Who we are

The digital marketing company PLATZKART works from 2004. Helping companies with complex (and not so complex) products and services in the sectors of В2С and В2В develop and implement marketing strategies, improve the effectiveness of marketing and marketing investments and sales. In addition, we make marketing clear at all stages for business owners.

We have experience in niche areas:

  • real estate,
  • agricultural machinery,
  • fire safety,
  • videoconferencing,
  • relational databases,
  • medical devices,
  • transportation companies,
  • and other high or low competitive markets.

Our company works with projects using the SOSTAC methodology. The methodology allows with \(80\%\) probability guarantee the success of the promotion of goods and services on the Internet. We rely only on quantitative data, thoroughly study the market, prepare and implement promotion strategies, use the accumulated data on the project for its optimization. project to optimize it.

Within the framework of the project we use the following tools to implement the strategy:

  • naming;
  • design: corporate identity development, website design;
  • creating a website or landing page to sell goods and services;
  • seo-optimization to present the site in search engines and to participate in the auctions of advertising systems;
  • setting up independent analytics to optimize sources of SEO promotion and advertising channels, as well as for reporting to our clients.

PLATZKART is highly recommended for further cooperation by \(90\%\) of our clients. We can provide their contacts upon request.

To find out more about our work email us:

Why do you need this case

The story is useful for companies that face the challenge of promoting a complex product in highly competitive markets with minimal budgets in a crisis. With minimal budgets in crisis conditions. Let's tell you what we did and how we did it during \(9\) months and what results we have achieved.

About the client

Client - a branch of the Trade House MTZ-North-West in Krasnodar. Official distributor of Minsk Tractor Plant ("MTZ") in the Krasnodar region and the Republic of Adygea. Are engaged in sales Belarus tractors and agricultural machinery.

The company supplies production to farms, collective farms, agricultural holdings and other enterprises engaged in agriculture.

Task

Get N-number of requests from Yandex Direct with minimal budget. No specific KPIs were set for the number of requests and cost. However, the following were defined:

  • monthly budget;
  • broadcast regions;
  • keywords involved in targeting (semantic core).

What was at the beginning

The company has its own site, created by our company earlier. The site had a clear structure and was prepared for SEO-support. This is important, because when you start advertising, the service evaluates the landing page by many parameters and if it is not of high quality, the advertising company will receive fewer impressions.

Project Challenges

  1. The company, after the "covid" times was not ready to to spend large budgets on attracting customers from paid channels.

  2. It is necessary to connect the CRM system to "Yandex Metrics" for transferring data on the quality of calls, but the company "TD MTZ" has had internal problems with the connection of "1C™" and CRM system. This integration was handled by a third-party company. As a result, before the launch of the project the problem could not be solved and it was decided to analyze the results of campaigns on the the results of campaigns at the level of inquiries, rather than at the level of "quality calls".

  3. Since KPIs have not been set, before launching advertising campaigns it is necessary to determine the theoretical possible number of requests to the company, as well as to know the "fork" on the budget to understand whether the established budget will be enough.

  4. It is important to understand what number of hits is possible to get, as it depends on the setting of analytics for training automatic strategies of advertising services.

Preparation

In our work, we pay great attention to the preparatory phase.

First, you need to conduct a Situational Marketing Analysis (SMA). The SMA is the foundation of a future customer acquisition strategy. Without a foundation, there is no guarantee that the strategy will work. It is necessary to understand with whom we will be competing, in which regions the competition is the greatest, what UTPs competitors broadcast in ads. It is also necessary to predict the theoretical volume of referrals and the necessary budget for setting KPIs.

The situation analysis consists of the following items:

  • market analysis,
    • semantic core analysis,
    • competitor analysis,
    • predictive analytics,
  • analytical systems audit,
  • site analysis.

Let's look at each point in more detail.


Market Analysis

The market analysis consisted of the following items:

  • semantic kernel analysis,
  • competitor analysis,
  • predictive analytics.

Semantic core analysis

Semantic kernel analysis provides an understanding of how many keywords are shown in advertising systems. Approximate budgets based on data from advertising services, as well as predict the possible volume of referrals to the company.


