Candour

Episode 112: Desktop page experience, new Google Ads features and MU

Play this episode:

Or get it on:

What's in this episode?

In this episode, you will hear Mark Williams-Cook talking about:

Desktop page experience: Google IO announcement confirms page experience update will come to desktop results

Google Ads features: Google launches a set of new features for advertisers

MUM: A 'breakthrough' in AI for Google generating results for complex queries

Show notes

Google New Bidding Features

https://support.google.com/google-ads/answer/10715180

Page Experience Ranking to desktop

https://www.youtube.com/watch?v=h00kn5J-F2Q

Search Engine Journal research

https://research.google/pubs/pub49971/

Transcription

MC: Welcome to Episode 112 of the Search with Candour podcast, recorded on Friday, the 21st of May 2021. My name is Mark Williams-Cook, and today I'm going to be talking to you about the Google I/O annual developer conference and some interesting information they gave us about page experience ranking. We're going to be talking about new Google Ads bidding features, and the big subject we're going to talk about is MUM, and that is the new Google breakthrough, as they call it, in artificial intelligence and search ranking.

Before we kick off, I want to tell you this episode is very kindly sponsored by our wonderful friends at Sitebulb. Sitebulb, if you have not heard of it, is a desktop-based SEO auditing tool for Windows and Mac. I'm delighted, actually, that we got the opportunity to get Sitebulb as a partner for this podcast because it's a tool that we've used in the agency for quite a few years. I've used it personally. I'm a big fan of it. So it's really nice to be able to talk about things on the podcast that I genuinely believe in and I was recommending anyway, before this podcast, to lots of people when I've taught SEO courses.

And that's something that I think Sitebulb as an auditing tool is really helpful for and it's quite unique for both agencies and in-house companies that are trying to do their own SEO in that, it's far more beginner-friendly than a lot of tools. And the reason for this is that when you actually run an audit, which you just do from your own desktop with the tool pointing it at your site, apart from they've got a really kind of nice help system, when this audit is complete, it doesn't just kind of kick you back data about, "This page is indexed. This page is no-indexed. This page has a canonical tag. This one doesn't." It actually tries to diagnose potential issues with that configuration. It will try and prioritise them for you.

But to top all of this off and as someone that's building an SEO course and has spent over a year doing it, I can tell you the biggest bit of work they've done is this kind of back-end knowledge centre they've built on the web. So every one of these issues, you can click on them and it will give you a pretty in-depth explanation on what this issue is, why it's an issue for search, how it's commonly caused, how it's normally fixed, and a general, obviously of context as all tools do, priority. And that's been a really helpful tool to help people learn more about SEO as well.

So when I've run SEO courses, we used to spend a long time looking at various crawling tools, and then we had to kind of go through; you run the crawl and you need to try and look for these kinds of patterns. Sitebulb does that for you. So I can't recommend it highly enough even if you are experienced. If you are experienced, hopefully you've heard of it already. They've got a special offer for Search with Candour listeners. If you go to sitebulb.com/swc, you get an extended 60-day trial of the software. You don't need to put a credit card in or anything. So you can just download it for free, give it a go, see how you like it. So give that a go. I'm sure you'll love it.

Google I/O, their annual developer conference happened virtually, of course, not in person, unfortunately, between the 18th and 20th of May this year, last year. And I think the year before, I did a roundup of things that were covered in Google I/O because sometimes there's interesting stuff for SEOs as it is a web developer conference. I'm still going through the announcements and videos that have surfaced for Google I/O, but there was one particular thing I wanted to share with all the SEOs out there, which is there is a very nice 13-minute long Google I/O session on preparing for page experience ranking. It's a really no-frills, in-depth, meaty look at page experience, not just core web vitals; so intrusive interstitials, safe browsing, HTTPS, mobile-friendly, all of that stuff.

