VK Advertising 2026: a complete guide to setup, audiences and optimization

VK Ads in 2026 is not myTarget with a different interface. This is a separate product, which in two years has grown to a level where it has look-alike by pixel, auto-bidding, conversion optimization and a normal API. I’m not looking at theory, but at practice: how to set up where CPL is normal and where it’s not, and what to do with ERID so that ads are not rejected.

I have been working with VK Ads since the closure of myTarget - since 2023. On the current six projects, VK takes up 30-60% of the performance budget, depending on the niche. The figures in the article are the median for these projects plus correspondence with eight colleagues with comparable volumes.

The main misunderstanding about VK Ads: they think that this is “the second Direct, only worse.” It’s more correct to think that this is the first and only tool for performance for audiences that Yandex does not reach.

1. What has changed: myTarget vs VK Ads

myTarget officially closed in 2024. The account stopped accepting new campaigns, accounts were transferred to VK Ads. At first glance, it’s just a rebranding. In reality it is a different system.

myTarget had “flat” targeting: you selected interests, set a CPC or CPM bid and launched. Automation was minimal. VK Ads works according to the logic of Facebook Ads from 2019: the algorithm itself decides who to show the ad to in order to achieve a given goal. You set a goal (conversion, lead, installation), budget and CPL that suits you - the algorithm works.

What else has fundamentally changed:

• Structure campaign → group → ad (previously it was “campaign → ad”, without group level)
• Look-alike based on pixel events - not just “similar to the base”, but “similar to those who bought”
• Conversion optimization - the algorithm receives data on purchases through the pixel and begins to look for similar ones
• A/B tests at the group level - several ads within a group, VK itself distributes impressions to the winner
• Frequency cap works better by default - audience burnout is slower

2. Account structure: campaigns, groups, announcements

The account is located at ads.vk.com. Three levels of hierarchy:

Campaign - top level. Here we select a goal: conversions, traffic, leads, awareness, installations. The goal cannot be changed after creation - therefore, we create a campaign for a specific task, not a “universal” one.

Ad group — this is where the audience, budget, schedule, placements live. One audience hypothesis = one group. If we test three segments, there are three groups. If we don’t mix different audiences in one group, we’ll lose the understanding of what works.

Announcement — format, creativity, text. Within one group we keep 2-4 ads for A/B testing. VK itself optimizes the distribution of impressions between them - after 7-10 days we see the winner.

A typical mistake for beginners is to create one campaign, one group and stuff all the audiences there. Analytics becomes unreadable, the algorithm gets confused. Rule: one hypothesis - one group.

3. Audiences: targeting, Lookalike, retargeting

Audiences in VK Ads have three layers. I understand each one.

Direct targeting — interests, keywords, demographics, geo. Interests in VK work based on user behavior within the platform (groups, posts, videos). Keywords - search queries in VK and related services. I usually start with a combination of interests + keywords, narrowing the audience down to 300-500K to be economical.

Look-alike - the most powerful tool. We load the database (phones/emails) or create an audience from pixel events (who made a purchase, filled out the form). VK builds a model and finds similar users. From practice: look-alike in terms of conversions works better than look-alike in terms of traffic - CPL is 25-40% lower.

Retargeting can be configured via pixel. We create audiences based on events: “visited the product page”, “added to cart”, “started checkout”. Each segment has its own ad. The budget for retargeting in my projects is 15-25% of the total, CPL is usually 1.5-2 times lower than that of cold traffic.

4. Formats: universal post, carousel, clips, lead form

Not all formats work equally in different niches. According to numbers from my projects:

Universal recording — text + picture or video up to 3 minutes. CTR: 0.6-1.2%. Works everywhere as a test. The cost of a subscriber through this format is 60-120 rubles in niches with an audience of 500K+.

Carousel — several cards with different products or offers. CTR: 0.9-1.8% in e-com. It works worse in services - people don’t choose between “service A” and “service B” through cards. I use it for catalogs when there are 3+ products.

