Good afternoon, dear readers! We all know that the collection search queries is the foundation for successful SEO promotion of any web resource in Yandex and Google. But as practice shows, this stage of search promotion is not very successful for many. Therefore, in my Notes, I periodically issue practical materials on this topic. Now let's talk about how to competently cluster a semantic site or blog with your own hands. In the article you will understand the importance of this process, see the options for its implementation, learn how to group keywords.

1. First of all, we form - phrases, which will then form the basis of future keywords.
2. At the parsing stage, we get the entire pool of search queries on our topic, .
3. With the help of the parameters found, we clean out the future semantic core.
4. Having a list of all found site keywords, the stage of their distribution among the pages of our web resource begins.

Site keyword clustering

It is the last step in creating semantic core raises many questions. The fact is that the first stages (except the first) are more or less automated. It does not take much effort to collect requests and analyze them. But the stage of grouping keywords requires a maximum of time and mental costs from the webmaster. Therefore, various errors occur. Our task is to prevent them!

What is keyword grouping

Grouping (or clustering) keywords is the process of distributing search queries of the same topic (a group of queries) to promote one page. Why one? You will find the answer in my practical.

In other words, with the help of this stage, we form the found phrases into separate semantic groups. Each group is embedded only on its promoted page and solves one task (common for all requests of this group):

• gives an answer to the user's question from search engines (blogs, info-sites);
• offers commercial data on specific products (online stores);
• provides information on the service (commercial sites, specialized portals).

This way, all group queries match the main theme of a particular landing page. All of them reveal the purpose of a particular site document from one side or another.

You can study in more detail about the concept of semantic core clustering. In it you will find the history of the appearance of this type of seo-work, you will see good example grouping requests.

Also, search query clustering refers to the automated collection of phrases through the interaction of the service with pages in search results. I will talk about this in more detail when we talk about the Topvisor service of the same name.

What gives competent clustering of keys for the site

Each group of queries found is not only keywords for the content of the landing page. It is clear that they need to be embedded in the site document (information blog post, product card of an online store, etc.). In addition, query clustering allows you to get:

• vision of the future full-fledged structure of the new site (or the old one);
• a guide to topics that are of interest to users from the search;
• understanding the current demand for goods/services in a specific niche;
• seo-promotion plan (what content to form in the first place);
• material from which requests for page optimization are visible.

I will explain one important point. Clustering into groups provides a huge opportunity to use all resources to attract maximum search traffic! Without grouping search queries, we thereby cut off our site from the consecration of topics that users from Yandex and Google need.

What promises the wrong distribution of keywords

So, without having distributed groups on a large topic, the owner of a web resource does not see the full picture of the promotion. This is the biggest problem that arises when the clustering step is ignored (or incompletely done).

But even the presence of this step in your promotion plan cannot guarantee you the achievement of all assigned seo tasks. This can happen due to errors that occur when grouping requests into landing pages. Here are the problems that the wrong distribution of keywords gives:

• the appearance of duplicates in search engine indexes (due to cannibalization);
• loss or non-receipt of places in the top ten search results;
• loss of money spent on the formation of "extra" content;
• worsening behavioral factors not achieving the set goals.

As one famous movie character said, “Oil Painting”. There is no other way to say it here. In modern search promotion you can’t do “a little” or leave something for later. Everything must be done on time and with meaning. Semantic core clustering is the stage after which “meat” (content) is strung on the site skeleton. And here, any mistake turns the SEO-promotion of the site into a real apocalypse. Problems arise where they are not expected.

Ways of grouping the semantic core

To form groups of keywords of the site semantics, manual and/or automated methods are used. The first is the most routine and difficult. But the most reliable - there is no better checker and proofreader than a person.

Manual way very laborious. It's one thing to distribute 500 requests to a number of articles new topic. Another thing is when you need to scan 10,000 keywords and form appropriate groups from them. This requires endurance and patience.

Automated way takes over almost the entire routine - the webmaster or specialist only needs to check the result of clustering. But other difficulties arise - some of the requests are not in their own groups, the distribution logic may differ from the one that a person imagines.

In order to somehow level the sometimes "stupid" logic of the machine, it is used semi-automatic way site distribution. In this case, the specialist pre-forms general groups according to the found requests. And automation distributes requests already in these groups themselves. Thus, distribution errors are much less.

Now let's look at each way to cluster the semantic core, using the best tools (in my opinion, of course).

