 This is so exciting. I love MozCon. It's my favorite conference of the year. It's just so cool to be here Six years ago. I went to MozCon local and I sat in the audience way back pretty hungover Looking up at the stage and said I want to do that Dream big right. I wanted to say hi to everybody here in the audience and everybody on the internet because they can't be with us today It's just so amazing to be here and I'm just really stoked to jump in It's gonna be a deep dive in a topic maps and we're also gonna talk a lot about innovation work I pitched this idea thinking that I'd come and show you everything that you need to know about building topic maps TLDR, I'm not done yet as Tom Kapper said who would have thunk that it would take more time, right? So let's jump in Doesn't it feel like vacation here in lovely Seattle walking around sun light breeze Hanging out with our Canadian friends telling me just how hot it is when back home in Colorado It's 102 degrees So this story and this journey for me began on January 4th on the lovely island of Oahu I was sitting there taking a shower and I came up with this amazing idea for building a killer product all around topic maps I had this vision of a force-directed tree You've probably seen that kind of tool in sight bulb or screaming frog and I thought this will be amazing And that was the beginning of the journey which a lot of times I think of innovation work is being a journey into the great Unknown and that's really what we're talking about so Vacation for me is reading catching up on all kinds of articles I read this killer article by my pal Dan leapson, which is all about how local SEO guide Earns their knowledge and it's an amazing article. I've got a great resource web page which you'll see the link in a couple minutes and I strongly encourage you to read that article and this deck is my Journey to earn knowledge around topic clustering and part of this story is not just about innovation work And it's not about topics It's also about careers and I'm gonna touch on that and settle along the way in a couple moments Because I think it's teachable stuff for anybody doing any kind of work in SEO whether you do content Whether you do technical whether you do e-com. There's lessons for everybody along the way. Okay, so I'm in Hawaii and I think to myself What a great year 2021 was we built this product called Explorer It's amazing because it allows you to start with your Google search console data And it allows you to store 50,000 rows of data per day But the the really challenging part about it is what if we're on a website where we don't have any Kind of keyword footprint in a new topic area or what if I as an SEO have no experience working in an area? It was clear to me that we'd only solve some of the problems, right? So that's why we're digging into topic topic stuff So I started to think about topic clustering and the technologies behind it And I got really really stoked and I started to talk to a lot of my close friends Dave Sotomano Robin Lorde John Merch all the guys who are building and also some Just all kinds of people are building just the coolest stuff on the cutting edge of SEO because when you're starting these journeys I always like to talk to these folks because they have Done a lot of the hard work already and there's a lot of opportunities for me to save pain along the way and I knew that it was critically important because building content as We all know is hard and it takes a lot of time and it's really really tough Especially when you have no data. So what are we really talking about? We're talking about topics You can get a sense here Topics are the things that are made up of lots and lots of keywords, right? A topic is the subject It's the thing that we're talking about Whereas keywords are kind of like the important words that we use to talk about the subject the topic is the thing That's the thing that I was really really intrigued by Grab this bit Lee I built a whole bunch of free tools and you get to play with them and I really really want your feedback In my Twitter handles there. So please at me or DM me. Okay. So here's what we're gonna be talking about We're gonna be talking about client goals first. I Hope in your onboarding that you ask lots of great questions like this Who's your ideal customer? What services do you offer? What is most profitable? We want to market the things that make people money, right? What's a conversion worth? We need to know every single time when they close a deal how much money they're gonna make how? When where do you make your money? What questions are tied to conversions? What are those those kind of like hinge points that make people say yes or no? We also want to know and this I got to call out my friend Renee Bigelow She said this to me once and it just blew my mind because it was like I thought about content from a traffic perspective And I know publishers they have very different problems than what we're attacking today But really it's like when you're getting started. It's talking about content to support your sales process Gathering of the data. This is what gets me super super excited because I love building tools, right? And here's my first principle When I was thinking about building out topic maps The first thing that I thought was if I can build topic the way that Google understands knowledge Then my content is more likely to rank than not so what do we actually know? We don't know what Google knows we can't see inside the black box So all we can know is what they show us and they show us information or how things are related In a couple of different places that are really cool, and we're gonna stop on some of them and go really deep into Google Trends So they tell us lots of information through the knowledge graph search API One of the free tools that you can play with is a way to actually interact with the knowledge graph search API Which I think is really fun Great information in the SERPs as we all know Autocomplete which is really powerful, but I'll give some reasons why it didn't settle there as we go Keyword planner data, which I am not gonna do a lot on today And that's gonna be an area of deep research for me throughout the summer and fall and then Google Trends which Really rocked my world Okay, so first knowledge graph search API Super cool You can type in any kind of search term and then Google will respond the tool that I built It'll show you all of the things that are related to your head term in this case It was bicycle and then it'll show you