We analyze the mistakes of contextual advertising in the premium segment and increase the number of leads by 2.5 times in furniture production
What to do when there are few and expensive leads in the premium segment? How to avoid mistakes in contextual advertising and effectively train algorithms? In this case, we share strategies for combating non-targeted requests that eat up the advertising budget and the lack of data for training Yandex AI. We optimize the site and use microconversions. We win in competition by teaching algorithms to sell a service as a product and reach a VIP audience.
We analyze the mistakes of launching contextual advertising on your own
We were approached by a company that manufactures premium-class furniture based on individual projects, which also works with designers.
The customer launched the advertising independently. There were few orders and the production was idle without work. Therefore, they began to look for a contextual advertising contractor, read our case and turned to us.
We analyzed the customer's advertising campaigns and found critical errors.
Error 1. The “extended geotargeting” setting was enabled. Advertising was shown in irrelevant regions, which resulted in a waste of the budget. The client makes custom furniture and is limited to the region where he can go to take measurements.
Mistake 2. Using the same ads for all product groups reduced their effectiveness. For example, for queries about both a sliding wardrobe and wall panels, there was the same faceless ad “Custom-made furniture”.
Error 3. The absence of negative words led to impressions for irrelevant queries. For example, a customer's ad was shown for the query “buy a ready-made kitchen + in Hoff”.
Error 4. Low-quality sites in YAN were not minused. Most of the advertising budget was spent on mobile applications, where the bounce rate was 80-100%.
And the biggest mistake in advertising was the lack of configured analytics. There was no way to track the effectiveness of advertising campaigns. All these mistakes were critical and drained the advertising budget without the ability to analyze the results for further optimization.
That's why we started to launch contextual advertising again in our own way. standardtaking into account past results and the specificity of the niche.
We launched contextual advertising according to our standards. But it didn't give any results – for 100,000 rubles we got only 4 leads
At the start, we do not just launch all types of advertising campaigns that are suitable for the client's topic, it is important for us to set up offline (calls) and online conversion analytics (applications on the site). This makes it possible to evaluate the effectiveness of advertising campaigns, instantly respond and optimize them.
We set up tracking of advertising effectiveness to find growth points in the future:
Basic conversion tracking. We set up call tracking to track calls, and set up a goal in Yandex Metrica for requests.
Tracking user engagement, to understand where and why users lose interest.
We usually analyze the viewing depth, set up goals in Metrica for the time spent on the site (1.2, 4, 6, 8 minutes). But the client had an on-page landing page, so we tracked the scroll depth: 10%, 25%, 50%, 75%, 90%.
Collected all premium queries furniture. Worked out semantics for search campaigns in detail, segmented it into groups by service areas:
Brand queries (brand name + word “furniture”);
Requests about premium furniture (premium furniture + custom-made Moscow, elite custom-made furniture).
But since they have very little coverage, we also used broader semantic segments that can convert thanks to a good client site:
Individual sizes (furniture + custom-made, furniture production + custom-made + custom-made);
Furniture according to a project/drawings (furniture + made to order + according to the customer’s drawings, furniture + according to a design project + made to order);
Turnkey furniture (furniture + custom turnkey);
Furniture by room type (built-in furniture + for bathroom + made to order, veneer kitchen + made to order, sliding wardrobes + made to order);
Wall panels (veneer wall panels + made to order);
Hidden doors (hidden interior doors + made to order).
Created relevant ads for each group of queriesThe ads were made in such a way that they would best respond to users' search queries.
If the user types in the search “Hidden doors of individual sizes”, then in the ad we write about Hidden doors and unique trading advantages in this category of goods.
Based on our previous experience in similar niches, we taught Yandex to sell a service as a product. Usually, a Product Company is used for e-com (online stores), and in the service sector it is used less often.
For the Product Campaign, we created a Feed manually using our checklist. A feed is a file with information about products and services. Direct analyzes the file and automatically generates ads with product offers to display in search.
By turning a service into a product, we appear in positions where our competitors are not. And therefore we get traffic at a minimal cost.
We did everything according to the standards, but we encountered the fact that the algorithms did not have enough data for training, they “died out”. In the first 2 weeks, we received few leads – only 4, and they were expensive – 24,760 ₽ per one.
At the same time, the indicators of contextual advertising were quite good:
CTR (click-through rate for advertising): 9.23% on search, 0.88% on networks.
Bounce rate – the percentage of users leaving the site within the first 10 seconds: 24% (with the norm being 20-30%).