Automated way to group a site's CL

Currently, there are a lot of automatic options for clustering search phrases. All of them differ in interface, there are even differences in the distribution process itself. More details about modern web resource grouping tools can be found in the review article by Alexander Alaev. I will analyze only two options that I use myself and consider them the best.

Group Analysis in Key Collector

Key Collector is the most the best software to create a semantic core. Of course, he also has his own way of distributing requests - the "Analysis of groups" tool. It allows you to break down all found search queries. Here is the scheme of work with it.

Select the "Group Analysis" tool

Suppose we need to make a distribution of parsed queries on the topic of CCTV cameras. To do this, click the corresponding icon on the control panel in the "Data" tab:

In the window that opens, select the clustering mode. The Key Collector program has several such modes:

• by individual words
• by the composition of phrases
• by search results
• by the composition of phrases and search results

The first option is for very fine tuning- it groups those queries that have the same at least one word. For example, the search phrases “Minsk video cameras to buy” and “Minsk video salons” would be included by the software in one group.

The second option takes into account the structure of the found search phrases and the number of their matches (the “Strength by composition” element is responsible for this). Great way to cluster a large number of requests.

The "by search results" grouping mode combines keywords by the number of links in the search results that match between phrases. It works only if the Key Collector has taken information about the search results for the site (“KEI components”). There is also a grouping strength, which is responsible for the degree of connection between phrases.

The last way to distribute requests is a combination of the 2nd and 3rd, the degree of importance of which is selected by a special slider.

Grouping requests

From my own experience I will say that for most cases it is the second grouping option that is suitable. We will consider it in more detail. For example, I took the topic of CCTV cameras. Selecting the type of grouping, indicate its strength. I recommend putting the number 2 or 3. In my case, the first value came up to me. As a result, I received 434 groups with 1789 useful phrases:

If I had taken a larger number, then the binding would have been slightly different and the number of groups would have decreased. Namely, it became equal to 298, and the number of useful phrases, according to our condition, also decreased - 1207.

Let's see what groups we ended up with in the first case with a composition strength of 2. To do this, I will export the grouping made to Excel spreadsheet. Here is an excerpt from the exported table:

As you can see, in general, Key Collector was able to make a good distribution of the found search queries. But still, it will not hurt to file this plate with a file - there are some groups and individual words that are not in their place.

Clustering requests in the Topvisor service

So, the requests have already been found (they could also be imported):

All that's left is to group them. To do this, click the special icon on the Topvisor toolbar and select the desired settings:

If the choice of search engine and region (required!) Is not difficult, then indicating the degree of grouping may raise questions. Let me explain what kind of animal it is.

The degree of grouping is the number of page addresses that will be taken to check the similarity of our requests for one or another group. For example, if this parameter is equal to 3, then having a number of search phrases for grouping, each of them will be checked only for 3 pages in the search results, and not for the entire top at once. In principle, this is enough to see the structure of the semantic core (query group) as a whole. If it is necessary to have more precise groups, choose the number 8 or 9.

So, after ten minutes, we get the finished result of clustering our semantic core:

As a result, Topvisor was able to distribute 3181 found queries into 514 groups. Moreover, one group, the last one, is called "Requests without connections" and contains phrases that did not match in the top:

If this result does not suit us, we can immediately regroup - just click on the same button in the control panel. By the way, the regrouping will cost a penny (it seems that Topvisor immediately makes a grouping on great importance, therefore, it requires little material and time costs for redistribution):

As a result, after the new grouping, we get a different number of groups (less than last time) and the number of phrases in them. They are more detailed, but on the other hand, the number of requests that did not fall into these groups (“Requests without links”) has increased noticeably.

For final work with a clustered semantic core, you can export queries to an Excel table or text file. Here is how the table with the name of the group looks like in the end:

In general, Topvisor does an excellent job of distributing the semantic core. good tool for those who cannot spend their time collecting requests in the Key Collector program. But, really, expensive.

Clustering queries in the Serpstat service

Another sensible option for automatic clustering is the popular Serpstat service. But unlike Topvisor, this seo platform was able to develop its own unique query grouping technology. It is done in the following way.

We go in our account to the tool "Clustering and text analytics":

By clicking the "Create a project" button, we go through the cycle of steps, indicating the necessary data on the project in the appropriate columns. Let's take as an example the task of grouping requests for my seo blog on the topic "semantic core".

