Search Engine Marketing for B2B Company
The story of three years of search engine optimization of a B2B project with a prologue and epilogue.
SEO audit, research, recommendations, implementation and results
In mid-2020, SVEL Group, a major manufacturer of electrical equipment, approached us. And the conversation turned to the situation with the search traffic of the company's website.
The situation was this: search traffic had been growing for many years, but since the winter of 2019, it began to decline and could not be stopped.
The development of a new website did not improve the situation and even made it a little worse.
With these inputs, the audit of search engine optimization of the SVEL project began.
The objectives of the audit were formulated as follows: to develop feasible recommendations and an action plan for restoring and significantly increasing target search traffic.
To do this, it was necessary to understand the reasons that led to the current situation, identify the unrealized potential of the project in terms of attracting search traffic and understand how and in what time frame the identified potential can be realized. That is, to turn search demand into search traffic.
Research of search demand
The research period lasted several months.
The search demand by topic was seriously studied. In order to cluster the queries, we thoroughly studied the range of manufactured products, the semantics of the electrical engineering industry in general, regulatory and operational documentation, reread hundreds of articles and conducted dozens of interviews with the company's employees.
The result was 14 segments and a total of 2500 requests.
The semantic core was structured by types of queries (information, transactional, navigational) and by areas of activity. They are also sections of the catalog.
The total number of requests with these queries to search engines (the total frequency of the semantic core) at the time of the study amounted to more than one million requests per month.
We take the frequency of queries from Yandex Wordstat and extrapolate for Google through the search engine popularity index. We look at the popularity of search engines in the open statistics of the LiveInternet service for the .ru domain.
To forecast future search traffic for each segment, seasonality of search demand fluctuations was determined.
Visibility and Search Results Research
By measuring the current visibility of the site at that time, we determined what share of the information demand it covers and with whom it competes for semantic core queries.
At the time of the first measurements, the average site visibility was 3%that is, the site was seen in search results only in three cases out of a hundred.
At this stage, we gained an understanding of the potential for attracting search traffic by segment.
If the visibility of a segment is 100% or close to this value, then it will not be possible to attract much additional traffic, even if the total frequency of the segment is high. But if the visibility is, for example, 5%, and everything is fine with the frequency, then this means that the segment has a high potential for attracting traffic.
But there is one thing. Competition.
Competitor Research
After determining the positions and visibility of the SVEL website for semantic core queries, we gained an understanding of which websites compete with us in search results.
It was important to understand what part of them are occupied by commercial resources of enterprises producing electrical equipment, and what part are aggregators, recommendation systems, information portals, etc.
We were lucky, in almost all segments the share of commercial sites was high, and this inspired optimism, since it is much easier to compete with them than with Wikipedia, Ozon and Zen.
We took several dozen competitor sites we selected and analyzed them into molecules.
It was important for us to understand why their positions for the queries of our semantic core are good.
What kind of content is searched for by the types of queries, what type is it (product description, catalog section or article), how is it written, what size, how many illustrations, is it original.
The structure of successful competitors' sites was interesting. What is the site's volume in pages, what kind of pages are they, at what level of nesting are they located, how are they linked?
How many domains and pages link to competitors' sites, what is the quality of their link profile.
How fast and adaptive are competitors, how well are they indexed by search engines. How highly does Yandex rate them in its IKS metric.
Several dozen sites were dismantled, studied, and from them a virtual “ideal competitor” was assembled. A site that possessed the best features of all competitor sites. Mom's friend's son).
Thus, we have an understanding of which sites Yandex and Google consider worthy of high positions.
In addition to the internal and external groups of factors that we see and can evaluate, the ranking of a site in search results is influenced by user factors, the value of which is impossible to study for competitors' sites without access to their statistics. We can only assume that they are doing well with this. Therefore, in this part, the portrait of the son of a mother's friend was as ideal as possible.
Technical audit and statistics analysis
The SVEL Group website appeared in 2010. And thanks to the fact that the Yandex Metrica counter has been on the website since 2014, very interesting statistics have accumulated over 7 years.
Search traffic has been growing steadily for five years. It has grown by 50% over five years, month-on-month, between 2014 and 2019.
We cannot give absolute traffic figures, but we will only say that these are tens of thousands of visits per month. This is a fairly high ceiling.
But in 2019, traffic began to fall approximately equally in both search engines. That is, the problem is clearly in the site itself.
We understood which pages and sections attracted search traffic in the past, and how the dynamics of traffic to sections changed.
The reasons for the decrease in search traffic became clear. It began due to the restructuring of the catalog, which happened a year before the request to us. The addresses and content of the pages were changed, the pages fell out of the index, and the new pages were unable to return to high positions, losing to competitors.
The technical part of the audit was not ignored.
The site was checked by automatic services and by specialists. All identified shortcomings were subsequently corrected.