But the takeaway that I picked up from that video, which I hadn't heard anywhere else before, is that the page experience ranking is also coming to desktop. So up until this moment, although we've talked loads and lots of the SEO community have talked loads about core web vitals page experience, it's always been with the caveat that these are going to be for mobile searches only. Now as expected. Obviously, page experience is important even if you're on a desktop as well, and that's reflected in this video. And there aren't any specific dates given, unfortunately, but it does make clear that this page experience algorithm, these signals, are going to be used in desktop ranking as well. So now there is absolutely no excuse to back out of really tackling these page experience issues, I mean, as I've always said, hopefully you were anyway for reasons outside of SEO.

So in my opinion, this page experience stuff, this core web vitals stuff are primarily user satisfaction metrics. And it just so happens that they're so important as user satisfaction metrics that search engines are paying attention to them as well. So hopefully, nobody's played that, "Well, it only reflects mobile searches and we get a lot of desktop traffic, so we'll deprioritise it." Hopefully it hasn't been the situation you've been in. But I will link to the video in our podcast show notes, which will always be available at search.withcandour.co.uk. Click the link there and you'll be able to see the video. You can share that with your dev team, stakeholders, and it's right there. It's Google saying that this page experience is coming to the desktop. So that can hopefully strengthen your case if it was a bit weak internally before. So definitely something everyone needs to focus on. Just wanted to share that with you. Over the next week or so, I'll be having a look at the other sessions that came out of Google I/O and picking out anything that I think is important to share with you.

I also have some news for you in the shape of PPC and Google Ads. These updates and new features actually can... It's really easy to let them slip you by. So this was published on the Google Ads help blog a couple of days ago, on May the 19th, and it's about new bidding features which Google say will help you reach your goals. Now there's a few new features. Some of them I think are quite helpful to you, and some of them, I think it's fair to say, are probably more helpful to Google than necessarily the advertiser. As usual, search.withcandour.co.uk for the show notes and links to the post if you want to see it for yourself.

I'll just read what Google has written here: so Smart Bidding is the foundation of a successful automated strategy. Questionable, but okay. Especially when paired with tools like broad match and responsive search ads, because we all know broad match works really well. Just slinging that on your account and you'll be fine. To make it even easier for you to manage these bid strategies and drive performance, we're rolling out several new features. One, get more insights with top signals for target ROAS. So ROAS, for those that don't know, is an acronym for Return On Advertising Spend, and maximise conversion value. Top signals in the bid strategy report give more transparency into the factors that drive your campaign performance and can help inform your broader marketing strategy. Currently, you can only view top signals for campaigns that use target cost per and maximise conversions. Going forward, these signals are now available for search campaigns using target ROAS and maximise conversion as well.

Secondly, apply seasonality adjustments at the manager account level. Smart Bidding uses machine learning to automatically set bids at auction time to improve the performance of your campaigns. Though these strategies already take seasonality into account, there may be key moments when you expect significant shifts in conversion rate, like during sales or product launches. We made seasonality adjustments to give you more control in these situations to make these adjustments even easier to use. You can now set them at the manager account level instead of creating them for each individual account.

That's probably quite helpful, to be honest. It just shows that there are these gaps of future knowledge that you have as a fleshy human that these models obviously can't account for because they're going on historical data and they don't know exactly what's going to happen. So it's just allowing you to kind of poke through that model to make these changes at a broader manager account level. Two other features, which are implemented to maximise conversion value bidding with recommendations. We're making it easier... hooray... to opt in, to maximise conversion value bidding by surfacing new opportunities on the recommendations page. These recommendations will show for eligible search campaigns and can help you get more conversion value for your budget.

As long-time listeners of Search with Candour will know, myself, Rob especially, who manages a lot of our PPC accounts, and I imagine probably most professional PPC people out there are not fans of the Google recommendations in Google Ads. We had a whole podcast episode (episode 99) on this. Again, we'll link to it in the show notes. But the kind of recommendations and optimisation score you get within Google Ads, just kind of clicking all of those does not necessarily improve the performance of your account. It's saying your account isn't fully optimised, for instance, if you could spend more budget. Yeah. I've got a lot of thoughts about that. I don't think it's particularly helpful. This is definitely a change for Google to get more people to kind of just opt in to stuff.

And lastly, predict performance with target impression and share simulators. Simulators can help you understand how target bid and budget changes may impact performance. This tool was previously available for conversion and click-based strategies and we are now expanding support to search campaigns using target impression share. With this change, you can now understand how adjustments to your target impression share can impact metrics like cost, impression share, and clicks. So that is actually quite helpful.