First, we go through step 1, specifying the project name and domain address:

Then we add search queries to the Serpstat clustering tool. This can be done either manually or using a downloadable list in txt or csv format:

And now the most crucial step remains - to choose a scheme according to which the requests we have indicated will be grouped. To do this, we indicate the strength of the connection and the clustering option:

Seo-platform Serpstat for clustering examines all the phrases that are included in the project. And for a competent grouping, the service for these phrases studies their intersection in the search results of the search engines we have indicated!

You can study a more detailed Serpstat clustering scheme on your own in this service blog article.

After giving the last instructions, Serpstat starts grouping. After processing all the data, we get a set of queries, the phrases of which are grouped according to their identical attribute. Here is a piece of the distribution of requests made in my example on the topic “semantic core” (the picture is clickable !!!):

Manual method of semantic core distribution

By manual method, I call such a variant of core clustering, in which we independently indicate groups, put things in order in them, and form the structure of the CL on our own. Of course, without special programs did not work here. Rather, without one - Excel.

Distribute queries using Excel

Everything is simple here - we unload the already collected and edited search queries and form groups with pens and throw suitable phrases into them.

I described this clustering option in detail in . There I give 3 ways to group, choose yours and go for it. I personally combine them depending on the situation.

Keyword Clustering with the Core File

This option differs from the previous one in that here we are already throwing phrases thanks to the smart “Kernel” script made in Excel. Everything else is also done by hand.

The Core script was made by the guys from MFC (Made for content) to facilitate the task of distributing keywords. As a basis, they took the experience of seo-specialist Sergey Koksharov, who came up with the option with Excel. Let's see how this macro works. To do this, I will use the video of the guys from MFC:

In general, everything is clear. There is nothing complicated here. Therefore, if you are unable to use Key program Collector, and you have a lot of semantic kernels for distribution, use the “Kernel” script (google it). Even if you do clustering rarely for your site or blog, this macro will not be superfluous. At least you should start with it, and only then finish the grouping in the first manual way.

Yes, I forgot to say the most important thing about the Core file - it's free!

Bonus - my way of clustering queries

I call it semi-automatic - the role of a person here is important at the very beginning and at the end. I reflected it in a special seo cheat sheet, take it and feel free to use it:

I can only say that it was based on working with the correct search for a list of masks, using the Key Collector software and the usual logic.

This concludes my little educational program on the clustering of search phrases of the semantic core. If you have any questions or suggestions, please feel free to comment!

Where can I order an excellent semantic core?

By the way, if you plan to build a semantic core for your project, you can order semantics by contacting . Thank you!

Sincerely, Your Maxim Dovzhenko

Query clustering sorts (breaks) the list of the semantic core (SN) into groups by similarity, which makes it possible to further optimize the site pages for them.

How are requests clustered?

The tool analyzes Yandex results for each request and compares it with the results of other requests from the list. If the TOP-10 for different queries contains the same relevant pages, then these queries are defined as similar and placed in one group. This means that one page can be optimized for them.

The query clustering threshold is the number of matching relevant pages in the SERP for different queries. Simply put, if you enter two queries into Yandex and the TOP-10 results show two identical pages (two out of ten), then when you set the "clustering threshold 2", these two queries will be placed in one group.

Disadvantages of manually grouping requests

Keyword grouping, also known as slicing, is performed by SEOs immediately after collecting keywords.

1. With a large number of requests, it is difficult to manual mode to determine their similarity with each other, you either have to enter each query into the search, or rely on intuition/experience, which can play a trick on promotion and not give the desired results.
2. High cost, which was formed due to the duration of the process. It takes an average of 4..16 hours for a high-quality breakdown of semantics with 500 requests on board. It is necessary to subtract each request, determine its group (the presence of which must be kept in mind), if necessary, double-check with search or services ... brrr.

Advantages of automatic query grouping

1. The speed of staking is approximately equal to the speed of sound. The system will check the results of each of the requests, compare them and give you the opportunity to correct possible minor exceptions manually, after which the result can be uploaded to csv file(excel).
2. The accuracy of the result achieved by eliminating the human factor. A person can get distracted and lose a thought, forget, misunderstand, or simply not be able to do the breakdown correctly; such difficulties are not observed with the program.
3. The tool is provided completely free of charge; it does not require monthly wages, holidays, sick leave; also he does not have a work schedule: he works 24/7.

Breakdown is a very important process when promoting, it sets goals for optimizing each page of the project and the entire site as a whole.