General structure of the core

For the tractor ad campaign, they collected 239 keywords. Keywords were selected by frequency from the service "Yandex word selection", but this does not mean that all keywords from our list will be shown in YaDirect. Therefore, it is necessary to collect statistics on displays and cost per click from "Yadirekt". As a result, we got a table for further analysis. The first five rows of the initial table are shown below.

Table: example of semantic kernel

Phrase Group Shows for a year Click price
mtz price mtz cost 18271 19.9
mini tractor buy Krasnodar mini tractors 7877 42.3
buy a mini-tractor + in Krasnodar region mini tractors 5491 43.6


A regular table, but already at this stage, useful information can be obtained. Let's consider the general picture of the core.


Table: general statistics on the core

Shows for a year Shows for a month Av. price per click Mdn. click price Keys with shows Zero
82628 6886 39.4 35.5 131 108


Conclusion

We can see the differences between the median and average cost per click. This means that the core contains mostly keywords below the average cost per click. Let's check the assumption on the graph.

Graph: distribution of keywords by value of the click

Indeed it can be seen that most of the keywords from our core are grouped by cost per click in the range of 25₽ to 40₽.

We also learned that the spread around the average click price sd≈17 ₽, so the cost of a click can differ from the average value in the greater or lesser side by a specified amount.


Keywords with zero impressions

It often happens that the approved semantics, after collecting frequencies in an advertising service, contains keywords with zero impressions. These words should be excluded from advertising campaigns, as they will not be shown and it will be easier and faster to work in the process of analysis.

Table: example of keywords with zero impressions

Phrase Group Shows for a year Click price
Belarus mtz sale of belarus tractors 0 17.8
Belarus tractor price mtz Belarus price 0 30.0
Belarus mtz tractor sale of belarus tractors 0 26.0


Keyword distribution

Let's consider the distribution of frequency of impressions by groups, thus we will find out which groups are expected to receive more hits. This is useful to take into account when you have different products in your advertising campaign. For example, credit cards for individuals and legal entities or different types of agricultural equipment, etc.

The graph below is shown without group names due to NDA policy.

Graph: distribution of keywords by impressions in the context of groups

Conclusion

We found out that at least four groups have keys with high coverage. These groups are expected to receive more clicks, which means that these groups will be the first to receive requests. We also expect these groups to spend their budget faster.

The same graph can be created for the distribution of the cost per click, thereby finding out which groups contain the most expensive keywords and take the groups under control.


Competitors

The methodology for selecting competitors in Yandex's paid rendition is based on the hypothesis that the more impressions a competitor's domain has for the selected keywords (i.e. the greater the share of coverage), the more traffic the domain receives, which means that they are our competitors on the web. This does not mean that the domain can be our direct competitor. Direct competitors are companies that provide the same services and products as your company in the same region.


Competitors and outreach

In our semantic core, we found 52 competitor sites. The graph shows the top 10 competitors and their share of coverage (due to NDA we can only show the first three competitors on the graph).

Graph: competitor sites and coverage

Conclusion

We see the top three leaders in terms of coverage, which means that these are our main competitors in the "YaDirect" auction. Note that the direct competitor is only one "am-volga.ru". The site "spec.drom.ru" is more of an indirect competitor, but, generally speaking, it doesn't matter, as our ads will mainly compete for displays with sites from the TOP 10.

By the way, if the site is engaged in self-promotion, you should take it into account. We recommend placing yourself on it to attract potentially interested clients.

Let's see how competitors are distributed by groups. In the table we will see the number of sites and ads for each group, so we will know the most competitive group.

Since this is an example, we will show 3 groups of 23 and their averages.

Example table: number of sites and ads by groups

Group Number of sites Number of ads
mtz sale 22 29
farming tractors 21 26
mini tractors 15 22
On average sites 14 —
Average ads — 16


Conclusions

On average, we have 14 competitor sites and 16 ads per group. The variation across sites is ± 4 . High competition is observed in the first two groups from the table, so we expect high cost per click in these groups.