Clips - short vertical videos. CTR: 1.5-2.8%, which is 40-60% higher than regular posts. The audience watches clips on VK Video - this is a separate placement with a more loyal cost of display. The downside is that you need high-quality vertical video creative. Production by a contractor - 15-30K ₽ per video.

Lead form — a form right inside VK without going to the site. CPL is 20-35% lower vs landing page traffic in service niches (dentistry, lawyers, repairs). Minus - the quality of leads is worse, because the person did not read the landing page and did not “warm up”. The conversion from a lead to a deal is approximately half as much as from a landing page. Count according to unit-economics, and not just according to CPL.

5. Autostrategies: when to enable and when not to

VK offers three bidding strategies: “Minimum price”, “Target price” and “Maximum price”. The rules are simple, but they are constantly violated.

At the start - only “Minimum price” with a budget limit. The algorithm learns from your data. If you set the “Target price” from scratch, he does not know what they want from him and spends the budget in an inappropriate way.

Auto strategy “Target price” is turned on when the group has accumulated 30-50 conversions. Before this, you are simply burning money on training with the wrong signals. I usually wait for 50 conversions - it takes 10-20 days with a budget of 30-50K ₽/month per group.

When automatic strategies are not needed: a very narrow ICP (B2B product for a specific position), retargeting a small audience (up to 20K), a situation where each conversion is checked manually. In such cases, manual bid management gives more control with less CPL variability.

6. ERID and labeling: step-by-step checklist

From 2023, advertising in Russia requires labeling - ERID (External Register ID). VK Ads checks markings during moderation. Ads without an ERID or with an incorrect ERID are rejected. Everyone complains about this.

Step by step order:

1. Register an advertiser in ORD (advertising data operator). The largest ones are ORD VK, Yandex ORD, Media Scout. VK Ads is integrated with the VK ORD directly - through the account.
2. Create a “creative” (creative) in the ORD - upload text and image/video before publication.
3. Get ERID - a string like “2VtzqxQsTmj”.
4. Enter ERID in the ad. There is a special field in the VK Ads account - we paste it there. Additionally, it is customary to add the note “Advertising” and an indication of the advertiser to the ad text.
5. Submit statistics to ORD monthly. VK does this automatically if you have connected ORD VK.

A typical mistake: receiving ERID after publication, “retroactively.” This is a fine. The order is ERID first, then announcement in progress. Another mistake is to use one ERID for several different advertisements. Each unique creative = its own ERID.

7. Optimization: what to look at at what stage

Different metrics are important at different stages of a campaign. The mistake is to watch the CPL on the second day of launch and draw conclusions.

Days 1-3: We look only at CTR and cost per click. If the CTR is below 0.3%, the problem is with the creative or an irrelevant audience. If the CPC is abnormally high (3-5 times higher than expected) - most likely there is a narrow audience or fierce competition.

Days 3-10: look at CR (conversion from click to lead/purchase). If the CTR is good but the CR is low, the problem is in the landing page, not in the advertising. The advertising did its job.

After 10 days and 30+ conversions: look at CPL, subscriber cost, ROAS. We compare the groups with each other. We cut those where the CPL is 40%+ higher than the target. We scale the budget for working groups in steps of 20-30% per week - a sharp increase in the budget throws off the algorithm.

The numbers are the median for projects in the niches that I have. Yours may differ by 30-50% in any direction depending on LTV, competition and quality of landing.

8. Summary

VK Ads in 2026 is a working tool with a normal algorithm. Not the Facebook of 2018, but not the VKontakte Advertising of 2020 either. If you have an audience in Russia and a budget of 150K ₽/month per channel, you need to test it.

The main question is not “does VK work”, but “is your niche suitable”. Gaming, EdTech, e-com of the mass segment, medicine, local services - it works well. Narrow B2B with ICP like “IT Director of a large bank” - weaker there Yandex Direct search and Telegram Ads will be more accurate.

Related articles: CTR benchmarks in VK Ads, CPL calculator, comparison Direct vs VK vs TG, retargeting 2026, analysis of advertising.

If you want to analyze your specific situation, write to Telegram or through form. Starting consultation - 0 ₽.