Yandex Ads automation means letting Yandex Direct’s algorithms choose bids, placements, audiences and ad combinations toward a goal you define in Yandex Metrica. The parts worth handing over are bidding, placement selection and creative combinations. The parts to keep are goal definition, budget limits, negative keywords, offer and landing page quality, and the decision about what a conversion is worth to your business.
How far has automation gone in Yandex Direct?
Yandex Direct has moved steadily toward automated campaign types. The Unified Performance Campaign (in Russian, Единая перфоманс-кампания, often shortened to ЕПК) combines search, the Yandex Advertising Network, smart banners and image ads in one campaign. You supply a goal, a budget, keywords or autotargeting, and a set of headlines, texts and images; the system builds and serves combinations.
Older campaign types such as text and image campaigns still exist with more manual control, but new features generally arrive first in the unified format. Understanding what it does automatically is now part of running Yandex Ads at all.
What should you hand to automation?
Some tasks are done better by the algorithm because it sees far more signals per auction than any person can.
- Bidding: autostrategies such as maximise conversions, pay per conversion and target cost per action adjust bids per auction using device, time, location and user behaviour.
- Placement choice: deciding when to show on search versus the advertising network.
- Ad combinations: assembling headlines, texts, images and video into variants and learning which work.
- Autotargeting: matching ads to relevant queries beyond your keyword list, within categories you choose (target, narrow, broad, alternative and competitor queries).
Choosing an autostrategy
Pay per conversion (оплата за конверсии) charges only when a Metrica goal is completed, which shifts risk to Yandex and is useful when conversion data is limited. Maximise conversions with a target cost suits accounts with steady volume. Strategies optimising toward revenue or ROI need e-commerce or value data passed to Metrica.
Whichever strategy you choose, autostrategies need a learning period. Yandex generally recommends a minimum number of weekly conversions for stable learning; check the current threshold in the strategy settings for your goal type.
What should you keep under manual control?
Automation optimises toward whatever you tell it. If the inputs are wrong, it efficiently buys the wrong outcome.
- Goal definition: decide which Metrica goal counts. Optimising for ‘visited contact page’ instead of ‘submitted form’ produces cheap, useless results.
- Conversion value: tell Yandex what each conversion is worth, especially when different products have different margins.
- Budget limits: set weekly budgets you can afford, and do not raise them sharply mid-learning.
- Negative keywords and exclusions: autotargeting needs boundaries; review search queries and add negatives weekly.
- Site exclusions: block advertising network sites that send junk traffic.
- Creative quality and claims: algorithms combine what you provide, so every headline must be accurate and compliant with advertising law and labelling rules.
- Landing pages: no strategy fixes a slow page or a weak offer.
Setting up Metrica so automation has good data
Yandex Metrica is the data source for almost every automated decision in Direct. A few setup steps make a large difference.
- Install the Metrica tag on all pages and link the counter to the Direct campaign.
- Create goals for real business events: form submissions, calls via call tracking, purchases with revenue through the e-commerce data layer.
- Mark key goals as primary in the campaign so the strategy optimises toward them.
- Upload offline conversions from your CRM, so Direct learns which leads became customers, not just which ones filled in a form.
- Exclude internal traffic and test submissions.
For example, a B2B equipment supplier might find that half its form fills come from students and job seekers. By sending CRM-qualified leads back into Metrica as a separate goal and optimising on that, the campaign spends on fewer, better leads.
How do you supervise an automated campaign?
Automation reduces daily button-pushing, but it does not remove the need for review. A weekly check should cover:
- Search query report: add irrelevant queries as negatives.
- Placement report for the advertising network: exclude poor sites.
- Asset report: replace low-rated headlines and images.
- Cost per goal trend against your target, by device and region.
- Budget pacing: is the strategy spending, under-spending or hitting limits early?
Make one significant change at a time and wait for the strategy to adjust. Changing goals, strategies and budgets in the same week makes it impossible to see what worked and resets learning.
When to step back from automation
If conversion volume is very low, for example a niche service with a few leads a month, automated strategies have little to learn from. In that case optimise toward an earlier but still meaningful goal, consolidate campaigns so data is pooled, or use manual bidding on a tight keyword set until volume grows.
Common automation mistakes in Yandex Direct
The same handful of errors appear in many automated accounts, and each one teaches the algorithm the wrong lesson.
- Optimising toward several goals with equal weight when only one of them is a sale or qualified lead.
- Splitting a small budget across many unified campaigns, so none reaches enough conversions to learn.
- Uploading only two or three headlines, which leaves the system little to combine and test.
- Turning on every autotargeting category, including competitor and alternative queries, without reviewing what they bring.
- Restarting campaigns after a bad week instead of waiting for learning to settle.
- Forgetting regional and schedule settings, so ads show in regions you cannot serve or at hours nobody answers the phone.
Document the settings of each campaign at launch in a short sheet: goal, strategy, target cost, budget, regions and autotargeting categories. When results change, compare against that sheet first. Many apparent algorithm problems turn out to be a setting someone changed without telling the team.
Automation also changes what good account management looks like. Less time goes into bid adjustments and more into feeding the system better inputs: clearer offers, more creative variety, cleaner goals and fresher negative keyword lists.
Before scaling an automated campaign, check that the sales team can handle more leads and that stock or capacity can absorb more orders. Automation finds volume quickly once it works, and a sudden increase that the business cannot serve damages reviews and repeat custom.
Where to start this week
Open Yandex Metrica and list the goals your campaigns optimise toward. Remove or demote any goal that is not a real business outcome, and set up offline conversion upload if leads close outside the site. Then review the last 30 days of search queries in your largest campaign and add negatives. With clean goals and boundaries in place, Yandex Ads automation works for your business rather than for the cheapest available click.
Keep reading: Yandex Ads · Advertising