Distribution of competitors by position

Let's test the hypothesis whether there is a dependence of coverage on the average position. As a rule, the higher the average position, the more traffic a site receives.

Conclusion

The hypothesis was not confirmed in this niche. This conclusion is confirmed by p-value=0.97 (with the norm of p-value=0.05 i.e. at this value the hypothesis is tested). So in this niche it does not matter what position the ad is on.Traffic can be collected from both the upper and lower block in the search engine of Yandex.

Also note that the bulk of competitors receive a small amount of traffic, according to our semantic core, which gives us the opportunity to collect more than 80% of the traffic, which means that we have a chance to get the most out of our advertising campaign.


Regional analysisв

Let's take a look at what regions competitors are broadcasting ads in.

Example table: frequency of using regions in ads by competitors

Domain Region Frequency Coefficient
spec.drom.ru Krasnodar Region 200 3.8
ironfield23.ru Moscow and Region, Krasnodar Territory, Republic of Adygea, Rostov Region, Republic of Crimea, North Caucasus, Abkhazia, Donetsk, Luhansk Region, Republic of Crimea, North Caucasus, Abkhazia, Donetsk, Luhansk 129 2.5
am-volga.ru Voronezh Oblast, Nizhny Novgorod Oblast, Penza Oblast, Republic of Tatarstan, Samara Oblast, Saratov Oblast, Ulyanovsk region, Volgograd region, Rostov region 121 2.3
kat-russia.ru Russia, Abkhazia, South Ossetia 108 2.1
wiweb.ru Russia 69 1.3


Let's find out which regions are the most competitive.


Example table: frequency of using regions in ads

Region Frequency Coefficient
Russia 506 0.51
Krasnodar Region 421 0.42
Rostov region 322 0.32
Abkhazia 237 0.24
Voronezh region 193 0.19


Conclusion

By entering the synthetic coefficient in the corresponding column you can quickly determine the most popular region - Russia. Krasnodar Krai and Rostov Region are in second place. Increased competition is expected in these regions. When creating advertising campaigns it is necessary to take into account the number of regions in which it is planned to broadcast ads.


Analyzing competitor ads

By collecting data on competitors' ads from the search engine, we can analyze them. We will find out what words are used in the headlines and texts of ads most often. What they emphasize. From this data it will be clear how to separate from competitors or what should be used in the headlines necessarily.

Example table: analyzing announcements

№ One-word words Frequency one-word Bigram's words Bigram frequency
1 tractor 23 buy a tractor 6
2 buy 13 mini tractor 6
3 tractors 13 agricultural market 5
4 leasing 11 mtz tractor 5
5 mini 8 agricultural marketplace 4
6 credit 7 runmax rmx 3
7 buy 7 profitably tractor 3
8 mtz 6 wheeled tractors 3
9 runmax 5 buy a mini 3
10 apc 5 buy new 3


The table shows the frequency of one-word and two-word keywords from competitors' ads. Having studied the whole table, let's draw conclusions on what competitors emphasize and how they compose their ads.

We can also see keywords in terms of groups and competitor sites.

Example table: analyzing ads by groups and sites

Group Domain Words Frequency
mtz sale 1express-credit.ru loans 1
sale of mtz 1express-credit.ru urgent 1
tractor on credit 1express-credit.ru loans 1
tractor credit 1express-credit.ru urgent 1
where to buy a mtz DboRu.ru freebies 1
where to buy a mtz DboRu.ru board 1
where to buy a mtz DboRu.ru ads 1
tractor sales DboRu.ru freebies 1
tractor sale DboRu.ru board 1
tractor sales DboRu.ru ads 1


Analyzing the Universal Trade Offer of direct competitors on the site

We analyzed the sites of direct competitors. As a result, we received the following table with data.

Example table: analyzing competitors' landing pages

Competitor UTO Advantages SEO desk SEO mob Compliance principle Cost
"AgroMarket Volga "UTO 5/10 7/10 7/10 6/10 10/10 0/10
Competitor «N» … … … … … …
"TD MTZ" 5/10 9/10 8/10 7/10 NA 5/10

Summarized quality scores for all competitors, which gives insight into how things are going in the niche.

