searchmetrics email facebook github gplus instagram linkedin phone rss twitter whatsapp youtube arrow-right chevron-up chevron-down chevron-left chevron-right clock close menu search
1450114501

Google BERT Algorithm Update: Contextualizing User Search Intent – Jordan Koene // Searchmetrics

Episode Overview: Speculation and uncertainty surround Google’s sudden BERT algorithm update, raising questions among experts in the industry on how big of an impact it will have on search and marketing strategies. Join host Ben and Searchmetrics’ Jordan Koene as they discuss BERT’s initial impact, how it’s changing the way Google analyzes context behind user search queries and how companies should avoid making abrupt, reactionary changes to their marketing strategies.

Summary

  • Google’s rollout of previous updates show signs they may have begun testing BERT during previous algorithm updates.
  • The update modifies the language machine learning system behind Google search, which helps the engine better understand the contextual meaning behind both typed and voice searched queries.
  • Companies should avoid making abrupt changes to their strategies as it will take time to obtain actionable key insights regarding the update’s impact on search.
  • Google will continue to value strategies that identify the search intent of consumers and provide informative, valuable answers to consumer queries.

GUESTS & RESOURCES:

Ben:                 Welcome to an emergency Google core algorithm update edition of the Voices of Search podcast. I’m your host Benjamin Shapiro and today we’re going to reevaluate the changing landscape of Google search post the second October algorithm update. Joining us today is Jordan Koene who is the lead SEO strategist and CEO of Searchmetrics Inc. and today Jordan and I are going to pull back the curtain on what is being called the Google BERT algorithm update.

Ben:                 But before we hear from Jordan, I want to remind you that this podcast is brought to you by the marketing team at Searchmetrics. We are an SEO and content marketing platform that helps enterprise scale businesses, monitor their online presence and make data driven decisions. To support you our loyal podcast listeners. We’re offering a free trial, the Searchmetrics suite and content experience platforms.

Ben:                 If you’re interested in optimizing how your website content and SEO strategies perform, go to searchmetrics.com/freetrial to test out Searchmetrics suite. Okay. On with show, here’s my conversation with Jordan Koene lead SEO strategist and CEO of Searchmetrics Inc. Jordan, welcome back to the Voices of Search podcast.

Jordan:             For whatever reason it feels like it’s been a while since I’ve been on the show.

Ben:                 I was going to say, dude, we got to stop meeting like this. Every three weeks there’s a freaking algorithm update and we’ve got to scramble to figure out what the heck is going on. Google, come on guys.

Jordan:             Everyone at Google, once you get their annual bonus, they’re just pumping out updates.

Ben:                 Take it easy, will you? And this is the big one, big one. We thought that the core algorithm update was, you know, the big daddy. No, this is meaningful. Ten percent of all searches, all queries are being impacted by BERT. So, let’s start off the top. What happened? How did they announce the update?

Jordan:             Yeah. So, the update was announced I believe on October 25th through a blog post shared on the Google Search Liaison blog. And it basically, it was kind of an under the radar type of updates. But basically the summary of it is, “Hey, look everybody, we’re applying an update to the search engine that allows us to better understand your search queries. Better understand the meaning behind what you’re searching for.”

Ben:                 So, unlike the previous two updates where there was a little communication saying an update is coming, here’s the things that we’re going to be addressing. This was kind of old school Google where like, you know, you wake up, it’s Tuesday morning and have your coffee and oh crap, hey, there’s a massive change to how Google thinks about your queries and about your search.

Jordan:             Well that’s one way to look at it. But I actually think that and Google actually states this in their post, there was a lot of testing that actually took place. So, Google was testing how to evolve and innovate BERT into search results and use this new rigor in identifying the intent behind queries. And I believe that actually the recent core updates that we’ve seen in September and I believe July, are testing of that testing. So I actually think BERT was being used during those algorithm updates. And now what we’re seeing is just more of a cross pollination of what they learned from those updates across the entire search landscape of queries.

Ben:                 So, you kind of answered my next question, but what did they change? You mentioned how Google is essentially interpreting language. Talk to me about what BERT is.

Jordan:             Yeah. So let’s talk about a little bit about what BERT is. Because I think that this can be a largely misunderstood, you know, I think that even the posts that we shared, by Pandu on Google’s blog can be somewhat misunderstood or correlated to other updates Google’s done.

