Ad data is not forever: collect and store your own statistics

An ad account report is a narrow window, and it does not keep the full picture for you. Part of the truth sits in call recordings. Part arrives through channels attribution never sees. Part drowns while you scroll through ad groups one at a time. Gathering all of it in one place and keeping it on your side is separate work, and without it you make decisions on the scraps the account happened to show.

For a company selling into Russia from another country, this is sharper than it was at home. The platform is Yandex Direct rather than Google Ads, the sales conversation happens in a language your head office does not speak, and the account itself was often opened by a local agency. Everything below is about the same three leaks I see in Russian accounts every month, and about what changes when the company running them sits abroad.

Scrolling through groups one at a time hides the truth

In a large campaign, opening each ad group in turn and stepping back out eats time, and comparing groups by eye is close to impossible. By the time you reach the last one, you have forgotten the first.

The fix is a set of linked reports in a BI tool at the level of the whole account, or of all accounts, that you switch between in one click. Google advertisers do this at manager-account level. Yandex Direct has no manager account in that sense: it has agency accounts and shared access, and its built-in Report Wizard shows one account at a time. So the cross-account view is built outside the interface, by pulling data through the Reports API into whatever BI you already run.

What goes into it is simple. Broken down by campaign and by the largest ad groups, histograms of click price against clicks, click-through rate against clicks, cost per lead against leads, and conversion rate against leads. A separate report with totals by date keeps the daily volume of traffic and inquiries under control. If one or two groups dwarf the rest, build a copy of the histogram on a logarithmic scale next to the ordinary one: the distribution across the small groups stops hugging zero and becomes readable again.

This dashboard has an honest limit. For it to answer how ad quality affects the metrics, three conditions have to hold at once: each group has accumulated a reliable sample, each ad inside the group has enough clicks and enough leads, and the ads in the group genuinely differ in quality. Until then the dashboard is a navigator through the data, and it is not yet a source of conclusions. That is the first reason to pull statistics to your side: you export them into your own store instead of relying on what the account interface shows and on how long it chooses to keep it.

Foreign companies that advertised in Russia through Google before 2022 know how this ends. The campaigns stopped, and whatever had not been exported stayed inside an account that nobody opens for this market any more. If your Yandex Direct account belongs to an agency and you part ways, the history is theirs to hand over or not.

Call recordings: the data that lives in no account at all

Without access to call recordings and feedback from the sales team, a marketer works blind. The ad account does not show what happens after the click, and that is precisely where the fate of an inquiry is decided.

In Russia this gap is wider than in most Western markets, for a plain reason: a larger share of first contact still happens by phone and in messengers. Call tracking with recording is standard local practice here. If your Russian sales are run by a distributor or a partner rather than your own team, the request for recordings and feedback has to be written into the agreement up front, because asking after the first bad month rarely works.

From the recordings and from conversations with sales you pull out what appears in no report:

• what a conversation with a buyer actually starts with, and what worries them;
• which objections come up and how the manager walks the person down the funnel;
• what gets upsold, what happens after the service, how customers are retained and how word of mouth is fed;
• how good the incoming leads really are, and whether the call scripts make sense.

The base condition for meaningful optimisation is call tracking with recording, access to those recordings handed to the specialist, and feedback from sales on lead quality agreed in advance. The structure and navigation of the site are then designed to repeat the logic of the conversation that ended in a purchase: the same order of closing doubts, the same sequence of arguments. This is where “data is not forever” becomes literal. A conversation that was not recorded is gone for good, and with it goes the voice of the customer you could have used to fix the ads and the site.

One practical note for companies working from abroad. The recordings will be in Russian, so the marketer listening to them has to read the language, and the personal data inside them falls under Russian personal data law, which requires data on Russian citizens to be stored on servers inside Russia. Where the archive lives is a question for counsel at the start, before the first call is recorded.