Example table: results of the analysis of competitors' landing pages

— Com 1 Com 2 Com 3 Com 4 Com 5 Com 6 Com 7 Com 8 Com 9 Com 10
Average score 4 4,8 5,6 4 3,3 6,2 7,8 6 4,8 7,3
Sum of points 24 24 28 20 20 37 39 30 29 44
Completeness 100% 83% 83% 83% 100% 100% 83% 83% 100% 100%
Quality Rate 17% 33% 33% 17% 17% 67% 50% 33% 17% 67%
Place "Sum of Points" 2 6 7 8 3 9 1 11 4 9
The place of "Complexity" 3 9 10 5 1 7 2 10 4 8
Place “QR” 2 2 3 5 2 4 1 6 2 4
Points for "Total points" 91 55 45 36 82 27 100 9 67 11
Points for "Complexity" 80 20 10 60 100 40 90 10 63 13
Points for "Quality Rate" 83 83 67 33 83 50 100 17 75 25
Total Points 861 501 389 426 876 355 970 109 674 144
Place in the Rating 3 5 7 6 2 8 1 10 4 9


The resulting data is used to improve the company's own website.

Conclusion

  • We collected data on the ads of com-competitors. We drew conclusions about how and what we would write in the ads.
  • Comcurrency website analysis data was used to improve the company's website.


Predictive analytics

In this section we will calculate the actual achievable market volume (Serviceable & Obtainable Market). We need to understand the approximate number of requests we can get at the level of generated demand. This will help us to set a KPI to which we are going to be guided.

The approximate number of hits per month is determined by the Poisson distribution. It is calculated by two methods:

  1. Based on the advertising budget and average click price.
  2. Based on the frequency of keywords shown and average ctr.


Calculation based on budget

Based on the monthly budget, according to calculations, with ≈90% probability of getting between 27 to 50 requests per month, on average ≈40 . Provided that the average click price CPC≈39₽ and Comversion from click to conversion CR≈4% (data on CR are taken from the client's "YaMetrics").

Graph: range of receipt of requests based on budget


Calculation based on frequency of impressions and ctr

First, let's calculate the number of clicks based on the average frequency of monthly displays. We need to get the average value of clicks.

Assuming that the average ctr≈10%, and the average number of impressions per month for the approved keywords ≈7000. Then with ≈90% probability we will get from 640 to 750 hits on ads from search per month, on average ≈700 .

Graph: range of clicks received

Then with ≈90% probability we will get from 15 to 41 requests per month, on average ≈28.

Graph: range of receipt of applications

And let's calculate the probability of getting sales based on the average value of conversions. According to the Roistat research data, the average conversion rate from conversion to sale in B2B is ≈17%, then with ≈90% probability to get from 1 to 10 sales per month, on average ≈5.

Graph: range of sales receipt

Conclusion

The main conclusion is that it is possible to get sales from a semantic core.

СLet's form a "fork" of the budget. Let's calculate it based on the frequency of impressions and average ctr.

90% probability Click price Trunctions Budget month Conversions CPA Clients CAC
Lower boundary 52₽ 640 33 280₽ 15 2 219₽ 1 33 280₽
Medium 52₽ 700 36 400₽ 28 1 300₽ 5 7 280₽
Upper boundary 52₽ 750 39 000₽ 41 951₽ 10 3 900₽


Now we know:

  • expected range for the budget excluding VAT,
  • range of appeals,
  • sales range.

To be considered:

  1. The click price is influenced by many factors from the number of participants in the auction at any given moment to the quality of the site and ads, display regions and device type, plus another ≈10,000 parameters that are taken into account by the auction system. parameters that the system takes into account when conducting the auction. Therefore, the final budget may vary.

  2. These calculations should be understood in terms of cohorts. For example, select a cohort of people who contacted companies in August. From this cohort people can Comversion to customers during the year, however, if you consider how many people from the August cohort converted to customer status during the year, you can identify the Comversion coefficient for this cohort, as well as the total number of customers their value and other metrics. However, such calculations require communication with the CRM system.