One of the strong points of PPC, especially compared to SEO, is the ease of which and the kind of accuracy with which you can do some forecasting and build models from that. So anything to help predict performance is always welcome. But those changes have been made live, and as I said, are linked to the post if you want to have a look at it, but thought you would appreciate being aware of them.

Okay. This is the big story if you like and interestingly, I haven't had too many people talking about it or not as many as I would have kind of guessed. This is about MUM and that's capital M-U-M, which Google is describing as a new artificial intelligence milestone for understanding information. There's a really fascinating introductory post on the Google keyword blog under their product updates by Pandu Nayak, who is a Google fellow and vice president of search, that was made three days ago on May the 18th.

I'm just going to read through the post with you because there's not a huge amount of information they've given in detail. So I'm just going to go through everything that they've said and then pull out what I think is interesting in terms of SEO relevance and what we do. So this is what Nayak wrote.

"When I tell people I work on Google search, I'm sometimes asked, is there any work left to be done? The short answer is an emphatic yes. There are countless challenges we're trying to solve, so Google search works better for you. Today, we're sharing how we're addressing one many of us can identify with having to type out many queries and perform many searches to get the answer you need.

Take this scenario. You've hiked Mt. Adams. Now you want to hike Mount Fuji next fall and you want to know what to do differently to prepare. Today, Google could help you with this, but it would take many thoughtful, considered searches. You'd have to think about the elevation of each mountain, the average temperature in the fall, difficulty of hiking trails, the right gear to use, and more.

After a number of searches, you'd eventually be able to get the answer you need. But if you were talking to a hiking expert, you could ask one question. 'What should I do differently to prepare?' You'd get a thoughtful answer that takes into account the nuances of your task at hand and guides you through the many things to consider.

This example is not unique. Many of us tackle all sorts of tasks that require multiple steps with Google every day. In fact, we find people issue eight queries on average for complex tasks like this one." - I thought that s tat on its own was quite interesting.

"Today's search engines aren't quite sophisticated enough to answer the way an expert would. But with a new technology called Multitask Unified Model or MUM, we're getting closer to helping you with these types of complex needs. So in the future, you'll need fewer searches to get things done.

MUM has the potential to transform how Google helps you with complex tasks. Like BERT, MUM is built on a transformer architecture, but it's 1,000 times more powerful. MUM not only understands language but also generates it. It's trained across 75 different languages and many different tasks at once, allowing it to develop a more comprehensive understanding of information and world knowledge than previous models.

A MUM is multimodal. So it understands information across text and images, and in the future can expand to more modalities like video and audio. Take the question about hiking Mt. Fuji. MUM could understand you're comparing two mountains. So elevation and trail information may be relevant. It could also understand that in the context of hiking, to prepare could include things like fitness training as well as finding the right gear. Since MUM can surface insights based on its deep knowledge of the world, it could highlight that while both the mountains are roughly the same elevation, fall is the rainy season on Mt. Fuji, so you might need a waterproof jacket. MUM could also surface helpful subtopics for deeper exploration, like the top-rated gear or best training exercises with pointers to help for articles, videos, and images across the web.

Language can be a significant barrier to accessing information. MUM has the potential to break down these boundaries by transferring knowledge across languages. It can learn from sources that aren't written in the language you wrote your search in and help bring that information to you. Say there's really helpful information about Mt. Fuji written in Japanese. Today, you probably won't find it in the search you do, but MUM could transfer knowledge from sources across languages and use those insights to find the most relevant results in your preferred language. So in the future, when you're searching for information about visiting Mt. Fuji, you might see results like where to enjoy the best views of the mountain in the area and popular souvenir shops, all information more commonly found when searching in Japanese.

Understanding information across types. MUM is multimodal, which means it can understand information from different formats like web pages, pictures, and more, simultaneously. Eventually you might be able to take a photo of your hiking boots and ask, "Can I use these to hike Mt. Fuji?" MUM would understand the image and connect it with your question to let you know your boots would work just fine. It could then point you to a blog with a list of recommended gear." And then they finished with a section about applying advanced AI to search responsibly.