Unfortunately, there was no silver bullet. The site had no shortcomings, the correction of which would have given an immediate increase in traffic. Alas)
Different content for different audiences and different tasks
The audit of the texts showed that there were two problems with the texts:
The available texts are insufficient to cover the identified search demand
Of those that exist, 50% are not original enough.
At this stage, the main audit recommendations related to the site content were developed.
The target audience of the SVEL business is very diverse and turns to search engines to get answers to a variety of questions. The task of search engine optimization is to provide answers to these questions on the pages of the project site and in return receive requests from potential clients, recognition in the professional environment and consumer loyalty.
It was suggested to create a detailed catalog with product cards for those areas where the number of possible product modifications is maximum. These are, first of all, measuring transformers, of which there are hundreds of thousands of design variants.
These pages should attract traffic for a huge number of low-frequency transactional queries, for example, “TOL-10-1-0.2S/5P/10/5 UHL2”
Based on this catalog, create thousands of product group pages filtered by several features (facets) to accommodate higher-frequency queries like “TOL 10 2”.
To cover high-frequency queries like “TMG Transformer”, it was planned to create popular articles.
For requests such as “operating manual for switchgear”, create sections with documentation on the operation of manufactured devices.
For queries like “GOST 15150-69”, publish texts of state and industry standards
The traffic that is attracted to these pages solves various business problems.
Search Traffic Forecast
The forecast was made for three years. Yes, we are optimists)
Forecasting Methodology
Traffic was forecasted in terms of slots of intersection of semantic core segments and CIS countries.
Knowing the volume of information demand and the current visibility of the site, we determine what visibility the project can achieve for a segment in the region.
Our mother's friend's son, “the ideal competitor”, gives us the maximum visibility for the slot.
In a simple version, the dynamics of visibility growth are assumed to be ideally linear – from a point of 3% to a point of, for example, 50%.
We take into account the seasonality of demand for the segment, and from the frequency forecast through visibility and CTR in search we arrive at potential search traffic.
And so on several dozen times for each slot segment/region with a resolution of one month.
Example:
Currently, the frequency of search results for this segment is 50,000 per month.
Visibility 3%.
This means that the site was seen by 50,000*3% = 1,500 users.
CTR for search is 4%.
This means that the site had 1500 * 4% = 60 visits.
If the frequency increases to 60,000 in a few months during the season,
And by that time visibility will be 10%,
Then the site will be seen 60,000 * 10 = 6000 times,
And if the CTR on search is still 4%, there will be 6000 * 4% = 240 visits.
We observe CTR in Yandex Webmaster and Google Search Console.
As a rule, as visibility increases, CTR in search increases.
In 2019, we predicted traffic growth of 436% (month-on-month) over the next three years through October 2023. And of course, we were wrong.
Our forecasting method is good for transactional queries and, as it turns out, fails for informational queries because the segment of informational queries we collected was incomplete. We were unable to predict that the posted content would attract traffic for so many queries.
Now (October 2023) the site is searched for by a huge number of queries. According to Yandex Webmaster, the site's pages are shown in Yandex on average 1,020,000 times per month for 350,000 queries. Of these million searchers, more than half see the site in the search results.
Embodiment of ideas and implementation of elephants
Here it should be said that our ideas, for the most part, found approval and comprehensive support from SVEL marketing.
The SVEL Group and the people who represent it possess the main thing for the implementation of ideas – Will.
Over the course of several years (at the time of writing this article, 3 years have passed) the decision to develop the site through content has not wavered even once. And this is important, a strategy is a strategy for a long time.
The elephant was implemented in parts, but many processes were going on in parallel:
Development of product catalog functionality and creation of tens of thousands of product card pages and facets.
Creating popular articles.
Creating texts for the catalog.
Creating texts for product cards.
Posting on the website operating manuals and regulatory documentation translated from pdf to html.
Parallel processes accelerated the receipt of results, but complicated communications and burdened accounts on both sides.
Of course, there were some difficulties and problems along the way.
The main problem in creating expert content is the lack of skills and ability of competent people to write texts for the project and the lack of special competencies of people who can and want to write texts for the project.
This is typical for all projects where the product is technically complex.
We solved this problem from two sides.
First, of course, they looked for and found authors with specialized technical education. And after they sorted out the range and features, the texts of popular articles became acceptable for posting on the project website without damaging the company's reputation.
Secondly, we introduced the concept of “popular articles” – texts explaining the basics of electrical engineering and classification of devices in relation to the holding's products. They are designed for a non-professional and poorly prepared audience, optimized for high-frequency information requests.
The professional requirements for these articles are much more flexible than for texts in the catalog, on the pages of categories and product groups, which made it possible to create a corpus of these articles, which became the basis for covering requests at the top of the needs funnel.
Another content issue is the uniqueness of the texts for product cards. More precisely, their non-uniqueness.
The project contains several tens of thousands of product cards and several thousand facet pages (filter pages with static URLs).
