le shares 10 types of data you should include on your wish list for keyword clustering.
Everyone is talking about keywords clusters. It’s really quite simple: group keywords that are related together. It sounds easy, doesn’t it?
Free tools can help you learn the basics of Natural Language Processing (NLP), which helps deduplicate keywords and identify semantic similarity. It’s fine to start there, but they are limited. Google has a much larger database of data that it can use to improve its algorithms. This includes on-page data, links, and other information.
You must collect SERP information to understand how Google views the world. This will allow you to determine which pages are ranked for what terms. By comparing the number of URLs that overlap in the 10 top results at scale, you can get a clearer picture about which SERPs relate. Keyword Insights has popularized this method, which is also available through Nozzle, Cluster AI and other providers.
Google does not show any URLs that overlap keywords I have manually grouped. It doesn’t matter if Google is “right”. We live in Google’s universe.
You can see the eight top results are identical when you compare the search results for “SEO agency” and “SEO company” with the ads removed!
It is almost impossible to manually find these overlapped pages at scale, but it’s trivial with the right tool. Since years, various tools have helped curate keyword lists. However, they do not go deeper. Even a newcomer on the block offers basic clustering but has a disappointing 2,000-word limit.
It’s great to automatically cluster your keywords. But most tools only provide a list of words, perhaps with search volume or rank. This is a list of 10 data types that are needed to understand keyword clusters. Most of these have not been available until now.
- Ranking URLs
- Refine by
- PAA
- The FAQ
- SERP Features
- Search for intent
- Ranking Position
- Share of Voice
- Entities
- The following categories are available:
1. Ranking URLs/pages
The tools that are available do not show you which pages are shared by all of the keywords within the cluster. This makes it difficult to understand what Google rewards. Knowing the number of URLs also gives a lot of insight into how tight or strong the cluster is. In the example, sharing 8 out of 10 URLs indicates a tight cluster. Only 3-4 pages that overlap are considered moderately tight.
You have to make a decision before you start using most tools about the number of overlapping URLs you want to count. This is difficult to do before you actually see the data. This value should be dynamic and change as you explore the data, without requiring that you pay for a new clustering process.
You can often see the top search results for one keyword if you are familiar with tools that help you write content. Scraping the top URLs for all the keywords in the cluster is more efficient!
It can be helpful to see detailed information on their headings, the schema and text statistics like word count/grade.
2. Refine by
The H1 hidden tag “Filters & Topics” is a great way to find relevant information about a topic.
Google will link to related topics after a few tabs such as Images, News and Maps. The topic is usually prefixed or appended to the keyword phrase. They are easy to distinguish manually, and they can be distinguished by HTML markup/CSS class.
People Also Ask (PAA) and frequently asked questions (FAQ) are two of the most common ways to ask questions. Questions that are frequently asked (FAQs) by people (PAAs)
Google’s People Also Ask Questions are a goldmine to content creators. It gives them a blueprint of what they should include in their content. The PAAs also tend to be more volatile than the traditional search results. By aggregating these over time, rather than using a single scrape, as most tools do, you can determine which questions are most frequently asked. Even though the SERPs in our example were almost identical, there was no overlap between questions.
The first thing we’ll do is show you the 10 most popular questions from the past 30 days. The SERP count represents the total number SERPs that the question appeared on. Keyword count represents the unique keywords used to show the question.
The following questions are similar to PAA and Google has deemed them relevant enough for the topic that they will provide a significant amount of visual real estate on the site if you implement correct schema.org markup.
5. SERP Features
Content strategy will be affected by the presence or lack of certain SERP features in a cluster. PAA and FAQ are usually very visible on SERPs – in this cluster they rank at positions 2.5 and 4.2 respectively. If you can capture them, adding the right markup and answering relevant questions can generate significant traffic. Maps are shown 65% of time, which indicates a fragmented intent. It is only displayed on one SERP, but it could be a great opportunity for growth if you optimize.
6. Search for intent
The search intent impacts your entire strategy. Knowing the cluster intent and mixed intent is key to creating a successful strategy. In addition to the overall score, search intent should be displayed per result to identify potential opportunities to rank several pages on one SERP. It is useful to have data which informs this intent, such as Google Ads metrics.
7. Ranking positions
It is important to report your current ranking. You might not be ranking at all. If so, you may have some low-hanging fruits. For example, if you are lacking in content, but possess enough topical authority to rank, simply by publishing. If you are ranked 8-15 you may be able 10x your traffic by adding some additional optimization.
You can earn bonus points by comparing newer metrics, such as pixel depth or above-the-fold percentage.
It doesn’t matter if the rank is available, if you cannot visualize it in a way that allows you to see opportunities.
The bubbles are colored according to the rank and the CPC. You can quickly identify clusters that match your criteria and drill down for more details.
8. Overview of the competitive landscape/Share in voice
It’s great to see your own ranking, but even better is to spy on your competition using the same data. You can dominate your competitors by switching between domains.
As each cluster may have a different group of competitors, you should be able to report the percentage of voices per cluster.
9. Entities
Google no longer uses exact-match keywords to search the web. It’s about semantic similarity, which is represented by entities extracted using Natural Language Processing (NLP) from the page content.
To identify important text parts, you can use Google API. They have a demo that shows how to do this.
- Salience : importance of entity in the text.
- Sentiment score: ranges from -1 to 1 with -1 being most negative, 0 being neutral and 1 being most positive.
- Sentiment Magnitude indicates the emotional content of a document.
You’ve probably moved past keyword density as a writer of content that targets a keyword group. It’s important to target the entities that Google is aware of and cares about.
Let’s pretend that we are writing for Bruce Clay and have chosen to target the “SEO companies cluster” example. Normal workflows would have the writer scan a few key pages and then update an existing page. We can now optimize content using entities. We can use NLP to extract the entities from the mapped page and compare them to the cluster entities.
It is clear in this case that the two pages are completely different. If the page that you are comparing is already ranked for other clusters, then this is a clear indication that you should create new content in order to target this particular cluster.
10. Content categorization
We can categorize content in a similar way to how we do entities. By comparing it to the cluster we can find mismatches.
Google offers a categorization API that has nearly 1,100 categories and works in multiple languages. SEO is, unsurprisingly, the most popular category in this cluster of matching URLs. However, category #4 is Product Reviews & Prices Comparisons.
If you feel confident, it is not necessary to add a table of comparisons to your website, but it may be worth considering whether this would provide value to your readers.
The conclusion of the article is:
You’ve been using different tools until now to get a small fraction of the data. Nozzle launches today our keyword clustering on Products Hunt. All this data is at your fingertips. Try it for free on your keywords!
The post Ten types of data you should include in your wish list for keyword clustering first appeared on Search Engine land.