By the way, you can pay attention that there is even a .

All we need is to indicate in front of each phrase in one or two words what this phrase is about. There is no rule, just put such words or phrases that you immediately remember and, looking at which, you will immediately understand what the article should be about, which includes the keys opposite which you put them.

It will be easier to look at the screenshot and everything will become clear:

Got it? Now all that remains for us is to enable the filter in Excel and filter by group. Each such group of keywords in the semantic core is a separate article!

Everything is very simple! It may seem to you that this is a very long time, but in fact, it will take only 20-30 minutes to ungroup 1000-1500 requests, with some skill!

Paid ungrouping service - TopVisor

After that, we add all the existing semantic core as a list

We configure the kernel clustering parameters in the service and click "Start"

Some time passes, and we get ungrouped requests. Next, we unload these queries and, if necessary (and in the case of parsing a real semantic core, this will definitely be required), we group the queries by combining some groups that are similar in semantic load with each other.

A huge advantage of this service is the ability to pay using XML limits.

Free Query Clustering Services

In addition to the paid grouper from TopVisor, there are also free ones. online services search query clustering.

Their functionality and usability are much poorer than those of their paid counterparts, but you can break down requests for free without registration and SMS, which, in the absence of funds, will help close your eyes to the interface and the presence of various chips.

Clusterer from py7.ru

The interface is very simple, it is impossible to get confused. Add requests and click "Group"

Ungrouper by Contentmonster

The so-called assistant from ContentMonster is still in test mode, which is why for some reason it didn’t ungroup anything for me, maybe you will be more lucky 🙂

Conclusion

As you can see, there are many ways and tools for clustering search queries. You can split phrases into groups either completely manually or by resorting to specialized services for free or for money.

However, be that as it may, no matter what you use, it is important to remember that clustering is an important stage, only by collecting and disassembling several semantic cores, you will learn how to do it correctly, it is experience that is important in this matter.

By the way, recently a very convenient and effective program for, I recommend that you read my review.

And if you don't want to spend time collecting semantics and clustering, then you can always visit me.

Good afternoon friends. Today our guest is Anatoly Ulitovsky, famous SEO Specialist Runet. Anatoly will tell us about keyword clustering.

Any proper promotion starts with the semantic core. The main purpose of which is to evaluate the frequency, potential traffic and the level of competition.

Paid and free services who do this work on the internet great amount. But the most difficult begins after compiling a list of keys. When the received keys need to be divided into pages of the site. This work must be done either manually or using specialized clusterers.

What does clustering give

Clustering helps to create a convenient site structure, facilitates relinking, and increases the relevance of the page to promoted queries.

A bit of theory

Webmasters use two fundamentally different approaches to clustering:

1. According to the composition of key phrases. Queries are combined into groups based on the analysis of their constituent words.
2. By search results. For each request, the TOP issuance is found and the matching threshold is set - for example, 50%. Those keys for which at least half of the pages in the TOP of the issue match are combined into one group. You can take any matching threshold, you can analyze any number of search results: TOP-3, TOP-5, TOP-10, TOP-20.

The second clustering method - based on output analysis - is more popular than the first:

The first 6 sites consider clustering to be synonymous with grouping based on search results. For 4 sites, this is already visible in the snippet, the rest (2nd and 6th) write about it on the pages.

What to choose?

Proponents of grouping keywords by search results are ignoring two things.

First, each search engine their ranking algorithms. See what it looks like search results for the phrase "what is query clustering" for Yandex users from Moscow:

Let's compare it with issuance by Google given earlier.

Clustering queries by TOP results means that we will focus on website promotion in only one search engine.

When there is already a list of requests, this is not yet a semantic core - you should first scatter requests across pages in order to have an idea of ​​\u200b\u200bhow to fill the site. Without good semantics, it will be very difficult to get traffic from search.

What is query clustering

Query clustering is just the distribution of search queries of the same topic into groups to promote the landing page.

Clustering includes the following processes:

• grouping requests depending on the user's intentions (intent);
• checking the compatibility of keywords for promotion on the same page in the Yandex top.

requests with the same intent- these are different requests through which a person, in fact, is looking for the same thing. The obvious examples are [Parker pen] and [Parker pen]. The situation is more complicated with such synonyms as: [ desk lamp] - [night light], [birth certificate] - [metric], [monitor] - [screen]. The difficulty lies in the fact that when searching for synonyms for keys through the Yandex dictionary, the system does not always offer an adequate selection.