But the CR (conversion rate) was only 0.75%, which is low for this topic. This means that users liked the ads, but they did not leave a request.
Having collected the data, we started thinking about how to fix the identified problems: low conversion rate on the site, and insufficient data for training algorithms. We decided to start with improving the site so that it would be as prepared as possible for further relaunch of strategies.
The site was not bringing in enough conversions – we made some changes and increased conversion by 27%
We analyzed user behavior on the site. With high CTR and low bounce rates, CR was low. That is, users liked the ads, they clicked on them, the site also responded to their request, they stayed on the site, but did not leave a request. Therefore, using the numbers, we convinced the client to make changes to the site.
The client's website was initially aimed at a premium audience. The design conveyed the mood of the exclusivity of the offer, so there were no questions about the visual component of the site. But users did not reach the final goal – an order. We analyzed the usability of the site and the user path, and made small adjustments to shorten the user's path to conversion:
Added conversion elements to the menu header (contacts): phone number, WhatsApp icon, callback or request button, “Calculate cost” button.
Added anchor links to site sections to the menu.
Added a link to Yandex.Maps and a “get directions” button for mobile users.
We redesigned the Application Form, added a title, description, and call to action.
Added a block with a video and a call to action.
The updated website immediately showed itself well. Conversion for the second half of the month increased by 27% (CR was = 0.75%, became = 0.95%), and the cost of a lead decreased from 24,760 ₽ to 17,363 ₽. While the updated website was being tested, we were already solving the main problem.
The main problem in premium topics is the lack of leads for Yandex training. We optimized advertising based on user behavior on the site and increased the number of leads by 45%
When training Yandex algorithms in the premium segment, there is often a problem of insufficient number of macroconversions (orders or calls) for the effective operation of automatic strategies. Yandex algorithms require within 1 advertising campaign at least 10 macro conversions per week for training, and with less data they “die out” and stop showing ads.
To accelerate advertising campaigns and collect data when there is a lack of macro conversions, we used optimization for micro conversions, which were initially specified in the goals when setting up analytics.
Microconversions are intermediate user actions on the way to achieving the final goal (macroconversions).
We analyzed user behavior on the site to find a relationship between changes in the number of micro-conversions and macro-conversions. If the growth in the number of micro-conversions coincided with the growth in orders, we found a correlation and could use this data to optimize advertising.
As a result, we identified two micro-conversions for Search and YAN, which led to macro-conversions within the project:
Clicking on the button to open the application form. Indicates the user's interest in further interaction.
Active time on the site is 2 minutes.
Thus, we configured Yandex Direct so that it would bring users to the improved site who are more likely to perform these microconversions and subsequently make an order. As a result, the number of leads in 2 months increased by 45%, and the cost decreased to 13,129.
There are still few leads – we combined search campaigns in a package strategy, Yandex learned and brought in twice as many leads
After we switched campaigns to micro-conversion optimization, Yandex algorithms had more data, but not all campaigns had 10 micro-conversions for training. So we decided to test the following hypothesis: we combined several search campaigns into a Package Strategy. It is used for campaigns that bring few conversions. Campaigns remain separate, but they have a common advertising budget, common strategy settings. And Yandex starts counting conversions for all these campaigns in total. This way, the algorithms work more stably and do not “die out”.
The Package Strategy combines 4 search campaigns:
Priority categories (wall panels, interior doors);
By room type (bathroom, kitchen, bedroom, dressing room);
Individual pieces of furniture (cabinet, chest of drawers);
Competitor queries (Competitor names).
And the system began to count conversions not for each individual campaign, but in total for all four.
We used the “Maximum Conversions” strategy to increase the share of impressions to those users who are most likely to perform the target action on the site.
And as a conversion, we indicated the goal in Metrica – “Unique leads”.
As a result, the use of the package strategy helped to increase the number of leads by 2 times. And the cost, CPL decreased from 13,129 ₽ to 8,785 ₽.
But everything broke down again – the New Year holidays dried up the algorithms. Yandex was manually sent and the flow of requests was restored
The New Year holidays arrived, and from the second half of December, a drop in leads began. There was especially little data for training automation during this period, so after the holidays ended, they began to revive the advertising campaigns that had dropped the most.
The Product Campaign and Campaign Wizard were initially set up for pay-per-click. But after the New Year holidays, we saw that the algorithms in these campaigns began to bring more traffic from the ineffective YAN. There were impressions, clicks, but no leads. Therefore, we decided to redirect the algorithms to the Product Gallery and Search.