Audit of analytical systems

An analytics audit is the preparation and validation of analytics systems to receive and analyze incoming traffic. It includes the following items:

  • analyzing the structure of analytics;
  • correctness of meter installation;
  • which goals are tracked in YaMetrics;
  • which targets are tracked in YaDirect;
  • what segments are set up and whether they are applied in YaDirect;
  • пWhether "end-to-end analytics" (ROMI or DDR analysis) are used.

We will not describe each point in detail here. We'll just outline the key points.

  • It was decided to use "YaMetrics", as the source of traffic will be "Yandex Direct.

  • According to the results of the sieve analysis we found out that the number of calls to the company is expected to be in the range from 15 to 40. . This is not enough for training automatic strategies. Therefore, it is necessary to make additional settings for tracking micro and macro targets.

  • There is no end-to-end analytics. That is, there is no possibility of transferring data from the CRM system to YaMetrics. This means that we cannot optimize advertising campaigns at the level of "quality conversions". "Quality referrals" - "50% -qualified** clients" or "qualified clients" that meet the "four P's" criterion or the "BANT" technology. In other words, potential customers who are ready to listen to the terms of sale of the company's products. However, according to the results of the advertising campaign, we have found a way to check how much quality traffic comes from the RC. We will talk about this in a separate case study.

Website audit

The main task is to check the readiness of the site to receive traffic by the following parameters:

  • SEO website optimization:
    • technical optimization,
    • usability,
    • commercial factors.

The obtained data were compared with the Comcurrencies quality table. We made made the necessary corrections to the website.


Implementation

In this part, we will describe how the implementation was carried out based on the of the situation analysis.

Abstracts and hypothesis

Let us formulate a number of statements from the situational analysis:

  • The budget specified by the client falls within the fork in the budgets calculated by us. So we can launch advertising campaigns.
  • The semantic kernel we have selected has the ability to bring in customers at the level of generated demand.
  • After grouping the keywords, we have a number of groups that will bring in customers in the first place:
    • minitractors,
    • group A,
    • group B …
  • We are dealing with a strong competition, especially considering the advertising budget. Therefore, it is necessary to choose the right regions and pay special attention to the composition of ads.
  • Taking into account the forecasts on achievement of goals on references it is necessary to carry out a number of adjustments in analytics systems for training of automatic strategies in advertising systems.
  • A number of works should be carried out on the site to improve the parameters, which will help to participate in the auction on more favorable terms.

The hypothesis in our case is very simple. From a statistical point of view, the null and alternative hypotheses are as follows:

  • \(H_0\) - Yadirekt advertising system will not be able to generate conversions, and those that will be a fluke.

  • \(A\) - the advertising system steadily generates conversions and operates at a stable capacity./p>

Structure of advertising campaigns

The structure was guided by the following parameters:

  • It is necessary to collect as much data as possible on macro and micro targets to train automatic strategies. Therefore, it is not possible to split semantics into small advertising campaigns.

  • We have a limited budget, so we work with a limited number of keywords. We pay special attention to keywords with high frequency.

Bottom line:

  • We got one advertising campaign consisting of 23 groups.
  • The keywords are grouped by meaning.
  • The number of keywords in a group varies from 2 to 22.
  • In each group from 2 up to 4 ads depending on the total frequency of keywords in the group. The more ads in the group, the more time is needed to gather statistical information to determine the best ad.

Implementation of analytics

In this project, the analytics scheme is as follows:

The main task is to take into account all sources of contact with the company.

All data from the site and third-party services are collected in Metrica. "Metrica and YaDirect exchange data. Data from Metrica and Direct are transferred via API to RStudio, where we perform basic analytics on the effectiveness of our strategy.

RStidio has a wide range of reporting functionality. Reports can be transferred to Google Spreadsheets, sent via email or Telegram™. Or transfer data to various BI reporting systems. You can also set up reports on your own servers.

In Metric, you set up goals:

  • handling the forms on the website;
  • sending or copying company emails;
  • tracking incoming calls;
  • access through the widget;
  • plus a number of microtargets have been set up.