In practice, similar queries can have a lot of different characteristics, which prevent them from being placed on the same page. Clustering queries by tops comes to the rescue. The clusterer finds the same URLs in the top of the search engine results, thereby signaling the presence of the same intent. The result of the work is expressed as follows:

• the presence of the same URLs in the top by queries means the possibility of their promotion on the same page;
• the absence of common URLs indicates, with a high probability, the impossibility of such promotion.

Why clustering is needed

With the help of automatic clusterers, even the largest semantic cores can be quickly grouped. If earlier it took weeks and months to disassemble the kernel, then thanks to clusterers, the work is reduced to a couple of hours. A big plus of clustering is the distribution of requests across pages in such a way that they can be promoted at the same time. It is difficult to imagine a manual analogue of high-precision clustering, since even an experienced optimizer makes up to 30% of erroneous distributions. It follows that keyword clustering is necessary in almost any case.

When I was a teapot webmaster, I made a website where there was a separate article for each request. Of course, he did not receive traffic - only a file turned out. And this is the problem of really many beginners - wrong queries or wrong clustering.

Clustering methods

When grouping requests, there is uncertainty in the methodology for combining them based on tops. In practice, there are two main methods: “soft” and “hard” clustering.

Soft clustering is based on the formation of a group from one "central" request. Everyone else is compared with him in terms of the number of common URLs in Yandex's top 10. Soft clustering forms groups of a fairly large size, but errors often occur in determining the possibility of joint promotion of requests on a page.

Hard clustering is characterized by combining requests into a group when there is a common set of URLs for all requests, which is shown for all these requests in the top 10.

There are two criteria for evaluating clustering:

1. completeness- the number of requests in the group that have the same "intent". If all requests with the same intent fall into one group, the completeness indicator is 100%.
2. Compatibility requests among themselves, falling into the same group. For 100%, they take the case when all requests that are in the cluster are compatible with each other.

An important role is played by such a parameter as clustering threshold". This is the minimum number of common URLs to form a group. Big number means a high accuracy of the groups, but at the same time they naturally decrease in size. The experience of using semantic clusterers shows that the minimum working threshold for "hard" clustering is 3 URLs, for "soft" - 4 URLs.

Even with a threshold of 3 URLs, hard clustering provides over 90% accuracy. For comparison: without the use of tools, the accuracy of an experienced optimizer, at best, will be 70%, and a beginner - no more than 30%. Despite the high accuracy, the "hard" method gives only about 40% of the completeness.

Soft clustering has high rate completeness, but significantly loses in accuracy. Thus, "soft" and "hard" methods are inversely proportional to each other. The use of one method or another depends on the goals of the optimization process.

With “traffic” promotion, when it is important to display as many requests as possible on the page, soft clustering is better suited. If “positional” promotion is carried out, then hard has the final say.

Hard-clustering is also used for text analysis of a page. Any text analysis on a group of requests for a page correlates quite strictly with the quality of this group. Only the "hard" method provides groups of the desired quality.

How to make a grouping of the semantic core

I usually do clustering in two steps. In the first stage, I throw the kernel into some automatic clustering service / program, and in the second stage, I finish the kernel manually. Through Excel. Here's how these guys are:

In these videos, in principle, it is clear how to do manual finishing, but as for automatic clusterers, everyone chooses what he likes best.

Semparser

Topvisor's automatic query grouper is an alternative to Rush Analytics and Semparser, and its interface is similar to the latter. The degree of grouping and saving the project in an Excel file is present.

The Topvisor clusterer has a "regroup" operation. After its application, the number of groups increases, and the number of requests in them noticeably decreases. This function useful for those who are not satisfied with soft-clustering and the hard version will do.

“Regrouping” is paid here, although it takes no more than a couple of rubles.

The advantage of Topvisor is based on the high grouping speed. The clusterer will distribute the semantic core of 1000 queries in a matter of minutes. Disadvantages: the high cost of grouping and, of course, the need for manual editing.

Grouping via Key Collector

Another example of an automatic clusterer is presented as an online tool at coolakov.ru. The breakdown of requests into groups is based on the similarity of the Yandex top 10.

Plus: free online service.
Cons: low grouping accuracy, lack of uploading to a file.

Summing up, you can confidently opt for automatic clusterers offered by various online services. But, unfortunately, the operation of any clusterer requires manual refinement.