information in that article body Column, and it'll also show you what the source of data is This is why it's important to go deep and to look where your data is coming from Because you'll start to see things like wiki data and wikipedia Making up an enormous amount of Google's knowledge, right? You're also gonna see things that you're gonna have to pay attention for because you'll have to disambiguate Ambiguate against them later and that's like to make make it clear for Google so that they understand what you're writing about and what's maybe Potentially related, but that we don't want to write about like if I'm writing about bikes. I don't want to write about motorcycles Okay, wikipedia. Why is it important? wikipedia makes up 67% of the trusted sources for knowledge panels this month if I want to know things and I think Google is tapping into like different sources for data. I'm probably gonna want to start in wikipedia So I went deep into wikipedia first couple things that jump off the page. Look at all the blue, right? Those are all of the related topics to the term on the page. I was like That's so cool. Look at the top of the page This is really neat too because there's a section called disambiguation on every single page in wikipedia They have links to another page that tells search engines and users What we are talking about which is bicycle a pedal driven two-wheel vehicle, but we're not talking about Bicycle crunch. We're not talking about bicycle kicks and we're not talking about a playing card corporation At the bottom of the page, you're gonna see these really cool purple accordions And if you were to open those up, you'll see all kinds of really cool things that are all related to your topic There's a link in the resource deck for Google sheets add-on So you can get all of the inbound links outbound links Mutual links between wikipedia pages and wiki data. It's cool. It's really cool. Install that put it in Google sheets You'll love it okay Tons of information in the service right We've got related searches. We've got people also ask. I like to put a search into Google on a desktop Scroll to the bottom of the page Click in and then I get this cool view that I can screenshot. I also got related entities in the bottom right. That's pretty cool When I started to think about building a topic clustering tool I assumed I was gonna use the search results pages as my main data source and a lot of People who are building tools like keyword insights, for example, they're using they're using the SERPs as their data source But as we'll get into that's a lot of work That's why I got really into Google trends autocomplete super powerful I think we all know what it is you type in and it gives you up to 10 results The thing that jumped off the page at me when I started to play with this and tools that are built on top of it like keyword Sheeter Is that the data is super noisy and it's also super localized if you're working in local SEO And you want to do keyword research at scale And you want tons of ideation I think autocomplete is super interesting if you want to build knowledge around an entire keyword space Probably you'll find that the data is too noisy to work with this is what really jumped off the page at me Just how personalized it is based on the language location and past user history and look at the bottom How Google in their own documentation is comparing autocomplete to Google trends and when I saw that I started to get really giddy Because I love Google trends and I love building tools and last year I built a Google Trends API So I could play with Google Trends data and I built it because I was like really wanted to build something for Lily Ray so that she could do stuff with news data And so I had this tool on my shelf sort of like a hammer that I was just like waiting to grab and start whacking stuff Okay, so let's see just how tough working with manual is right? Let's say we enter a query you can update the region from country to region or state or city or DMA You can change your time range and also the category And then you could like download the related topics and then you could add those to Google Sheet and Man that would stink wouldn't it have to like try and do something at scale So for me I got really excited because I knew that I could get that data via API's if you're like dorky And you like using Python you You could use something called PyTrends and If you are into JavaScript, you could use something called Google Trends API Or if you like Apps Script, you could use a data provider like data for SEO to get this kind of information And I subscribe when I'm building products to a philosophy called make it work Make it right and make it fast and I got that from my pal John Merch. I think it's super super cool So when I'm building tools I always start with the Google Sheets interface And then I use Apps Script to get rolling and then if I have to get more complex I do that in this interface you can see how we can put the bicycle as a search term the dates The categories you could see how there's a city whitelist that column it's column K That means that there's also a city exclusion list Which means I'm excluding all of the cities in the United States that have at least a thousand residents So I get rid of any kind of localization in my data And it's all just like going deep deep deep in a bicycles the L column is stop words That's like big box stores Some brands some city names disambiguation remember when we looked at at Wikipedia we saw Bicycle corporation and motorized bicycles well I might want to put in mopeds motor, you know all kinds of stuff that I know is going to take my tool off off off course, okay So with Google Trends there's a bunch of different ways of getting data out of the API You can get interest over time in almost every SEO deck that you've seen ever It's almost always interest over time. That's that time series chart. You can also get interest by location You can also get daily trends. You can also get Related trends related topics and related queries. You would think since this is a thing about Topic clusters that we'd be working with related topics, but through a lot of experimentation. I came to the Belief that related topics was too noisy What do I mean by too noisy like lots of people? Brands corporations it didn't tell me about bikes But it told me about all kinds of stuff that would take me off course and out of my main topic So I settled on related queries. Here's what we're gonna do. We're gonna get data We're gonna use We're