"Whenever we take a leap forward with AI to make the world's information more accessible, we do so responsibly. Every improvement to Google Search undergoes a rigorous evaluation process to ensure we're providing more relevant, helpful results. Human raters who follow our search quality rating guidelines help us understand how well our results perform. Just as we've carefully tested the many applications of BERT since 2019, MUM will undergo the same process as we apply these models in search. Specifically, we'll look for patterns that may indicate bias in machine learning to avoid introducing bias into our systems. We'll also apply learnings from our latest research on how to reduce the carbon footprint of training systems like MUM to make sure search keeps running as efficiently as possible. We'll bring MUM-powered features and improvements to our products in the coming months and years. Though we're in the early days of exploring MUM, it's an important milestone towards a future where Google can understand all of the different ways people naturally communicate and interpret information."

So what do we need to take away from all of this? There is a lot to unpack here. Firstly, in terms of the actual technology, as far as I could see... and I did read an article as well about this on Search Engine Journal... there doesn't seem to be any specific patents filed by Google that specifically mention MUM. So this is likely to be the name given to a combination of other algorithms and systems and sub-patents, not kind of a specific technology itself.

The Search Engine Journal did find a Google research paper, which we'll link to in the show notes at search.withcandour.co.uk, which is called HyperGrid Transformers: Towards A Single Model for Multiple Tasks. And this reading the abstract kind of sounds like the same eventual goals as MUM, although this paper doesn't mention that specifically. And of course, just because Google has a paper or even a patent on something does not mean it's necessarily applied in any way to search. And obviously, even if it was, we wouldn't know exactly where in that system it sits. So I wouldn't put too much weight on that, but we're linked to the paper if that's kind of your thing. I know quite a few SEOs really enjoy just reading the source and trying to understand the tech behind it.

The most interesting thing for me that I took out of this announcement is that Google talks about language generation. And we've seen loads of examples and we've had some examples on the podcast. I went through a whole bunch of really cool ones for GPT-3, which is the language AI model where you can feed it questions and it will generate its own unique answers just written in English. It's not just scraping that answer because it's found somewhere on the web. It's actually "understanding", for lack of me knowing a better word for it, kind of the query and predicting what the answer should be.

So this basically means that Google is building this database of knowledge, as we know. We've talked about knowledge graphs before. Google talks a lot about entities and things, not strings, and working out the connections between those things. It's still taking this information from the web, from websites, and it's understanding the query and basically generating content around it. Now that means potentially no website's going to be shown. So if we take, for instance, that query they gave an example about how would I prepare to hike Mt. Fuji after Mount Adams or whatever it was, even if there was maybe an article about that, Google is building a system that can just generate the answer from multiple sources cause that's kind of the point, and almost write the web page, if you like, for you.

Now, I think there's some very dangerous ground there in terms of, well, who's going to be credited with that answer, right? Because the information the system has been trained on has come through human-written content that probably someone's been paid to write, a company's invested in it with the hope that people will read it and search engines will point them to it.

Now there's again been lots of back and forth with SEOs about Google scraping content for things like featured snippets. There's been lots of discussion about whether you should use schema and allow Google to just show the answer to your queries in the search result. And certainly, people like Rand Fishkin have been very vocal about zero click results and Google's kind of retorted on that about how people use search. But the fact is there is a portion of searches where Google just does scrape the content from a page, shows it to the user, and there's no benefit for the website for that. There's only the benefit from Google in terms of people therefore finding their search engine helpful, use it more, and it secures their ad revenue for search, basically. So I'm very interested in how that future is going to roll out because the line obviously Google falls back on is around it's better for the user.

So even if you say, "Okay, well, I actually have that content about how to prepare for this training to go from this mountain to this mountain," Google could say, "Well, actually our research shows that because it loads faster, people prefer to stay within the Google ecosystem and just have that answer generated for them from multiple sources. Therefore, maybe it's more trustworthy." So there's kind of a lot of get-out clause arguments there for Google, I think, they'll take when it comes to language generation and actually creating websites that they've used as training data.