As we already know, data on micro-objectives are collected faster and there are significantly more of them. By conducting correlation analysis together with multivariate analysis of variance (ANOVA) we can determine which micro goals precede the achievement of macro goals. Thus, by setting up "YaDirect" to receive micro-targets we train the system, which will eventually lead to macro-targets.

In our advertising campaigns we try to use automatic strategies, as they show the best results compared to manual strategies. The thing is that in auto mode the system takes into account tens of thousands of factors, and a specialist is physically unable to take into account such a number of factors when managing campaigns in manual mode..

Challenges in implementation

After data collection, a series of analyses were conducted to determine the antecedents of the macro objectives to the micro objectives. However, a problem was encountered. According to the results of the analysis, none of the macro-objectives correlated with the micro-objectives. Then we decided to create a special goal that would collect data from all sources of appeals. This step proved to be effective in the end.

Also in the process of project implementation there were months in which, for various reasons, data were not collected from some sources of contacting the company, which affected the total number of goal achievements in the statistics.

Results

Consider the results of our work.

Graph: overview of requests for the whole period

The graph shows that in August and September we had problems with the collection of data on requests. Some channels were not tracked for various reasons. In October, the analytics have been set up again.

Consider the density of the distribution of requests.

Graph: density of distribution of requests to the company

We see that advertising campaigns more often did not receive requests. This is to be expected, since given the specifics of the product we should not expect to receive requests every day. Most received one request per day. Less often 2-3 requests were received. It is quite expected F - distribution of requests in this niche.

Consider the number of applications by month.

Graph: number of requests by month

Note that our calculations on the number of requests per month fall within the calculated range, except for inactive months. On average, we received 30 applications per month. Our projection is 15 to 41 applications per month.

Graph: circulation of leads with accumulation

From this graph, you can quickly determine which months had the most or least referrals.

As a result, our system of attracting traffic based on situational analysis worked. Requests from the advertising campaign on the search came steadily and they were not random. Thus, we accept the alternative hypothesis formed earlier.

Next, let's analyze our traffic attraction system from a six sigma perspective.

Six Sigmas

Each traffic source generates calls within certain boundaries. The boundaries are called "six sigma boundaries". They are calculated with the help of the "Shuhart Comtroller Card Standard" and statistically significant indicators (sales, conversions, registrations, clicks, etc.).

Knowing the limits of six sigmas:

  • know the capacity of the traffic source i.e. within what limits we can expect daily sales / registrations and so on;
  • we know how much control we have over the system;
  • It becomes clear what KPIs should be set for traffic managers and sales managers;
  • We use the data to build predictive analytics to track trends, which allows us to make timely changes to advertising campaigns.

We will not describe the calculation methodology in detail; we will write about it in a separate case study. Let's just look at the graph and draw conclusions.

The red dotted line on the graph shows the upper and lower Comtroller limits. They show the range in which our system of attracting requests works. In our case, the range is from 0 to 4 requests per day. In other words, this is the capacity of our system. We do not know whether it is good or bad, but this is how our system works. You should not worry if on some days we do not receive requests. This is the norm for this system of traffic attraction.

However, you should pay attention to the days when the number of bids goes beyond the upper limits. Such days should be analyzed and tried to scale. In our case, such outputs were right after the launch of the company. Most likely there were additional tests on sending requests from the site by site developers. We also noticed outages in May and in Comtse of the year. As our client informed our client, there is always a rush at the end of the year and this is the norm.

This analysis is especially good for companies whose requests come mainly from physical outlets: catering, real estate, auto and other niches.

What's next

Since we have a problem with determining the quality of requests, the next step is to develop a method of determining the quality of requests without linking to the CRM system. We already know how to do this. The method is based on text analysis of incoming calls, requests through widgets and requests from websites. Third-party systems are able to determine the sources of incoming requests, so we will link our campaigns and sources, and then a lot of code and as a result improve the effectiveness of advertising campaigns.

That's all. Thank you for your attention!

Want to run advertising campaigns based on our system? Write to email: