Keyword clustering types and levels
To explain how clustering works, let's imagine our project has three keywords: A, B, and C. First, we check the volume ("!Volume" in Yandex) for all keywords and sort them from highest to lowest. Let's say keyword A has a volume of 100, keyword B — 50, and keyword C — 30.
Next, we collect the Top 10 websites in search results for each keyword. We'll refer to these results as results A, results B, and results C. The next stage involves comparing URLs in these Top results with each other. The clustering type determines the principle of URL comparison.
SOFT
With the SOFT type, we first compare the results B with results A (the keyword with the highest volume is always taken as the basis in all type). If there are identical URLs in the Top B and Top A (as many as specified in the clustering Level), keyword B is added to keyword A's group. The same happens with results C: we compare results C and A, and if they have identical URLs, adds keyword C to keyword A's group. At the same time, it doesn't matter whether there are common URLs in the results for keywords B and C — they are not compared with each other. As a result, with the SOFT type, all keywords in the group will be connected to the most high‑volume keyword, but other keywords may not be connected to each other.
MODERATE
With the MODERATE type, keywords are compared in pairs. First, the results A are compared with results B. If they have identical URLs, then keyword B is placed in the same group with keyword A. Next, the results B are compared with results C. If they have identical URLs, keyword C is added to the group with keywords A and B. With this type, keywords A and C might not have common URLs, meaning all keywords will be pairwise connected to each other, but the URLs in Top results for different keyword pairs may vary.
HARD
With the HARD type, search results are compared across all keywords: A, B, and C. To group them together, all three Top results must contain matching URLs. This means identical pages (number of pages = clustering Level) must appear simultaneously in results A, B, and C. As a result, all keywords become maximally interconnected through these shared URLs.
Clustering level
Clustering level — refers to the number of matching URLs in search results required for keywords to be grouped together. For example, if the clustering level is set to 2, the system will look for two matching URLs in the results and group such keywords into a single cluster.
The probability of matching 1–3 URLs in search results across keywords is higher than matching 9–10 URLs. Therefore, the lower the clustering level, the more complete the groups will be, resulting in fewer total groups. If you set the level to 8‑10, you'll get many small groups with only a few keywords in each.
How to test different clustering types and compare results?
To select multiple clustering levels, tick Send to email when running the clustering. You'll receive results for all levels via email as CSV files.
Within 7 days after clustering, you can change the Clustering Level and Clustering Type at a discounted price. To re‑cluster, click on the toolbar, select Level and Type and push Change.