The one thing they did mention that I read out in the article was obviously they said, "We might be able to link you to helpful articles or helpful videos." And again, they haven't gone into any sort of depth about what that's going to look like. Again, I think this is going to fall into this... It's even more important to have that entity footprint in Google so they know which brands, which products, which subjects you're talking about so you can get linked into this graph that they're building and become a source for that.

I think the thing we have to face as search marketers, as digital marketers, is if users do prefer that and they take that convenience over... I don't know over what... over from what our point of view can look like, I guess, a monopoly on that information, whether they accept that deal is out of our hands, much like how some people, for instance, are very against buying from Amazon because they don't like what they do with tax or workers. But then I've had a lot of people say that, but then they still do it because it's so convenient, right? Because they've got Prime and they just press buy. And then it's with them the next day. I think the same applies in a lot of ways to Google.

So I used DuckDuckGo for a year as an experiment on my phone, and one of the most-used commands I ended up using on DuckDuckGo was the exclamation mark G, which actually pushes my query to Google because I found actually a lot of the time, I couldn't find the information I wanted to as quickly with DuckDuckGo. So I fell back on that crutch of, "Well, it's just faster, more convenient for me to use Google. Look, I'm busy. I'm in a rush. I don't have time to do this." So I think that Google might be relying on that from a public usage point of view of if there's this discussion around, "Look, actually come to our news websites. Come to our informational sites." Users may just be like, "Well, look, I'm busy. I just want to know this. Google just tells me, so sorry. That's what's easiest for me."

So again, this isn't anything I think that's happening immediately, I think this is a natural extension of search. It's not something really that should surprise SEOs that have been thinking longer-term. I mean, I was doing talks several years ago now talking about originally 10 blue links and comparing it to... For those that are old enough, when you used to go to a library and you had to type in the name of the book you wanted on the computer and it would tell you which shelf it's on, and you'd have to go find the shelf and then crawl around on your hands and knees, whatever, to find the book.

That's kind of what we've been doing with search engines and 10 blue links, right? We do a search and then Google or whoever just chucks us back a list of webpages. And then it's up to you to pick through those, find a page that you think is good, and then find the answer on that page. And over the years, we've taken lots of small steps with stuff like featured snippets with local pack results, with images in results, with top news stories in results, with videos and results where Google has been shortening, shortening that journey. So rather than list the website, it just shows you the video. Or rather than just listing the website, it shows you a featured snippet that answers your question. And we've seen intelligent personal assistance come into the home, smart TVs, and we are approaching that, I think, reality where we're moving from typing stuff into search engines and sort of filtering through results ourselves to the point where we just kind of shout at our TV or just shout into the abyss of our house to all the electronic devices that are listening and then get the answer.

So I'm not massively surprised. And again, long-term digital strategy-wise, search strategy-wise, we've been telling all of our clients it's not just about keywords and webpages. It's about establishing this digital footprint across different devices and being there and having that content and having that schema, having that structured data so you are prepared for when this... I call it maybe a different type of indexing or different type of search model happens. And that's quite interesting as well in terms of the search model. There's a really interesting diagram on Search Engine Journal about this unified model that they've pulled out the paper, which compares the current retrieve then rank model, which is this index where they retrieve answers and then rank them and generate results, whereas this unified model is just the query goes into the model and then you get the results. There's a nice picture we'll put in the show notes to demonstrate that.

But I think if you're not thinking about that model now, and by that model, I mean, that changed behaviour of how people are using search technology and you're still just thinking about web pages and keywords, it's definitely time to start planning for that before it's too late because those things are going to change. We've had over a decade notice of Google saying it's moving from becoming a search engine to a knowledge engine/answers engine, some people call it. I don't think it will be the whole part of search, but it's going to be a big part of the future.

So have a read through everything you find about MUM, and yeah, watch this space. And that's everything I've got time for in this episode. I'll be back, of course, in one week's time, which will be Monday, the 31st of May. As usual, if you're enjoying the podcast, please tell a friend about it. Do subscribe. It's on pretty much every platform that you can listen to podcasts on. And I hope you all have a really great week.

More from the blog