Series 6 Episode 8 - How and why SEO and Experimentation should work hand in hand

Speaker A: You're listening to the Big Lift, the podcast of WebTrends Optimize, the CRO solution that enables marketers and developers to maximize the ROI on their digital properties. WebTrends Optimize is a powerful, feature rich and easy to use solution, all delivered within a fixed price contract with no additional cost for increased functionality ever. During these podcasts, we meet some of the key influencers within the marketing and conversion world to understand their roles and examine their challenges. Today I'm in conversation with Luke McDermott, um, head of digital optimization at digital marketing agency Modo25. His expertise lies in SEO, UX and CRO, which gives him the perfect insight. Excuse the pun, on our topic today, why SEO and a B testing should work hand in hand. So, good afternoon, Luke. Let's kick this off by giving you a little background about you and your experience in digital marketing so far.

Speaker B: Yeah, good. Uh, thanks for having me, John. Um, uh, I'm Luke, header of Digital Optimization at MODO25. Um, I've been doing CRO UX experimentation for about 10 years now, actually. Started off originally in paid media and then moved over into, um, experimentation and then about five years ago started, uh, working in the SEO space as well and really trying to combine a B test, that type of thing with the SEO space, um, and just seeing how we can improve, uh, websites from, well, one in Google's eyes and obviously for customers as well.

Speaker A: And we're getting hopefully a better customer experience and making them a bit more sticky along the way. Is that right?

Speaker B: Yeah, that's the plan, at least.

Speaker A: The title we've got is how SEO and a B testing can work together. But one of the things that I'm, uh, cognizant of is that in larger organizations, A B testing and SEO are often handled by different people. So how can those organizations bring together the teams for a common goal? How does that work?

Speaker B: Yeah, so it's really a common thing, right? Like most, uh, AB testing, that type of things may mean product teams and then obviously SEO normally sits with performance, maybe marketing. So what we've, uh, started to do is really try and speak to both teams at the same time and really say, okay, well, how can we get the best bang for buck? I think historically a lot of SEO agencies haven't really done a lot of a B testing. And it's almost like a, uh, just do it and see what happens. So we try to come in and prove, obviously with experimentation, a B testing, that something can be of benefit. So I guess it's just about starting those Conversations and then trying to work on, um, okay, well, how can we all work as one team with one goal? Um, and again, there's no better way. I don't think there may be testing, putting a little bit of skin on the line in terms of, um, betting on what's going to work and what's not going to work and go from there.

Speaker A: So would you say that you would start an initial conversation with people at one level, say at the SEO, or would you do it together? So when you're doing your first introduction, do you gather the teams together and say, this is what we're trying to do a bit?

Speaker B: Both. Uh, so we do have, um, a lot of clients that we've worked on historically that are just, uh, CRO come to us to want to optimize their website. And I guess what we've started to do with that, uh, is go along the lines of not running out of tests, because we're never running out of tests and experiments. Ah, but starting to look at, okay, what else can we do that's the bigger picture than rather just the classic change the color of a button, change the structure of a website. What can we do that's actually going to get you the best bang for buck for optimization? And a lot of that is actually, can we optimize the website for a certain keyword, keyword set, make sure that we're getting the right customers to the website? Um, so we do it that way and we've kind of evolved our strategy from that perspective with the customers. But then also, it's a big, big part of our, like, uh, our SEO offering is actually. Okay, well, we can actually do these things as well that maybe a lot of SEO teams or agencies haven't in the past. So we kind of come at it from both angles. So there's not a one size fits all, but it's kind of. So the answer is it depends. But I suppose that's always the answer, isn't it?

Speaker A: I think it's the answer I get so many times with these conversations along the way. But SEO and A B testing are, um, if you like, two elements of a common goal, I suppose. But SEO, if I can remember correctly, SEO is very much with regard to getting the right keywords and things like that. And I know it's changed a lot with influence and author authenticity. But how do you manage to be able to start off? Is it starting with SEO and then testing it to see if it works? Is that how it puts together?