gonna personally monitor it, you know, human intervention or human curation during the make it work phase And we're going to remove irrelevant terms and you'll see why that's important in just a second And then we enrich the data with You know another API that might give us things like search volume and it might give us cost per click and You also can get intent Which is really really cool because you can use it to filter your data down the road, right? And then you turn that process multiple multiple times all right human curation is really important And here's why let's say we start with bicycle. I love this. This is my world. I know all about this keyword space Check it out. Look at about 12 o'clock. Do you see how it says Trek bicycle? You'll also notice this whole way of visualizing the data is called a radial tree cluster And it's from the D3 library part of going deep is learning about all the Visualization libraries that I could use to potentially tell a story with the data because down the road end Users are gonna have to look at this and say like oh, what does this mean? What does this mean? So radial tree cluster was really interesting at first But like after two cycles this data is getting really useful Oh, the reason why Trek is important here is that let's say I feed Trek in The next turn of the cycle could very well turn into Star Trek and then the next one after that is Jean-Luc Picard And now we're not talking about bicycles anymore. So that's why human curation early is really important So after two turns of the cycle It's really interesting right we're up to a hundred and thirty nine search terms This whole process has taken me three minutes to build in terms of like running the tool The visualization is really cool. It's clustering. I get to see how things are related to each other I've got search volumes. It's pretty interesting one more turn check this out. We're around a thousand search terms Do you see how the visualization is becoming less efficient? So this is what you learn when you go deep you learn What you can do and what you can't do right so within 15 minutes of Doing that manual curation in running the cycle a whole bunch of times you get to build something completely. Oh my god That's awful right Pump your brakes yo 4000 search terms equals messy messy messy data viz. I could never do analysis that way that would totally bum me out That would be a total fail for a tool Which takes us into analysis. This is when we've gathered all our 4000 terms and we need to like make lots of decisions This is a force-directed tree This is what I envisioned building. I was standing in the shower and I was like oh the different sizes of the nodes It'll be meaningful the different colors. It'll be meaningful People will be able to make decisions. It'll be awesome 10 15 20 hours of experimentation Big hot steaming pile. Oh, couldn't get it to work Circle packing. This is another D3 library Much closer really interesting. I can get a sense of scale. I can see big topics smaller topics I get search volume Right really interesting. There's different ways of interacting with the data So you could click into one thing and it'll zoom right into that, which is really interesting wasn't quite there yet This though Rocked my world. Okay brief segue into careers so I Love hanging out with SEOs and I love talking to really smart people because I know that there are things that I don't know right and I talked to probably 50 SEOs before Moscon to show them this because I was so excited about it And if you do one thing in that resource library look for the Noah's Sunburst tool And please follow the link and play with it and please give me feedback because I think it's almost Addictively fun to play with And down below there's there's like a table and what's really cool about this is that as the user clicks into a segment It zooms in so like on the left if the user had clicked Trek bicycles it would have zoomed into the trek bicycle term the size of the arc tells you how much search volume is Downstream that's related to that topic So for that term bicycle in the center the stuff in the inner ring or all of the search terms That are related to it and I got really really interested in this because remember I showed you how we went from 1 to 22 to 139 that was like five or six different API calls that is such a simple technical solution and To me that was like really really really interesting Here's a question. Let's say you're talking to someone who's not technical Let's say you're talking to decision-makers in a meeting Do you think you could get buy-in with a really visual tool like that and you could play with it and interact with the data? Or let's say you're in a sales meeting and you're just you walk in and you're showing people their entire keyword universe Do you think that that would blow them away potentially? Maybe maybe not. I think it's pretty cool Okay, what I found building tools is that there seems to be two or many more flavors of how SEOs interact with data I like spreadsheets, you know Paul Shapiro sees music when he looks at a spreadsheet David Mim sees music when he looks at a spreadsheet, but there's a whole cast of characters who when they look at that They see things just swimming around the page and that user group when I showed them the sunbursts were like Oh my god, it was like Joker smile. It totally worked So most SEOs might prefer something like this, especially if it has Moz colors This is a digital studio template that I built for this. I thought it was really fun Okay, so what are we seeing? We're gonna jump in in a second. I'll I'll go deeper into this But okay, so clustering methods a couple things How do we do it how do we cluster things you could do it manually I think the goal for me anyway was to get away from that as quickly as possible One use that's really interesting though as keyword planner that interface is really really cool and very quickly You can filter down. Let's say Well Nico is a guy that I work with he's amazing totally brilliant super analytical a great PPC mind What he'll do is put throw one to ten search terms into keyword planner comes back with one to 10,000 different terms and he uses the filtering mechanism over on the right and Then he'll add each of those different sets of terms into their own ad groups download the campaigns delegate the workout to the team Boom topic maps. I thought that's pretty cool Serp similarity is the gold standard right now That's when you let's say you have a bucket of 10,000 keywords You feed each