Jordan:             Well, let’s talk about what this really is. What it’s based on is understanding national language. So it’s using an LP, which is a technique, it is a training module. It’s like a machine learning training module that allows you to understand or represent, transform the meaning of certain terms or language that we use to apply it back to search queries. So I know that that was a mouthful, but in its simplest form, this is the way that Google can better provide results based on what the full context of the word is that you use.

Ben:                 Jordan, I’m a little confused. Give me an example.

Jordan:             Absolutely. Ben, let me give you an example. So, you know I have kids and my kids play a lot of Pokémon. So, you know, you could do a search query of “Playing Pokémon with game console” or “Playing Pokémon without game console.” And in the past you get essentially a mashup of the same results for that query. But now today you get very distinct results because the real intent behind that query is based off of one word and that is the word “With,” or “Without.”

Jordan:             And so, BERT helps Google understand these things and understand what is the real desire of the search query and thus what results should I [inaudible].

Ben:                 So, I guess the underlying update is really, and putting it in non-SEO terms, Google’s language brain got an upgrade, right? They understand or it understands more context than it did before.

Jordan:             That is absolutely correct. And I’m glad that you used the work brain because Google I believe in 2016 released, or maybe it was in 2015, released an update, a machine learning based update that … It was called RankBrain. And it essentially is in the same vein. And rank vein… excuse me RankBrain is also closely connected to Hummingbird, another update where Google is looking at the genuine validity of their entire search engine based on understanding the meaning of language. And I think that that’s a really powerful statement behind these updates.

Ben:                 So, as we take a step back and we think about what’s happened in the last couple of months with the three updates, there was Google looking at and addressing specifically how they were going to handle news, right? That’s what we were calling the Real News update. There were some changes around how people can tag their content and no follow and how people are handling links. There was the core algorithm update and now Google is essentially updating what I’m calling the language brain related to RankBrain. How do you think about the difference in the overlap between these three updates and what should we take away from all of these updates happening in such a short timeframe?

Jordan:             You know, I guess one simple way of thinking about this is like building a car. You know, a carmaker can focus on making the engine faster. They can focus on making the interior more spacious. They could focus on improving the amenities, like the audio system, or the lighting, and what Google basically just did in a matter of three to four months is they just basically fixed all, and they just made the whole carpet. And in this particular case, when it comes to BERT this is really an engine update. They made the engine stronger and faster and more reliable. Because at the end of the day what consumers are getting here is a much cleaner set of results based on what they were looking for.

Ben:                 So, this is something that’s going to benefit consumers because Google has a better understanding of the context in which their users are typing in their queries or speaking their queries. Talk to me about the impact you see on SEOs. It’s obviously early days. Have we seen an impact and if we have not, what do you think the impact is going to be?

Jordan:             Yeah, so we’ve seen some fluctuation in ranking results. In particular where we’re seeing some of the biggest volatility is in heavy content based websites. So for example Wikipedia. Google itself, you know like Google.com and its own assets in the way that they showcase their own assets. Other big sites are also seeing impact that are content creators and content aggregators such as Instagram and other rich content type sites that have UGC.

Jordan:             But the reality is that volatility in these kinds of changes isn’t necessarily the best indicator. And if RankBrain is any indication of the volatility that we should expect off of BERT, we won’t necessarily see like these monolithic, you know, fluctuations for specific domains. But we will see some volatility.

Ben:                 I believe you and I don’t because everything I’ve seen from all of the last three updates … and look, you’re the SEO expert. I’m just a talking head sitting in front of a microphone. But everything Google has done recently has somehow had this massive trajectory changing impact on YouTube. And all of these changes and maybe just obviously there’s some separation of Church and State and Google can’t unfavorably favor YouTube, but boy, YouTube is doing really well in terms of SEO visibility over the last two months. And there’s been a lot of changes. You mentioned that some of Google’s other properties have seen an impact as well. Is it just YouTube or what are the other changes that you’re seeing in Google’s properties?

Jordan:             Yeah, so there are certainly other changes that we’re seeing across the landscape. YouTube being one of them. But also, I think that most notably we’re seeing some volatility here is in the utilization of featured snippets and other elements that show up in the SERP. So with BERT, one of the beauties of having a natural language filter for lack of better terminology that provides you with better results is that in some cases it may seem more relevant or less relevant for Google to showcase these featured elements. And so obviously many of those featured elements are rich with in some cases Google properties. But the reality is that this is essentially one of the main reasons why Google is evolving the relationship to search queries so that they can showcase more of these elements in search.

Ben:                 So, Google is essentially trying to better understand what consumers are looking for so they can create more rich snippets so they can essentially keep traffic on their pages. Is kind of the, “I don’t believe in Google or I think they’re evil,” view of the world.

Jordan:             Right? I mean that’s certainly one way to look at it. But the other way to look at this, and I think this is why at least from many of the brands that we’ve spoken to post-BERT here, they’re not seeing huge traffic impact is because in many cases these are not queries in which there’s a high volume of click behavior. You know, people are searching for an answer, they’re trying to explore an answer. And that answer isn’t necessarily solved by clicking on 10 results in Google. The answer’s typically resolved by one click or in many cases zero clicks. And so the reality here is that this isn’t necessarily to dictate click behavior, but it’s allowing Google consumers, their users, their searchers to identify the problem, the solution, the expectation, the anxiety that they have behind that search query.

Ben:                 It sounds like it’s more of a step towards getting ready to make voice search a more dominant part of Google’s portfolio as they’re moving more towards, you know, position zero type content where direct answers. The more Google is able to understand what the answer to your question is, the more they can keep that in their experience, the more they don’t need a webpage in front of you.

Jordan:             There is no doubt in my mind that voice search data was used to [inaudible] BERT and we’re seeing this evolution of RankBrain. They even state in their own public update around BERT that they are trying to understand the relationship of other words in a sentence. They didn’t even use the word keyword. They just said sentence. And then using the word sentence is its total clue and giveaway that when they’re looking at search behavior at a voice level. Because nobody searches on full sentences when they’re typing. They only do that when they’re doing voice search.

Ben:                 Well, I think the only time you type a full sentence is when you have a question. “What is two plus two?” Technically, that’s a full sentence, but yeah, I understand what you’re saying. It’s the type of queries that replicate what you would be saying in voice. Jordan we’ve seen so many of these updates happen in a short period of time. Do you want to just record the Ernie podcast now we already have Bert. Is Ernie coming?

Jordan:             Yeah. Well we’ll see if they can find an acronym for their next NLP pre-trained instead of technology, but yeah…

Ben:                 Is BERT actually an acronym?

Jordan:             It is.

Ben:                 What is it an acronym for?

Jordan:             Bi-directional Encoder Representations from Transformers.

Ben:                 What in the hell does that mean?

Jordan:             BERT.

Ben:                 BERT. And that again, that’s the language brand got updated, right? That’s what that means.

Jordan:             I had to add a few more new keywords to my brain after that one.

Ben:                 All right. Jordan, any last words on the BERT update? Anything else SEOs should know? Any advice as they try to interpret how this is going to impact their web properties?

Jordan:             I remember when RankBrain came out and there was just a lot of hands thrown up in the air. There was a lot of speculation about changes coming down pipe. People were living in this kind of like eggshell, like environment, waiting for Google’s rankings to drop. And I want to encourage you to really reconsider the fear factor here and go back, look at your plan, look at what are you executing on and ensure that you’re addressing the expectations of users.

Jordan:             Because again, many of these updates, these semantic updates, they are intended to help consumers. And so if your SEO strategy is focused on helping your end consumer, you’re going to be a winner in the long run. And so I think that not a lot of folks should scared right now. I think they should be looking at their analytics, looking at their data, but remember that if you’re reaching the intention of your consumer, Google will find that value and it’ll be part of these BERT updates as along with others.

Ben:                 No need to panic everyone unless you’re trying to figure out what BERT actually means. All right. And that wraps up this emergency Google algorithm update edition of the Voices of Search podcast. Thanks for listening to my conversation with Jordan Koene lead SEO strategist and CEO of Searchmetrics Inc.

Ben:                 We’d love to continue this conversation with you, so if you’re interested in contacting Jordan, you can find a link to his LinkedIn profile on our show notes or you can send him a tweet where his handle is JTKoene. J, T, K, O, E, N, E. If you have general marketing questions or if you want to talk to me about this podcast, if you’d like to be a guest on the Voices Of Search Podcast. You can find my contact information in our show notes or you can tweet me at Ben J. Shap, B, E, N, J, S, H, A, P.

Ben:                 If you’re interested in learning more about how to use search data to boost your organic traffic, online visibility, or to gain competitive insights, head of the searchmetrics.com/freetrial for your trial of the search metrics, SEO suite and content experience software. And if you liked this podcast and you want a regular stream of SEO and content marketing insights in your podcast feed, hit the subscribe button in your podcast app and we’ll be back in your feed later this week. All right, that’s it for today, but until next time, remember, the answers are always in the data.