if you’re doing things right, good imagery, good reviews, good customer service increasing with good posts and good Q&A — you’re getting increased engagement but you’re also getting increased conversions when it’s also leading to increase visibility through increase rank and, thus, creating a virtuous cycle. Would you agree with that idea?
Cindy: Yes, I would. And, maybe, we need tinfoil hats for this theory. But I mean Google knows where our phones are, physically — especially, Android phones but I think, also, to some degree, iOS phones. If we’re a searching for a local business and then we go to the local business, they know that we’ve been there, especially my phone because it asks me for reviews about how the food was — that’s engagement.
Mike: No, that’s the ultimate engagement.
Cindy: It’s not web engagement but that’s not only engagement — it’s success. So if I, then, give a five-star review, does that help it rank? I mean not the five-star review itself but the whole virtuous cycle of she searched for it today and went and then gave a five-star review. Is that more powerful than me going either not reviewing or me searching and not going? Or, me searching and going next week?
Mike: I’m not sure you need a tin hat for that one but I think it’s…given how much Google knows about Android and Android users behaviors — I don’t doubt one bit that that information is feeding into this entity’s understanding cycle.
Cindy: I mean do you believe that that’s an SEO tactic that we could gain? Could we go get a bunch of Android phones do this and game it?
Mike: Well, I actually had this idea once…I think it was about a year ago maybe where you would have a virtual scavenger hunt that required you to go to actual physical locations. So you’d have like five businesses in the wedding industry and you would have to go to each location and check-in on the map. You have to search that location, go to that location, check-in on the map, and go to them, And you’d have a contest and, thus, drive up engagement across five locations across a series of mobile devices encouraging people to log in. And I thought it would increase rank, actually — search…
Cindy: …and, did it?
Mike: Well, nobody’s ever tried it, yet. It’s a big idea. It’s all ready to try.
Cindy: Let’s try, this. Let’s do it at Austin — we’ll have a scavenger hunt.
Mike: All right. Good. You organize it — let me know. I’m the last of the great delegators. I think of ideas but I don’t want to actually have to do all the nitty gritty work to get them done.
Cindy: I’m going to delegate, Mary.
Mike: Oh, she loves that. She loves that. I could tell you, she’s really big on that. All right, so I think it’s time for these entities to say goodbye. I really appreciate you being, here. I want to, again, encourage people who have enjoyed the conversation, Cindy, to join us at LocalU Advanced, the night of April 11 for a meet-and-greet and all-day the 12. Some great attendees — Mary, being one. David Mihm is going to talk about what the agency of 2020 is going to need to look like, two folks from Google — Marissa Nordahl and Allyson Wright, myself, Mary, Ed Reese, Tine Reese. Joy Hawkins is going to be there so you’ll be able to actually solve Google problems that day. So anyway we’re looking forward to seeing you and I want to thank you for joining us. And with that, I’ll say goodbye. Thanks, again, for joining Deep Dive and Local.
Cindy: Thanks very much.
Male: Bye-bye.