There is a second effect of this connection, less obvious than the first. A steady exchange between client and specialist produces insight, and it also produces priorities: it becomes clear which segments should receive budget and what to fix first. Site strategy and advertising strategy align, trust grows, and the client hands a growing share of the work to one contractor with less anxiety. Both sides gain: the client stops hunting for a new contractor for every micro-task, and the contractor stops chasing new clients. A communication break leaves both sides without this data, and the work goes blind.

The best channel is often invisible in attribution

Sometimes the cheapest and most convertible traffic is one the account does not connect to the budget that produced it at all.

In one online store, the main share of conversions over a period came from brand traffic: visits from searches containing the store’s name. There was no direct advertising on brand queries. The audience learned about the brand from videos by bloggers and then searched for the store by name on its own.

That traffic turned out cheaper and more convertible than the performance channels. A person deliberately searching for a store by name reaches payment far more willingly than a cold click. The funnel had a quirk too: an order was logged as placed before payment, and the manager sent the payment link separately, yet the share of placed orders that reached actual payment was very high. That strengthened the value of brand traffic further. The seeding with bloggers was financed from a separate budget, kept apart from search advertising, and over time the emphasis shifted to one main video channel as the more stable performer. In attribution reports, this brand traffic was formally tied to no advertising budget, although the seeding was exactly what created it.

For the Russian market this pattern is common, and it hides in a specific place. Seeding here runs through Telegram channels, VK Video, YouTube and Dzen, while the brand search it produces lands in Yandex Metrica as organic Yandex traffic. A head office reading the report sees a healthy organic line and a seeding budget with no measurable return, and cuts the seeding. The organic line then thins out a quarter later with no report explaining why.

The conclusion for a practitioner: when judging performance channels, do not lose the contribution of indirect sources. Seeding, awareness work and influencers feed direct visits and brand search, and those may turn out to be your best channel, the one that standard attribution says does not exist.

Collect it and keep it on your side

One thread runs through all three stories. Account data is incomplete and short-lived: some of it the platform does not show, some it hides inside aggregates, some is lost if it is not captured in time. So the full picture is something you assemble and store yourselves. Your own statistics store in BI, your own archive of call recordings, your own manual stitching of indirect channels to the budgets that created them. The platform does not keep this archive for you, and counting on the needed detail waiting for you in the interface is a bad bet.

For a foreign company this comes down to ownership questions. Who owns the Yandex Direct account, you or the agency. Where the Reports API export lands and how often it runs. Whether Metrica raw visits are being pulled through the Logs API or only viewed as aggregates in the interface. Who holds the call recordings and who is allowed to listen. Where the seeding spend is logged next to brand query volume. Each of these has a default answer that favours whoever set things up, and rarely is that you.

What to do about your analytics in the coming weeks

• Check whether the account’s statistics are exported into your own store, or whether the whole history lives only inside Yandex Direct.
• Build a dashboard with histograms by ad group so you see everything at once instead of scrolling one by one, and keep linear and logarithmic scales side by side.
• Confirm that call tracking with recording is in place, and that the specialist has access to the conversations themselves rather than to a call summary.
• Agree with the sales team, or the distributor, on lead quality feedback in advance, before launch rather than after complaints.
• Stitch indirect channels to the budgets that create them, so that brand traffic is not credited to itself out of nowhere.
• Settle who owns the account and where the archive lives, in writing, while the relationship is still good.

Pretty reports like to show what is easy to count. The most valuable things about your advertising often lie where the account does not look and does not store. Whoever collected and kept them is the one who sees the whole picture.

If you suspect you are deciding on a cut-down version of your data and half the truth is leaking away somewhere, come for a free review: I will show what is not being collected and where the picture drains out. And for how analytics fits into a working paid traffic system in Russia, see the Yandex Ads page.

Frequently asked questions

Sources

• Russian version of this article on belokrylovo.ru: Данные не вечны: свою статистику собираем и храним сами