Content Strategy: Everything You Need to Know About Topic Modeling

Keeping fresh content coming on a regular basis is obviously an important part of your online marketing process. But you can’t just throw anything out there and expect it to perform, whether your most important metrics are based around traffic or conversions. You need to create content that your audience actually cares about.

So how do we do that? How does a marketer identify the right topics to talk about—the things that their best possible customers are not only interested in reading or watching or listening to, but actively seeking out? The topic modeling process is one way to get there.

What is topic modeling?

Topic modeling is the process of analyzing the relationships between words and phrases—that is, keywords, keyword variants, and related topics—as they pertain to specific content topics. There are a number of different tools that can be used to figure out what those relationships are (and we’ll get to that, don’t worry), but whether you use some or all of the available processes, the end product is typically a topic cluster.

Start with a focus topic: for example, when we went through this process earlier this year with one of our clients, the main focus topic was composite bearings. That’s at the center of the topic cluster, and branching out from there are several related topics—think subheadings under the focus topic’s title. In this example, some of the topics related to composite bearings included types, benefits, applications, production, and materials. Then from there, each related topic had a number of secondary topics that branched off from there—like, some applications included automotive, agriculture, construction, and civil engineering.

Depending on the quality of your investigation and the detail you want to get down to, these can branch out further and further for quite a while. This is the end product of the topic modeling process. You get a web of subjects, from the most important in the center to the more esoteric out at the edges—and a deep knowledge of what your audience wants to learn about and how all those subjects can interrelate.

How are topic clusters created?

As you might guess, this isn’t simply something that one can pull out of thin air. The whole point is to make sure the content you’re creating is the content that your prospects and customers actually care about. No matter how smart a marketer is, that’s not something that can just be made up (although as we’ll see, marketer’s intuition does play a role). So how do you actually figure out what your audience is looking for?

Step 1: Talk to your audience

It may sound a bit simplistic, but yes, one of the best ways that you can determine what kinds of content your audience is looking for is simply to ask them. Especially for those of us who have been in the game for a while, it’s easy to get a bit of hubris about our audiences. “I know what I’m doing, I’ve been talking to these people for years and I have a pretty good idea of what they want from us.” But really, sometimes the deeper you are, the harder it is to really know for sure what people who aren’t on the inside think of you, what they’re looking for, and where they find you lacking.

That’s why brand and audience research are so important, and why it’s a process that you can’t just execute one time and then forget about forever. People might not be looking for the same things today that they were ten years ago, or five, or even last year. Things change. 2020 showed us just how quickly that can happen, and while you certainly don’t need to expect that level of change every year, you should not expect stagnation either.

Audience research can tell you a lot. First of all, it’ll help you determine just who your best-case audience actually is. You might assume that’s something that never changes, but as your business grows and prospers, the people you’re best suited to serve might not be the same people as when you were just starting out. That’s definitely something we’ve seen here over the years at Element Three, which is part of why we rebranded last fall. Once you have a better idea of who exactly you should be talking to, you’ll be able to find that person and, both through this general research process and by literally talking to people and asking them, you’ll know what they want to hear from you.

Step 2: Data and analysis

Once you have some ideas about what people want to learn about in your industry, it’s time to sort and analyze all you’ve learned. There are a couple of ways you can do this, and the first is to lean heavily on the technology and tools that are available to marketers.

Artificial intelligence, for example, can do a ton of the heavy lifting here. You’re likely aware that at this point, AI is not just a scary science fiction villain, it’s also a part of our modern technology landscape. AI can be a massive help as you construct your topic clusters because it can compile all the potential topics that you can cover and analyze their relationships itself. It can take all the factors into account and help you determine what’s most important, what is related but slightly less important, and what can be ignored. That’s a great foundation for effective topic modeling.

Keyword tools are also super useful, as they can show you what people are talking about in your industry—and, perhaps more importantly, what they aren’t talking about but should be covering. That’s where you will find the biggest opportunities, because if you’re the only one talking about something that your audience cares about, they’re obviously going to end up listening to you and you alone. But the one thing that keyword tools won’t do that AI can is analyze the number of co-occurrences between terms and their semantic relevance, which means they may not show you the degree to which these topics are related.

If you’re going to dig into the tech, a smart way to execute would probably be to concentrate on using keyword tools to identify what you want to talk about in general, and then engage AI to tier topics by importance. But there’s also another way you can do this, if you’re confident in your own smarts and skills.

Step 3: Trust your own instincts

The other way that you can analyze your topic modeling data is, simply, to do it yourself. Once you’ve spoken with your audience, take a look at what they’ve told you. Identify the main things that come up again and again, the most important pain points that they’re seeking to solve. Look at how they relate to each other and how they might connect. Determine what’s important, and determine what needs less attention.

Simply winging it isn’t the answer here, but that isn’t the same as making educated decisions based on the data you’ve gathered to this point. You’re smart, you know what you’re doing. The intuition of the individual marketer is part of the secret sauce that makes great marketing great. And while artificial intelligence is absolutely amazing at what it does, it lacks that spark that an actual person has. Use yours.

It’s best, of course, to use some of column A and some of column B. Don’t abandon the revolutionary tech that’s available to you, and also don’t abandon your own knowledge and opinions about what will and won’t work. As long as the decisions you’re making are set in the foundation of the research you’ve done, you’re going to be fine.

Keep the content fresh—and relevant

The content flywheel never rests. It’s important to consistently create new content, and that content has to be quality. Otherwise, why bother? One of the better ways to determine whether or not you’re hitting the right beats on content is to dig deep into what exactly your audience is looking for, and topic modeling and topic clusters will do just that.

If you’re talking about things that people care about, they’ll listen. And not only will that drive traffic to your site—it’ll drive revenue for your business.