Keyword Strategy by Keyword Clustering

I would have loved to have such a tool...

During my journey as an SEO agency leader, I sought to optimize every process within the company. My goal was to improve return on investment and the quality of deliverables for our clients. And, I must admit, keyword research was one of the tasks that scared me the most! That is, until I understood the importance of semantic fields and keyword clustering! This led to the launch of SEOQuantum 😉.

🥵 I struggled with keyword strategy for a long time: a lengthy and tedious process

We usually start by establishing a list of strategic keywords for our client's business. The goal is to present our client with a keyword strategy based on SEO content. Once this list is determined (from 100 to over 1000 keywords), the team would spend long days enriching, sorting, and grouping them by semantic themes, search volumes, and competition levels...

In short, a tiring task prone to errors. I quickly realized that this task was not fulfilling for anyone and added no value...

💡 I looked for a solution to develop a keyword strategy...

An important reminder: the "old-school SEO" approach, meaning one keyword per page, is outdated! Back then, we optimized phrases like [restaurants in Italy] and [Italian restaurants] using two different pages, one for each keyword. That's probably how we all started...

Google's search algorithm (via updates like Rankbrain, Hummingbird, or Helpful Content) has moved beyond this practice, and thankfully so! Of course, it can still work. Sadly, most SEO plugins (like Yoast for WordPress) still rely on this "1 word = 1 page" approach.

Nowadays, in-depth content covering a variety of related concepts and entities is increasingly winning over "old-school SEO." Of course, you still write about a main keyword. However, you will work on the entire semantics surrounding this key idea to target an unlimited number of related queries!

Faced with this realization, I looked for a solution that would reduce the time spent on strategy development and provide a clear view of my client's SEO market challenges. There are some tools like SEMRush and Majestic, but none of them are sufficient, none of them allow for grouping keywords by semantic proximity...

🔧 Creating the solution to solve the problems

Unable to find the perfect fit, I decided to embark on a crazy project: automating the creation of keyword strategy by keyword clustering. In practice, after creating a list of keywords, SEOQuantum can analyze the pages ranked on the search engine for each keyword to:

1. Group them by semantics
2. Assign a theme to each cluster (group of keywords)
3. Estimate search volume per cluster (group)
4. Suggest new keywords

You will no longer make mistakes in developing your keyword strategy! In less than a minute, you'll find the list of terms to use and have a complete idea of the words. Our product offers you a significant service to rank on Google and attract qualified traffic to your site!

Unlike other tools on the market, you become aware of the semantics needed to rank for multiple keywords (sometimes hundreds!). Short or long tail, everything is possible to be visible on the Web!

🎁 Example: Case study for the "courses and learning" universe

The semantic study on the universe of online courses and learning** is based on a sample of 1,300 generalist keywords. Choosing keywords conditions the rest of the analysis, which is an element to keep in mind for the rest of the case study.

The selection of keywords was done using tools like Google Keyword Planner and SEMRush:

• Most generic keywords containing "Course [XXX]", "Learn [XXX]" are included
• Associated and derived keywords like "salsa YouTube" and "salsa beginner" for the key phrase "SALSA course" are excluded
• Geolocated searches are limited (due to their large number)
• Searches on raw materials (lead prices...) and stock market courses (ORANGE, SFR stock prices...) are excluded

To conduct this study, I developed a new tool** capable of studying the semantics of a list of keywords. To do this, the tool browses and retrieves the texts from the TOP 10 pages ranked on Google France's SERPs, totaling 13,000 pages.

Once the dataset is created, the tool uses a **machine learning algorithm for data partitioning to create clusters, groupings of keywords by semantic affinity. Note: the algorithm determines the appropriate number of clusters "automatically"; some keywords have a "blurry" positioning, meaning they can belong to one or more clusters. The direct consequence is that some clusters do not have well-defined boundaries.

The clusters are then consolidated by data such as monthly search volume per keyword, organic competition level... After processing by the tool, we obtain the first results.

Give some height to your vision

For this study, the detected clusters number 91:

• Drawing
• Development (programming)
• Languages
• Writing
• English
• Driving
• ...

• On the ordinate axis: monthly search volume per cluster (on a logarithmic scale, thanks Walid for the advice)
• On the abscissa axis: organic competition level

Each cluster is positioned on a "search volume/organic competition matrix" allowing a quick overview of the entire semantic universe. The tool automatically labels the cluster name based on its content.

Example of 2 clusters: sewing courses and writing

To maximize the development of the keyword strategy, it is necessary to review the clusters. Notably, 142 keywords are not classified in any cluster (about 11% of the list). Some keywords (about 5%) are in clusters that are not correctly assigned to them (probably due to the blurry cluster boundaries). The tool does 80% of the work 😀.

In the example above, the word "guitar course net" is not in the right cluster, and the "Paris Argentine" cluster should be merged with "dance courses." The study's clusters are closed enough to draw some conclusions.

What lessons can be learned from this study?

The semantic analysis of the universe provides several insights, including: many users search for information on drawing, followed by development (programming). The "Learn to Draw" cluster alone represents over 435,000 monthly queries.

Users ask more questions in the clusters above.

Interestingly, the competition level is high for the "multiplication" cluster with very short content. One tactic to perform well in this cluster is to offer longer content than average.

Finally, note that it is pure players like Superprof and wikihow that cover the market the most.

## 🚀 How can this approach help SEO?

Can an SEO keyword strategy be driven by semantics? Yes and no at the same time. I would say it all depends on the dataset (list of keywords) used. In our case study, the list is far too broad to respond to an effective keyword strategy, as we miss many queries. It would be more of a semantic-based SEO market study.

However, this approach helps to break the ice and **provide decision-making metrics to leaders and the marketing team, allowing them to prioritize based on search volumes and competition levels to establish the right strategy.

Nevertheless, to have a clear and well-defined natural referencing keyword strategy that meets the marketing team's requirements, users' needs, and SEO, it is essential to repeat the exercise focusing solely on one cluster. The main step is to go vertically down the cluster and address all facets of a theme through an exhaustive list of keywords.

When I was in charge of SEO, I used this type of analysis in two ways:

1. either for pre-sales, when you meet your prospect. This gives you the advantage of knowing their business from an SEO perspective in just a few minutes: the different niches, competition levels, and search volumes;
2. or in determining the strategic SEO direction of campaigns, decide which niches to work on first and the effort required.

## 🙏 Sources used to write this article

https://www.searchenginejournal.com/keyword-clusters-seo-content-strategy/399414/

https://blog.hubspot.com/marketing/how-to-do-keyword-research-ht