In Yandex smart campaigns, there is no way to disable one of the platforms. Therefore, in the Product campaign, creatives (advertising banners) for display in YAN were removed, leaving only the headings and text for the ads.
Yandex algorithms saw that there were no creatives, which meant there was nothing to show in YAN. Therefore, ads began to be displayed more in the Product Gallery and Search.
At first, the Campaign Wizard was launched completely on automatic settings. It did not bring any results. Therefore, we set up a campaign by key phrases, manually directing the algorithms where we needed.
Before optimization, 1 lead came in the Campaign Wizard per month for the holiday month, after optimization for the same period – 5. In the Product campaign, before optimization, 2 leads came in the month, after – 6.
After a few months of decline, the total number of leads increased by 75% by February. And the cost of a lead decreased from 12,179 ₽ to 9,849 ₽.
Things aren't so bad anymore, but we want more. We tried other ways to reach a premium audience.
In search of growth points, we started looking for new sources of premium audience. Yandex provides the ability to target different target audiences. And we decided to test different tools to find exactly those users that are interesting to us.
We have configured the advertising account for specific audience solvency groups.
In the RK settings, you can specify the priority by the level of solvency. There are 4 segments:
Top 1%;
Top 2-5%;
Top 6-10%;
The remaining 90%.
They gave priority to 1% and 2-5% of the most solvent audience. And made a downward adjustment of 6-10%. But this did not give any result. Perhaps this happens because the solvent audience can delegate the task of repair to contractors. And the designer who selects furniture for his client may not be included in the audience of the most solvent audience in Russia.
Setting up with adjustments for solvency:
We tried to reach the desired target audience using geolocation in Yandex audience:
In Yandex Audiences, we drew a polygon – clearly defined areas on the map where the premium audience could be. We set ourselves up for the following locations:
Residents of premium areas (Barvikha, Rublevka, etc.).
A premium residential complex under construction, where those who were considering buying there came 1-3 times.
Screenshot with example polygons:
And then we used increasing adjustments for these audience segments in the Search Campaign and YAN.
Uploaded a database of designers to Yandex Audiencewhich the client provided to us. Based on this segment, we created Look-a-like audience and made upward adjustments to it in YAN.
The system found users who behavioral characteristics are similar to the busy audience and showed our ads.
After connecting the testing ground and creating a Look-a-like audience, the number of leads increased by 28%.
We found a new point of sales growth – we implemented a quiz on the site to collect contacts and caught up with those who did not complete it with retargeting
In search of ideas for increasing website conversion, methods for collecting and segmenting audiences, we decided to test Quiz.
We thought out our structure depending on the directions on the site and the interests of the client. We implemented the button from which most applications came. Initially, there was a form. There are many directions, but one button, so we created such a quiz structure and the client's path that leads to the necessary goal.
The quiz consisted of several stages.
First, let's find out which segment is sending the request:
Next, we find out what stage the customer is at and identify the needs:
And then the user selects a category that is on the site:
To motivate filling in, discounts, bonuses, and gifts were offered at the end:
Afterwards, we collected an audience of those who did not complete the quiz. And we returned them to the site by increasing adjustments for this audience in search campaigns. And those who completed the quiz to the end, but ultimately did not become a buyer, were returned using retargeting, offering an additional discount. As a result of the quiz, conversions increased by 30%, and the cost of a lead decreased from 9,921 ₽ to 7,449 ₽.
As a result, in 10 months of work, we increased the number of leads for premium furniture by 2.5 times and reduced the cost of attracting a lead from 24,760 ₽ to 7,449 ₽
The case demonstrates that even in a narrow premium segment niche, where leads are few and expensive, contextual advertising can become an effective tool. Due to work with the site, setting up detailed analytics, as well as constant testing of hypotheses in Contextual advertising.
Over 10 months of work, we increased the number of leads by 2.5 times, and reduced the cost per lead (CPL): from 17,363 ₽ to 7,449 ₽, that is, 3 times.
The company's case demonstrates that even in the premium segment, contextual advertising can be effective if properly configured and optimized. If you have questions about how advertising works, come to us for a detailed audit. We will identify errors and direct the algorithms in the right direction.
If you need to increase your conversions, order our deep audit of contextual advertising for 0 rubles on makodigital.ru
How will this be useful?
We will find technical errors in campaign settings and segment them by danger level: minor, significant and critical.
Let's check the statistics and find the reason for the increase in the cost of circulation and the fall in profits.
Important! We will recommend positioning, content and development (conversion, usability, functionality) for the site.
We will develop a contextual advertising strategy with new growth points.