one into a data provider like data for SEO or grep words something like that And it'll come back with a JSON response of everything that's happening on that page So the top 100 the related all of the SERP features etc And then the way that we're getting to our clusters is you basically compare the results of the different terms and Then you very quickly have a sense of how they should be clustered based on the similarity of the responses The engineering challenges are significant, which is why I didn't finish the tool And the costs are significant if you have 10,000 different things you have to make 10,000 API calls and then you have to store all that data process all that data analyze all that data And it has to be a really robust system when you have 10,000 API calls You can also use machine learning. This is a resource that I kind of updated a little bit. That's that's in the guide also Super cool really fun This is how it works. You feed it a CSV a search data. It'll then come up with clusters It's about 60 to 70 percent of the way there We felt like after we did a lot of experimentation that it was close not quite But if you modified it in certain ways we could get over 90 percent of the data to cluster But I like Occam's razor approaches. This is where we get into career stuff again I love leaning into folks's lessons that they're willing to share with me It saves me hours and hours and hours of time if you don't have a board of advisors for whatever your superpower is Go get one get a network work your tail off to build a network I'm super stoked to be in your network at me on Twitter. Come say hi. I'll do a zoom with you Elias Dabas. He is like One of the coolest guys working in SEO He built the advert tools library We had a couple meetings and we looked at how we could potentially cluster the data using regular expressions instead so it's like much simpler than doing lots of machine learning and We came up with the solution whereby you could take all the search data Come up with all these predefined buckets remember how I said we could get intent via API Well, this gets us down into sub intent, which I think is pretty cool bottom line. You got to find your own way for me Data analysis equals data studio, right? Let's get back to that template. Let's get through it So clusters on the left driven by Lee foots tool Intent oh, sorry clusters on the left and then we get all the queries in each cluster I got search volume that came via API and Then I also have intent via API Sub-intent via that like regular expressions driven tool and Then we can do all kinds of cool analysis on our data, right? Like that whole tool was focused on getting trends data Well, what if we compare our trends data versus the 300 or 700 thousand search terms That we're storing up in our data warehouse. You guys are storing your search console data, right? No You got to there's so much data there anyway, so you can compare the two The sequel to perform this analysis is in the resource deck So you can do it yourself and what you're gonna find is that there are gaps even if you only have 4,000 and trends and 400,000 in search console, which means that there's plenty of Opportunities for you doing that analysis is so easy in a tool called count If you don't know sequel and you want a really gentle way to get into it. I think count is super super fun You can also compare trends and search console to find that overlap And that means that our templates gonna change a little bit to do analysis all of a sudden we get position bucketing Right, which is super useful so you can find low-hanging fruit stuff. I'm on the first page stuff that's aspirational Talking with JR Oaks about all this stuff that I was building. He gave me the aspirational thing. I think it's awesome Okay, creating topics real easy right You basically let's get back to our client goals for a second We need to have that as a filter to think about this next step. Everything has to align with our client goals Things have to have a high search volume right so that there's a meaningful opportunity. I Know publishers have a different goal. They just want their ad revenue But for most of us we want conversion intent and We have to have a reasonable chance of ranking for things and That's the process We also want to go back to our sunburst right for ideas because then We can think about how to build our internal links that map the same way that our topics are related Via Google Trends, which I thought was really cool. It's a neat neat kind of hack Okay, so what's next then we get into our normal kind of like publishing flow build content briefs using all kinds of killer tools Whether it's content harmony Market Muse clear scope whatever Publish your content and then you get to use search console data to refresh and optimize Cool. You got to earn it. You got to go deep Uh Whatever it is that you do whatever your superpower is This is something I've been thinking a lot about like I spoke at search level a couple weeks ago and I Was really meticulous about masking There are three of us there at that conference who had masks on and I got cove it And it was a super bummer and I was lying in bed for 10 days And I was thinking about my deck and I was thinking about my deck and thinking about my deck And I was really concerned that I'd get up on stage here I'm just cough my ass off in front of you guys and I started to think about what's really really important really really important and I thought a lot about my friend Hamlet Batista who passed away last year from COVID and The last chat that I had with him. I asked him like hey What what was the most meaningful thing that turns your career and for him? He didn't hesitate at all It was give give give give give give give give give give if you have an opportunity to do that Go above and beyond you will be shocked how fast the karmic wheel spins In your direction Be curious Ask why ask what if and then test it Um Guys this has been a hoot. I don't want to like give you too much career advice or anything like that It's just I Didn't know if I'd ever have a chance to do this again You know to be in a stage like this and to try and share what's been meaningful to me because this whole concept of giving and Trying and growing a network being bold dreaming big and believing in yourself Has totally propelled my career and I think it's totally repeatable and they if you kind of embrace that mentality of Openness and caring and giving you'll be shocked how powerful it is. Thanks everybody. I really really appreciate it