How do you incorporate machine learning into your SEO strategy?
That's what we're discussing today with a man who's been a search account director for Ogilvy and SEO director for Havas Media Group. He is currently a Web three CX advisor. A warm welcome to the In Search SEO podcast Si Shangase. In this episode, Si shares six machine learning steps to creating successful SEO strategies, including:
Signing up for a Google Cloud account
Creating a Google Cloud platform using the Search Console API
Using BigQuery, create a schema for your data set
Get access to Forecast Forge for your business
Run your data set at regular day intervals
Connect to Bigquery with Looker Data Studio
Si: Hello, David, how are you doing?
D: Hello, Si, I'm very well. Thank you. How are you doing?
S: Doing well. It's good to see you again. I always love having these chats with you.
D: Absolutely. Always enjoy these conversations. You can find Si over at KuduHQ. Tell us about Hudu.
S: Hudu HQ is a way to support people or founders to drive new users to your projects, and to also help them fundraise. The biggest issues or the biggest pain points that founders have is essentially either outreaching or getting their project to market. And the best way to do that is obviously through Google SEO. And CX means being a customer experience advisor. So what do we do when people get to the website, we do all the fun things that SEOs love to do content, tech, and outreach.
D: So today, we're zeroing in on machine learning. Should every SEO be using machine learning?
S: I think so because it's probably the best way to actually make more logical, informed decisions. A lot of the time, and as an SEO I'm very guilty of this because I got a lot of experience, but sometimes, it's all about the right experience, for instance. When you use some of the six things, you can essentially use machine learning to bring your data back, and analyze it at scale. And that's the most important thing. Having a lot of information, and a lot of scalabilities is going to allow you to actually see different scenarios from an output perspective. That is a good place to be in.
D: Great. We'll dive into six steps, just in a second, I just want to clarify exactly what you meant by something else. You were talking about using it to drive predictive SEO strategies. What do you mean by that?
S: Essentially, with predictive SEO, you have different input variables. For instance, what we're going to talk about today is when you analyze on Search Console, looking at clicks looking at impressions. What you can do as well with machine learning is you can have a look at content, different keywords, monthly search volume, and seasonality. And with the keyword research that you do and the type of content that you're going to create, you can actually start to predict how much of those clicks or those impressions are going to basically translate into clicks for your business, and eventually into sales. That's the best thing about this. You can actually use a lot of input variables, but at the same time, you can see what a constant strategy is going to do for you. And what a specific change in either a technical structure is also going to do for your website.
D: Understood. Essentially, you're using a model to demonstrate whether or not what you intend to do is likely to be successful or not.
S: That's correct. You want to see, for instance, if a specific change is going to have a positive impact or a negative impact. And what's that going to look like? If you're a business, you need to understand what's the business case to make the change. And that can allow you to actually put a monetary value to the work that you are going to implement. And that's obviously a good thing for all product managers who want to know why should they make this change from an SEO perspective. What's the bottom line of the business?
That's what we're trying to get as close to that as possible. It's not always accurate. But the accuracy with some of the machine learning applications is you get as close to 86% confidence. At least you know that from a confidence ratio you can share that with your leadership team to give some form of security in terms of how accurate that data is, or how accurate your plans would be. I think that's what we're trying to do here. We're trying to share a bit more visibility to other stakeholders within an organization of the importance of certain SEO activities or tasks.
D: Understood, so it's certainly a lot more accurate than a gut instinct or perhaps more than using your experience based on what I've done before. You're actually using data modeled upon your current circumstances, your current website, your current competitors, the content out there, and what is likely to be successful in the future. I understand why that's really essential to do. You're doing a great thing there.
So let's go into these six steps to using machine learning. Number one is to get a Google Cloud account.
1. Sign Up for a Google Cloud Account
S: Yeah, we're going to take it to the basics. The main thing is to get a Google Cloud account. My recommendation would be to use a Google Cloud account because a lot of people have used Google Search Console, so they're well aware of using ads, views, and Google Sheets. The reason why I would say to use a Google Cloud account is that it's integrated within that suite. And it's quite easy to actually plug in some add-ons into your Google Sheets. And you can also plug into what is now Looker Studio [previously Google Data Studio], and you can plug into that as well. So utilizing a Google Cloud account would be the best way to start out. And then as you increase, you can start to look at different other options, but that would be the first place to start.
D: And step number two, create a Google Cloud Platform using Search Console API.
2. Create a Google Cloud Platform Using Search Console API
S: This is what I was talking about in the first step. What you've got within the Google Cloud Platform is you can actually connect to the Search Console API. How do you get this? Well, the main thing is, for those of you who actually haven't done it, you can just Google search how to get Google Search Console API. And Google obviously writes very good blog posts around how you can actually pull that data from your Search Console account. Then you can plug that into the Cloud Platform console. That will be the first step would be to get on the platform by connecting the APIs together. That way, they all talk to each other.