These pages are needed to land a large number of low-frequency transactional requests.
The problem is that two product pages, the only difference between which is the number of primary windings, differ from each other by literally a couple of kilobytes.
Accordingly, there is a problem with such pages getting into the index of search engines. They consider such pages to be of little importance.
In Yandex this problem is less pronounced, in Google – more so.
At first, we went down the path of generating texts for product cards using templates from constant and variable parts, replacing variables with synonyms for different pages. But the differences between pages were still insufficient to be considered significant. The second stage was turning variable parts into small independent texts. Then we began trying to create product description texts using AI with different prompt options.
We cannot claim that the problem has been completely solved. But we have advanced far enough and consider this situation as a project that has not yet been fully realized.
Unoriginality of some content
For the segments “Designers”, “Equipment Operators”, “Students of specialized specialties” we have placed on the site a large array of texts of state and industry standards, equipment operating instructions and other regulatory industry documentation.
These are excellent texts, but they are not original at all. They are published on thousands of sites, many of which are more authoritative than the SVEL project site.
And the search demand for GOSTs, OSTs (Industry Standards) and PUE (Electrical Installation Rules) is huge.
It is impossible to cover even a small part of this demand by publishing unoriginal texts. But no, it turns out it is possible.
We did not take the texts of regulatory documentation from other sites. Instead, we manually translated all industry standards, operating manuals, GOSTs and PUEs into HTML.
It was a lot of work. The pages turned out clean, without any junk code, well structured and easy to read.
We made sure that the pages received good internal PageRank.
PageRank is a Google-designed metric for measuring the weight of a page, the value of which depends on the number and weight of pages linking to the page. Google calculates PageRank for all pages on the Internet, but you can calculate PageRank for pages on a site locally.
And these generally simple actions produced results.
Over time, the site began to attract a very serious amount of search traffic to these pages. Much more than we predicted.
Unexpected but correct decisions
We also made a seemingly controversial decision – we banned all PDF documents on the site from indexing and converted all texts to HTML. As a result, the site first lost positions for queries that were provided by PDF documents. This is, of course, sad, but pages with these same texts after some time occupied the same, and then much higher positions. Probably because visiting a page gives more information about the user and his satisfaction with the site content to the search engine than downloading a file.
And since the user behaves well on the page, does not leave the site immediately and does not return to the search, the page begins to rank higher.
This is not exact knowledge – it is an observation))
Search Engine Marketing Results
Three years of work have passed unnoticed. It is time to evaluate the results and compare the actual indicators with those we predicted during the audit three years ago and generally evaluate the dynamics of search traffic.
Search traffic returned to the average historical level 15 months after the start of work.
The historical maximum of search traffic was exceeded 17 months after the start of work.
Since the start of work, in three years, traffic has grown by 865%. The forecast was exceeded by 223%.
Compared to the low point in January 2021, search traffic has grown by 1144%. We have overtaken my mother's friend's son in almost all metrics).
The minimum point is actually quite decent traffic, so this is not growth from zero).
Unfortunately, we are not allowed to disclose absolute values.
There are reasonable suspicions that, despite the excellent traffic results, they can be improved. New growth points are visible.
What about search traffic conversion?
And everything is fine with her. She is falling, but the number of target actions is growing.
The average conversion rate for search traffic is falling because the share of traffic from informational queries is growing.
But since information traffic is also converted, the number of requests increases.
Conversion of traffic by transactional segments is largely controlled by the invisible hand of the market, seasonality, macroeconomics, etc.
There is a constant struggle to increase the conversion of any traffic in the project, or rather, systematic work is being carried out. The main direction is to increase the conversion into requests through forms by constructing exciting order configurators for complex multi-component products.
Interestingly, at some point the number of requests from articles and technical documentation exceeded the number of requests from traffic coming to the directory.
Results – what is the main thing in the project
The main thing in the project is people)
Those who think, do, convince and agree, criticize and suggest, argue and invent.
Moreover, these are people who work on both sides, on our side (the contractor) and on the client’s side.
On our side, the project involves an account strategist, SEO expert, editor and authors, optimizers and programmers.
On the customer’s side – internet marketers, specialised marketers, applied specialists.
The results achieved are the result of the efforts of all these wonderful people.
It is noteworthy that at the beginning of the cooperation, corporate seminars on Internet marketing were held for the employees of the SVEL Group. This was a very correct decision, which helped people from both sides to quickly come to a single conceptual field. Especially in such a difficult subject as SEO.
Instead of an afterword
Search engine optimization is a large part of our work on this project, but it is still a part of it.
The project includes a decent share of contextual and targeted advertising, end-to-end analytics, regular research and consulting.
And it is very interesting to observe the synergistic effects of the systematic use of tools in a project. We will talk about them in other articles.
Author – Rublevsky Roman – Head of SEO project.
October 2023
