Campaign structure in Yandex Direct: five decisions that cost or save you later

Account structure is usually built for tidiness: so it looks pleasant and a colleague can follow it. Tidiness is secondary.

Structure decides two practical things. Where you will find the data you need later, and what you will be able to control separately from everything else. Beauty is cosmetic. Below are five structural decisions, and each of them either saves hours in reporting or quietly costs you inquiries.

Campaign attributes belong in the name, not a label

Labels look convenient: hang one tag on a dozen campaigns, switch on a filter in a report, and the breakdown is ready.

The problem surfaces later. Filtering by label is unavailable in some standard reports and entirely absent from custom ones. A filter on the campaign name works everywhere: in any section of the interface and in any report you build yourself, no exceptions.

Hence the rule. Shared attributes that apply to several campaigns at once, region, topic, language, traffic type, go straight into the campaign name. Then in a report of any depth you filter them with an ordinary text filter. To avoid rebuilding conditions every time, save them as a preset.

One caveat about the desktop editor. Saved web filters do not travel there, while labels synchronise automatically. If you live in the campaign editor, labels are useful, but as a supplement, not as the basis of segmentation. The basis is the name.

For a foreign advertiser there is a bonus reason to name carefully. Whoever reads your reports in six months may not read Russian. A campaign name that carries region, language and traffic type in Latin characters keeps the account legible to your side of the business.

Split different domains into campaigns, not ad groups

Technically one account, and even one campaign, can carry several domains. Inside an ad group only one domain is allowed, so the minimum level of separation is different groups.

The temptation is obvious: put everything in one campaign, divide by groups, done. Resist it.

At group level half the levers are unavailable to you. The bidding strategy is set per campaign. So is ad scheduling: different display schedules for directions whose sales teams work different hours cannot be set on groups. Separate goals and target conversion prices, regional adjustments, shared budget, all of that lives at campaign level and above. Add the eternal fiddling with domain filters in every report and in analytics.

Different directions mean different campaigns. That is how you keep control and separate accounting.

Short keywords: split the campaign or label the groups

One-word and two-word keywords are dangerous because of their width: a lot of other people’s intentions fall under them.

The classic cure is moving them into a separate campaign to control relevance. The decision is clean, but it has a price. Automated bidding strategies learn from conversion statistics, and the more you fragment those statistics across campaigns, the slower and worse the learning goes.

The compromise looks like this. Short keywords are not moved into a separate campaign but separated into their own ad groups inside the same one. The group name is marked with a clear attribute, say “one word” and “two words”, and the keywords themselves get the same label so the breakdown is one click away.

To stop long keywords of three or more words from catching short queries, each group of long keywords gets the target short phrases added as negatives in exact match. You cannot do that with a shared list: negative lists attach to the campaign, not the group, so it is done by hand, group by group.

Which to choose depends on bidding. On automated strategies a single pool of statistics matters more than purity, and labelling by groups wins. On manual bidding the benefit of a shared pool is smaller, and splitting into campaigns regains its sense. That is the real fork around the single-keyword-ad-group idea: precise structure per short keyword, or statistics preserved for the sake of learning.

Master campaigns living next to the rest

Before any fine structural tuning, check one thing.

If a Master campaign, Yandex’s simplified automated format, runs alongside ordinary campaigns in the same account, it tends to take almost all the available traffic for itself. Its buying model is more aggressive, and the other campaigns start receiving fewer impressions, sometimes to the point of effectively stopping.

The conclusion is simple: do not put them head to head over the same audience. Either separate goals and budgets so there is no direct competition, or deliberately limit the Master campaign’s participation if the priority is your manually managed ones.

Noticing the effect after the fact is hard. You will be guessing why a search campaign suddenly deflated, while the cause sits right next to it in the same account. The platform does not document the tug-of-war mechanism anywhere, so check it on your own data rather than taking it on faith: set up observation, look at how traffic divides between formats, and decide by the numbers.

Naming ad groups: word order should follow demand, not the alphabet

When a keyword set of tens of thousands of phrases is clustered automatically, each group gets a name made of its significant words. The temptation is to sort those words alphabetically. That loses the meaning: the name no longer shows which word is the main one and which is dependent.

A different order works. Category words in the name are arranged by descending industry demand for the cluster they belong to: first the main object of advertising, then the subcategory with the largest reach, then qualifiers and additions, also descending.

The benefit is threefold. Implicit duplicates with different word order collapse into one spelling. The list of names immediately shows which characteristics are popular and which are secondary. And you get a ready queue of priorities: which categories to launch first, which landing pages to improve first, in what order to put menu items on the site. Structure stops being guesswork and becomes a cast of real demand.

What to do with your structure

1. Move cross-cutting attributes into campaign names and save a filter preset. Leave labels for the desktop editor.
2. Check whether different domains are separated by group instead of by campaign. If they are, split them: you get back bidding, budget and scheduling.
3. Choose deliberately for short keywords: group labelling on automated strategies, separate campaigns on manual bidding.
4. Make sure a Master campaign is not eating its neighbours’ traffic. Separate the goals or limit it.
5. Build group names by descending cluster demand, not alphabetically, and read the priority map off them.

Structure rarely makes it into a client presentation: you cannot show it as an achievement. But it decides whether you will see where the budget leaks and whether you can reach the right lever when it matters. A bad structure does not break loudly. It quietly hides data from you and ties your hands.

If the account feels impossible to navigate and you cannot tell what to grab first, that is one of the things the free review untangles.

Frequently asked questions

Sources

• Yandex Direct help, campaign types and settings: yandex.com/support/direct/en/campaigns
• Yandex Direct help, statistics and report filters: yandex.com/support/direct/en/statistics
• Russian version of this article: Структура кампаний: SKAG, короткие ключи, спрос