D: And step number three, using BigQuery, Create a schema for your data set.
3. Using BigQuery, Create a Schema for Your Data Set
S: Yeah, a lot of SEOs do schema markup. So if you haven't, please give it a go. It's a lot of fun. There are tools out there that allow you to actually create schema markup for your data and web pages. I know it might not be something that's a high priority for you, but it's going to help you with this next step. What you're doing in this step is you're telling Google Cloud Platform and BigQuery which data you need. For instance, because you're going to be using a lot of Search Console data, the main things that we're going to need are dates, impressions, clicks, and CTRs.
So put the data in and understand what they are. For instance, when we look at impression data, it's numeric data. When we look at CTR data, it's float data. And the reason why you do that is that you have a decimal point in CTRs. So just learning a little bit more about schema and how it works is going to help you go a long way in this regard because the data that you actually mark up in the schema that you want to pull in is going to be the stuff that BigQuery is going to be pulling in from Search Console. And that's the data you're going to be warehousing for your analysis later down the line.
D: That's a perspective from a hardcore SEO. Use schema data because it's a lot of fun. That's the number one selling point.
S: Yeah, it's a lot of fun. Think you're going to love it.
D: On to step number four of getting started with machine learning is to get access to Forecast Forge for your business.
4. Get Access to Forecast Forge for Your Business
S: Yeah, this is a big one. Where does Forecast Forge come from? It's actually created by guys at Google. I was listening to an audio podcast and apparently, one of the guys who created this is a super-sized genius Ph.D. guy. And he's helped them create this machine learning. It is a predictive forecasting behemoth of a tool. And the great thing is, you don't have to have millions to have this at your doorstep. You can just use something called Forecast Forge. It only costs about 100 quid so it's not a lot of money. And it's something that you can plug into your Google Sheets. And the reason why you do that is what you've done is you've got Google Search Console, you've got your Cloud computer, which has BigQuery, and then you have Forecast Forge. Forecast Forge is actually an add-on that you can use within your Google Sheets which allows you to create the predictions of your datasets.
Let's go back, you’re warehousing all your data in BigQuery, you’re analyzing it using Forecast Forge, and then using Looker to plug into Google Sheets to visualize your data.
D: And step number five is to run your data set on regular day intervals.
5. Run Your Data Set oat Regular Day Intervals
S: Yes, I think this is an important one. You can probably miss this when you're looking into how to actually put this together. I'd say it's like going down a rabbit hole of how you can create the schema of your data. You should definitely run it in dailies. If you are a bigger business, you might want to write it every hour but this is obviously going to obviously take up a lot of your computer credits. So definitely run it on a regular daily interval, especially with Search Console. And that's just going to give you better accuracy, especially when you're doing the reporting that's plugged into your Looker Data Studio.
D: And lastly, we’re up to step six, connect to BigQuery with Looker Data Studio.
6. Connect to Bigquery with Looker Data Studio
S: Yeah, I think we sort of jumped to this one. The great thing about Google's suite of platforms is that you can connect them all together. And if you're using a different platform, you need a third-party provider to sort of basically plug into that. So within Google Data Studio, now Looker Data Studio, you can connect through the insights platform with the hardcore data that you pull in from Search Console that you're analyzing. It'll be the main last step. Once you've done that, it's basically playing around with how the data looks in terms of the insights that you're pulling in, and how it looks visually. So that's the last and final step. But I think it's always about playing and testing. That's the main key thing, to test and see how the data is being collected.
Pareto Pickle - Crawl Budget Optimization
D: Superb. Well, let's finish off with the Pareto Pickle. Pareto says that you can get 80% of your results from 20% of your efforts. What’s one SEO activity that you would recommend that provides incredible results for modest levels of effort?
S: I'd say crawl budget optimization. And the reason why I say that is, just optimize your EuroStoxx file from how Google crawls it. Think about your most important category pages or your most important pages. And you want Google to access or get to the pages that are most important to you as fast as possible. And you could do the rest later, which is obviously making sure the content is right. Look at the internal links and all that fun stuff. But the most important thing is just making sure when you do click that request indexing button in Search Console when Google comes in to actually crawl your website to your web pages, that it's getting the information that it needs as fast as possible.
D: I’ve been your host, David Bain. You can find Si over at hoodoo hq.com. Si, thanks so much for being on the In Search SEO podcast.
S: Thank you, David, was a pleasure.
D: And thank you for listening. Check out all the previous episodes and sign up for a free trial of the Rank Ranger platform over at rankranger.com.