Speaker B: Um, again, it depends. Um, but I think what we always do is we always do an audit of the website, first and foremost. So, um, a real look at what's actually going on, what's good, what's bad. We also have a tool in house which basically looks at landing pages and looks at where the traffic's coming from and where ultimately we think there is value or additional value in testing. So we start there and then work out what's actually going on so we can then decide, okay, well, actually, is there some actual SEO that needs to be tested on that page to start with and then work things through from that perspective? But it always starts with the audit and working out, well, one, is there something to optimize or is there not for a landing page? And then obviously, whether that be SEO purposes or whatever, then go from there.

Speaker A: Do you think there's an element of what we used to say, calling your baby ugly by going into their website and saying, actually, you're not doing this right, you're not doing that right? How well is that received?

Speaker B: Um, again, I think it depends on who they are and what they want. I think nine times out of 10, when most people speak to us, um, they know they've got a problem. Or in your analogy, they know that their baby's a little bit ugly, needs a little bit of makeup, maybe to tidy it up a little bit. So I think most people who we speak to at least, um, know that there's a challenge or something they need to improve. And it's about them coming to us and saying, okay, well, we've got a large, large set test ideas, databases, where, like, we've, we've worked on this before, and I think that's where the value and what a lot of people see is in Modo 25. Like, we've done this a lot. Like, obviously it's about pulling the levers that work for you, whether that be starting off with a CRO campaign or whether that be an SEO campaign, whatever it might be, or even ppc, um, and working out what's the best strategy for you? And again, not being a one size fits all, but, yeah, like, I can understand why potentially some people wouldn't be too happy with you saying your website's rubbish.

Speaker A: I get the feeling that the answer to this next question is going to be it depends as well. But how can SEO, uh, and a B testing work hand in hand? What are the kind of the levers that you need to pull?

Speaker B: So I think the first thing that we, um, see is obviously the content angle. And I think that was the first Thing that we really went into and I started off and actually the reason I started looking at SEO itself was in a big agency beforehand where, um, big SEO agency and SEO was kind of influencing the website a little bit and it was the tail wagging the dog, so to speak. Um, and we were having huge, huge sections of copy that were just ruining the experience for customers. But obviously that's what Google needed. So we found out that roughly uh, 17% of uh, all words on websites are being actually read. Um, only 17. Only 17, correct. But obviously Google needs all of them because of that. Obviously we said, okay, well what can we do to not reduce them but also what can we do to move them around, manipulate the page to a B test and change what that looks like, whether that be hiding in accordions and such or uh, just even moving that content around on the page. And that's when we started to see a real, real good uplift in the value of testing from that perspective. And then that led onto multiple other areas of SEO that we started to see test.

Speaker A: And what's your kind of criteria for success if you like?

Speaker B: Um, again just, it's all statistical based. So what we kind of do is set up a structure of what we expect to happen from an experiment. Um, and then we'll work it out on two words. One, we run the actual a B test itself but then obviously what we'll do is track that within multiple of the tools as well, such as like Semrush, Ahrefs, that type of thing and obviously see and understand what the value of that that can be. We also do a lot of um, SEO testing where we'll almost build two pages of different structures, launch them at the same time and see which ranks better so we can start to see. So the success I guess is deemed on what we're actually trying to do. But ultimately the main thing or the core metric that we're looking for is more visibility to the page. Um, as well as obviously the classics of um, um, any CRO campaign. So your conversion rates, your revenue increases, your AOVs, that type of thing.

Speaker A: And you measure time on page as well.

Speaker B: Um, we do it again, I feel like I'm saying it depends so much now, but it depends on the client. Like we got, we've got everything from fast fashion right through to um, well with like super niche high end bespoke real uh, estate. So obviously it's a lot different uh, for that and even what we're testing. So ultimately like if we were looking at a blog for instance, we'd expect that time on page and things like that would be something that, a metric that we would share and report back to our clients. Whereas if it's selling a pair of shoes to somebody else, maybe we're not too bothered about that time on page.

Speaker A: So I can see you know, um, a B testing working very, very well on pay per click. Obviously you can see when somebody clicks on the page how long they stay on the page, where they click on from that point on. But SEO is quite generic by comparison to um, a landing page which specifically talks to where the customer came from, etc. Etc. How do you be able to utilize SEO and a B testing without kind of creating uh, an imbalance between either one's winning or the other's winning?

Speaker B: So yeah, I think like there's varying degrees. What we always look for again I think it comes into, well, I suppose the argument is anything ever winning? Yeah, um, but I think what we always look at again in comparison to PPC is we look at those KPIs that we look at for a page. So what we're always looking at is ultimately has the test been successful by those visibility metrics, by ah, those conversion metrics and work on from that perspective. Um, does that answer your question?

Speaker A: It does to a degree. But one of the things that I would suggest is user experience has a lot to do with pay per click. Is that the same with SEO? Does it refer back to Google to be able to say actually people were either longer on the page or they click some links or something like that? Because SEO seems to be a kind of a draw from your website outwards, but not necessarily, and I'm not sure if I'm explaining this correctly, but not necessarily something which is um, uh, how would you describe it a selling point? It's because you're trying to be able to provide information and education. Whereas a pay per click page obviously is trying to sell somebody to do something beyond that page and you can see that whether they actually click through or not. Is it the same for um, SEO? And, and how does that influence SEO in a kind of circular motion? Does it do that at all?

Speaker B: Yeah, so yeah. So ultimately depends on the keyword. So um, I think obviously like from what you're saying with pay per click, nine times out of ten, um, it's more shorter tail, M m more short and snappy. So it might be a black dress for instance that you're searching for and then you'll obviously pull through the landing page for that. So the principle is exactly the Same. If anything, obviously it's just more free traffic from an SEO perspective. So the principles are exactly the same. Um, again, what we've done a lot is testing actually quality scores with the PPC team and working out that actually, uh, the landing page element of that is the most important of the three components of quality score. And that ties into the whole SEO element as well. So is the page giving the right information, correct information? Is the website fast? Um, so the priorities are still the same for short, uh, shorter tail keywords. In terms of the way that we're working. I understand what you're saying in terms of a little bit more, uh, longer tail, it is a little bit more like upper funnel and like again, things like blogs, stuff like that. Um, yes, like maybe not as much of a drive in terms of what the conversion will be, but we also have worked on a lot of things with our clients in terms of structure of blogs and the way that they're presented so that they can actually lead to conversions and trying to commercialize the blogs. I mean, one of the things that we get quite a lot is people asking us that, like, how do I make money from my blog? And we see a lot of it is in terms of still traditional a B testing, but actually where do I put my banners promotions within that structure? Um, does that answer your question?

Speaker A: Yes, it does. One of the things is that, um, a lot of the things, at least in my day when I was doing SEO, was off page content, um, rather than on page content. You know, things like you're looking at external expertise, authority and trustworthiness. Eat content. How can you, if you can at all, I don't know if you can influence that sort of thing to be able to drive better responses.

Speaker B: Yeah. So again, like, one of the things that we've tested quite a lot around that is, um, particularly again on the blog posts, is things like authorship, that type of thing, removing authors, adding authors into it, like, and improving or seeing what the value is, I can tell you, which will be no surprise that having authors on blogs, more relevant, updated blogs, that type of thing actually does improve. And Google sees that as a signal. And again, we've done lots of a B testing with that that I'm happy to have various degrees of conversations with people about. Um, but yeah, that's one of the things that we've seen is around that expertise and authority, people want to know that there's somebody else there that has been there, ah, reviewed it, that type of thing. So we have seen a lot of values in terms of that again, always the classics in terms of headlines and such as. Well in terms of uh, reviews, that type of thing as well. Like Google one likes them. But also that's something as well that a lot of customers like as well. So that's something we can have a B tested quite successfully around that Eat content equally as well at the same time with that. Uh, um, you can, well we can do that off page as well. Um, but I feel like we might talk about off page as well a little bit more.

Speaker A: I was just about to say what about the things like keywords, meta descriptions and title tags?

Speaker B: Yeah, so that's the other thing that we have a um, regimented process that we work through with our clients when we first start to look through. I mean it's no secret that everyone in the last six months um, because of AI overviews and such has probably seen a drop in clicks in their search consoles account, that type of thing. So one of the things that we really want to do is try to maximize click through rate. It's a real simple metric alongside conversion rate that if we can get eyes onto the page then we've done half the job. So we do a lot of testing of meta descriptions, page titles, ah, H1s, H2s, all of that as well to optimize the landing page and see what actually works and improves obviously rankings.

Speaker A: And is that an easy task to do or is that more complex than actually doing the on page content?

Speaker B: M. I, um, think it depends on who you asked in my team. Um, I think we definitely, I think that's it's the easiest to track I think in terms of from a performance perspective of value. I think there is a little bit of looking at different reference points as well with the on page content side of things and again which I think we're going to come on to in a second. But like actually the actual engine or search engine that we're actually looking at. Um, so yeah, as always it depends but I think it's easier to track the meta descriptions and such in terms of the performance and increases.

Speaker A: We kind of, when we're in our previous discussion we talked about more and more organizations are becoming more reliant on AI. That must be I suppose great in one sense because you can get content quicker but providing relevant content into the mix around SEO as well, I'm sure that that's a bit more challenging. So how does that work with you and being able to introduce content via AI?

Speaker B: So we do use uh, some elements of AI for our content. Um, again it depends on who the client is and what we're actually writing about. So we found historically uh, that um, or currently even that where like certain topics are better to write on than others. So when it's things like uh, product recommendations, that type of thing, maybe reviews of like the best five car seats in the industry that's really, really good for AI to write with us just dropping in. These are the, the products that we want to potentially recommend or things like that. Um whereas when it's things a little bit more longer form and a bit more detailed experienced uh again we've got some clients in the holiday space. Um, I don' would it be class of the holiday space but you take my point and it's a little bit more tailored in terms of an AI machine we found can't really tell you how a lake looks or feels or that type of thing. Um so it depends. We are a B testing or we a B test content and LLMs uh quite consistently. So the main three that we use are uh, ChatGPT, Gemini and um, Claude um as well as our own content writers in house currently. What I can say which m I feel the content writers are happy with, um, but they are beating the AI engines to actually rank um, particularly uh, with those search engines and Google still. So it does look like human content is still the way forward. Currently however we are seeing that over the last even month, two months like the scale of how well those um, blogs or pages that are being written by AI are starting to rank uh better than they have done before. So humans are still in charge or that's at least what we've seen. But it doesn't look like it might be for too much longer before the uh, machines can uh, take over or compete with us.

Speaker A: Because if you use two different AI tools and ask the same set of questions you invariably get different answers. Somewhat different. They won't be completely different but how it it's okay in say moda25 where you're actually using 2 or 3 so you can aggregate the results between 3 but generally speaking if you like, people who are doing it by themselves are probably only going to use a single AI tool because they're looking for fast and efficient use. So how do you get your clients once they've if they've stopped being ah a moda client to be able to continue to use AI but know they're getting the best bang for their buck from the AI engine they've did they just follow one route?

Speaker B: So there's two ways we're Actually, um, training, um, some of our clients on this. The first one is actually the way that you actually behave with the actual engine that you're using, whatever that might be. Um, you're totally right that some are better uh, than others for certain things. So we've seen that Gemini is quite good for like uh, data analysis and research, whereas things like Claude and ChatGPT are a little bit better or look like they're a bit better for blog content and that content, uh, content creation. So there's that element of it, but then it's about what you're feeding into it. So there's almost like the uh, reverse questioning which we've seen is quite successful. So almost like asking the search engine to ask you questions around the topic and area and then feed in things like pretending as well around them being, ah, an SEO expert for instance. So uh, they'll pretend to be an SEO expert, they'll ask you 10 questions on uh, what keywords, different varying degrees of things that they'll want which ultimately then lead to a better result. Um, and again we've done lots of experiments in terms of actually those questions. You know, is it 10, is it 12, is it 15? Obviously it feels like, or from what we've seen, the more questions in that reverse questioning that you get, ah, the search engine asks you the better. But ultimately, do you want to be satisfied, um, answering questions from ChatGPT all afternoon? Probably not. So normally we start with a minimum of 10, uh, and then go from there depending on what we're actually writing.

Speaker A: Have you got an example that you could use without naming names of how you've actually used that kind of questioning technique? Say if it's a holiday company or something like that, what are the kind of questions you'd ask them to be able to provide you with the questions?

Speaker B: Yeah, so we've got ah, um, a client in the cycling space, um, and basically we wanted to write some content for them. So what we do is say I want to uh, uh, uh, ask me 10 questions in the cycling space around the top 10 products. I want you to behave like an SEO professional. Um, and I want the content to be around 800 to 1,000 words. And then the machine, well, chatgpt in that case would then ask you questions around what are the keywords that you particularly are ah, interested in? Are there any other competitors that have similar blogs that would um, that are of interest to you? What tone do you want to use? So obviously just like, well, have a conversational um, chat with the LLM M And then it answers that and then comes back. And what we've seen is that gives you a far better result, particularly because it's giving you better information than just asking. I want a blog on cycling for.

Speaker A: But these, these LLMs, presumably you're using generic LLMs. You're not, you haven't got your own um, AI engine or anything like that.

Speaker B: No, no, no.

Speaker A: Surely you're just feeding the Internet with

Speaker B: potential competitive information so you can close it off. So what you can do is, so there's no personal information that in that example that I just gave, but you can actually close it off itself. So you can, I think that's only on pro versions actually. But you can close that off so that any information that you wanted, uh, you could just make sure that that wasn't available for ChatGPT or whatever alike. So it gives you that security in itself. One of the other things as well that we have done is we actually have an account for each of our clients so that it doesn't um, cross reference anything. Uh, um, because that's one of the things that it does like to do quite a lot. Particularly ChatGPT is hallucinate in terms of almost picking up things, previous conversations. So we always make sure that our clients are all regimented in one account so that it's not talking about holidays and cycling, which from time to time can be useful but not, not always.

Speaker A: So it's interesting you, you say you ask questions or you ask uh, uh, an LLM to be able to provide you with questions which you can then answer in the, I won't say the early days, but certainly a few years after Google came on board, the use of long tail questions gave better answers for SEO. Does that work in the same space as you say in um, using ChatGPT or any one of the other LLMs?

Speaker B: Yeah, yes, absolutely. So what we've seen is particularly with AI overviews, that type of thing, um, a lot of the engagement is particularly from that informational content angle and those longer tail questions a lot more people asking about um, well, longer tail solutions than what before. So what we're seeing as well is that that's, that is the content as well that, that we're seeing the best bang for buck in terms of getting uh, engagement for one of the things that we also are tracking now is our uh, how, how many searches, queries, terms are people being shown up for in chat, GPT and such. And all of them are like you said, those longer tail questions about you know, what is the best car Seat for me, what's the best road bike for me? That type of thing.

Speaker A: Yeah, there seems to be a lot of that on the, on the Internet at the moment. Isn't there, um, anything you're looking for? Here's the top 10 and hey presto, there's links to Amazon in nearly every one of them.

Speaker B: Yeah, exactly.

Speaker A: To be able to do so, it's probably a little biased, to say the least. So you've, you're using this collaboration of SEO, A B testing and LLMs or artificial, um, intelligence. How often should you be checking back the effectiveness of these things for your SEO? Is it something that goes on every week, every month, or is it something that you could say actually is a continuous assessment?

Speaker B: So we're very fortunate to have our own bespoke reporting tool called AskBosco. And what that basically does is pull in all of the information from GA4SEMrush, whatever tool you're using from an SEO perspective, and pulls that in. So we look at it from two different angles, the first one of which is a, uh, landing page. So we're actually uh, religiously checking or. Well, it gets pulled through to us weekly to see what that performance is looking like. So is it improving, is it decreasing, is it, is it whatever's happening that obviously then ties into when we make changes. So like, what the team will do is have change logs. So we'll know that if we change the headline, it got changed on the 1st of the month and we've seen the impressions increase, decrease. So it depends on sample size, obviously, but like, we're always checking. I'd say, well, we have the data available daily, but we uh, but weekly is what other team are checking in on, mainly depending on the sample. And then obviously we work on what's actually happening, happened from that result.

Speaker A: So do you think it's imperative that, um, SEO and A B testing are, uh, working hand in hand, or do you think it's something that people would like to get towards?

Speaker B: It's funny. So I would say two years ago we did little things, A B testing, maybe a little bit longer, actually, maybe like three, four years ago. Whereas, uh, a B testing in this space, whereas I would say now, uh, it's quite close to 40% of all our tests are, uh, related towards SEO, uh, testing. I'd also say maybe an extra further, like 10% is around PPC. So again, the classics of uh, you know, what you test with them and PPC test and improving quality scores and then the rest obviously is just actual, just stereotypical a B testing. So I don't think it's a required thing, but I think it makes life so much easier. And just again, the whole principle of that, uh, that's why I was so interested when I first started off in marketing in the CRO space of it's concrete evidence of something that's actually happened and you know that this was the, you know, the right decision. So I think doing it in way that we've done from an SEO team as one made us quite successful, but also two, like just gives you a peace of mind that you're doing the right thing. And it's not just all right, we think this content is going to be the right thing, like we know it is.

Speaker A: Well, I think that actually that's quite a good summary because one of the things that we mentioned throughout this podcast is you've used the phrase it depends and testing is the only way to be able to say, well, it depends on what and you make those changes. So I think from my perspective, testing whatever it is, whether it's um, SEO or whether it's testing landing pages, until you've tested it, you can't prove that actually what you're doing is effective or not. So I think it's pretty good. Before we close off, is there anything else that we've not covered that you'd like to cover?

Speaker B: No, I think the only thing as well is just that with the off page side of things is just links into the page. Again, something that we've seen that's been hugely successful is actually monitoring the performance of Anchorage anchor text and links into the website. And that's just the only other thing that I think we probably haven't touched on. We touched on the content side and then as well, obviously page speed, technical SEO as well. Obviously. It goes without saying, but the faster your website is, uh, nine times out of ten, the better the conversion rate is going to be. But I think they're obvious things to cover. But I would definitely encourage anyone to really jump into it and start to a B test. Um, their SEO proposition as well as uh, any other, any of the other channels as well. You know, it goes into, we're doing ppc, we do affiliates as well. Um, like really a B testnet paid social as well is another great one. So, um, I would definitely say it's, it's got massive value and I'd encourage anyone to do it more.

Speaker A: Well, Luke, thank you very much for your time. It's been a subject which I'm sure is going to evolve over time, because it's one of those things that AI is going to have more of an influence over us. Ah. As we go forward anyway. But I think that the ability to test everything, including SEO, is one of those things which actually is going to hopefully, um, show real benefits in the long term. But for now, Luke, thank you very much for your time, and we'll speak again soon, I'm sure.

Speaker B: Thanks, John. See you later.