GoPractice spoke with Andrey Novoselsky, the head of advertising at VK (the biggest social network in Russia) from 2014 to 2021.

Andrey told us:

  • How the advertising systems work and why it is important to maintain the balance between three parties: users, advertisers, and the ad platform.
  • How UAC-like campaigns will affect the interests of these parties;
  • What will be the future of advertising without IDFA, GAID, and cookies; and what marketers and advertisers should prepare for in the future.

For the convenience of our readers, we are presenting our conversation with Andrey in Q&A format.

→ Test your product management and data skills with this free Growth Skills Assessment Test.

Learn data-driven product management in Simulator by GoPractice.

Learn growth and realize the maximum potential of your product in Product Growth Simulator.

→ Learn to apply generative AI to create products and automate processes in Generative AI for Product Managers – Mini Simulator.

→ Learn AI/ML through practice by completing four projects around the most common AI problems in AI/ML Simulator for Product Managers.

Getting rid of IDFA, GAID and cookies. The imminent future of ad systems

The advertising system balances the interests of three parties: users, advertisers and its own business interests

GP: Tell us a bit about yourself. What did you do before joining VK? What were your responsibilities when you came to manage the VK advertising system?

Before VK, I launched two projects of my own, one in travel and the other in native social media advertising. Thanks to the latter, communication with VK began, and I eventually received a job offer from Andrey Rogozov (CEO of VK from 2017 to 2021 and previously the operating and managing director of the company). For the next seven years, I was responsible for the growth of advertising efficiency and advertising revenue at VK.

GP: What did you manage to build in VK and what were your achievements?

The main thing that has changed since 2014 is the transition from profile targeting of advertising to machine learning algorithms. Now, the ad rate in each advertising auction depends on which user we show it to, the time of day, the planned budget, and a dozen other parameters. Thanks to ML, all this happens automatically instead of being manually formulated. These changes allow you to rely less on manual campaign settings and automate large parts of the process.

Over the years, VK’s revenue has grown eight-fold, with the bulk of it coming from advertising. Almost all advertisers believed in and switched to ML algorithms. If two years ago about 10% of advertisers used automation and optimization, now it is more than 90%.

The main goal we have been aiming for over the past two years is the number of active advertisers in the small- and medium-sized business segment. I cannot give exact numbers, but we’ve observed a multiple-fold growth and we also see a huge room for improvement.

GP: What product principles does the advertising system use in its work? Whose interests are your priority?

To answer this question, you first need to draw a line between ad systems that work with their own audience and those that work with external ones. This is an important distinction: in the first case, the system delivers ads to the users it values the most (that is, its own audience); in the second one, the user experience is no longer so important to the advertising system (ads are shown on other people’s sites or apps).

The largest advertising systems such as Group (Note by GoPractice: VK is part of Group), Google, and Facebook usually combine these models. As expected, it is the internal part of ad network that brings in most of the revenue. Later, we will talk about advertising systems based on our own audience.

These products are based on the desire to balance the interests of advertisers, users, and the business interests of the advertising system. Advertisers should get good ROI; users should get interesting, relevant, and non-annoying ads; and the ad system owner expects to maximize margins and revenue.

Disrupting this balance will cause the advertising system to fall apart: it will either annoy users to such an extent that they leave the site, disappoint the advertiser with ineffective ads, or simply fail to make money.

In an ideal world, there would be a table listing all the “checks and balances” mechanisms that help maintain this balance. But in reality, this is always a semi-manual process, where, depending on the short-term and medium-term goals, the interest of one side or the other prevails.

At VK, we have chosen the interests of users as the main and undeniable priority: the long-term success of the advertising system depends on them. As long as there are users, there will be advertisers and revenue.

GP: You said that growth in the SME segment is one of the key goals. Can you use this segment as an example to illustrate how the balance of interests of the three parties works?

Let me give you an example. The advertiser wants to attract customers for $10 each. At the same time, we see that the cost can be reduced to $6. In this case, the advertising system will necessarily sell customers for $6, and not try to add up an additional $4.

In this case, the balance of interests of the parties will be as follows:

  • The advertising system will sacrifice instant profit, but at the same time will receive a loyal client who is ready to use the product for a long time and even increase their budget;
  • The advertiser will receive customers at a lower price than expected, which may become an incentive to continue working with the channel and allocating the general budget in its favor;
  • The ad will be seen by users who are most interested in it: this is what the ad system’s algorithms do. This minimizes the negative impact of advertising on user retention and engagement.

One of the important principles for us is not to maximize instant profits, but to optimize for long-term relationships and future revenue from advertisers.

An advertiser from a small or medium-sized business evaluates the result of its ad campaign only by sales. For the advertising system, this means that at every moment you need to bring concrete results to the advertiser instead of trying to spend the budget as quickly (and ineffectively) as possible.

Here, second-order effects also appear. The focus on long-term relationships leads to an increase in the number of active advertisers, an increase in the number of relevant advertisements, and an increase in competition for user attention.

UAC-like campaigns: Why ad systems are taking settings and control from advertisers

(Note by GoPractice: Several years ago, Google began to gradually migrate all types of mobile app ad campaigns to a single Universal App Campaign (UAC) type. In UAC campaigns, the user is only required to add creatives and set the goals of the advertising campaign. Then Google independently decides where, to whom, and how to display ads to achieve results. Thus, advertisers no longer have the ability to independently control targeting, placement, and other campaign parameters.)

GP: You said that you are actively moving and growing in the segment of small and medium-sized businesses. Did you have to redesign the product for such advertisers?

Yes, it is essential to do so. All advertisers from this segment are self-service (self-serve model), as opposed to large clients, where there is always a personal manager. At the same time, representatives of SMBs often don’t have sufficient experience in working with advertising channels, or are afraid and firmly rely on 200% ROI by tomorrow.

To tailor the product for their needs, all the inner components of the advertising system had to be rebuilt. The key goal of this overhaul was to replace a huge amount of customization features with a simple and straightforward interface, while automating a significant part of the work of setting up and running campaigns through ML and other techniques.

There are other ecosystem-level challenges associated with the growth in the total number of advertisers. For example, we had to figure out how to separate advertisers so that multinational companies don’t crush small businesses. In such a situation, quotas for business sectors are sometimes used, but in my opinion, the more correct way is to see how the advertising suits a particular user. To do this, the system may even decide to sell a particular impression for less if it is more relevant for the user. This is roughly what Facebook does with Relevance Scores. And again, everything rests on finding the optimal balance between the interests of advertisers, the ad network, and users.

GP: Your description gives the impression that you are heading towards UAC-like campaigns. What is it? What are the pros and cons?

In my opinion, the advertising system can and should help advertisers (and especially the not-too-experienced ones, such as SMBs) to avoid mistakes and spend their budget more efficiently.

It is important to understand that any advertiser has little information about the current state of the advertising system, so they can make mistakes and make unreasonable bids. Only the system itself can optimally choose the strategy for the advertiser and help ensure their money turns into the maximum number of actions.

In this context, UAC is a good choice, though the initial reaction of the advertiser is “that’s scary. What will wind up there in my ads? What are the sites? What will the creative look like?”

Some people try to experiment with UAC—it’s fun to watch them do so.

GP: Almost all major ad systems are moving towards a UAC-like model. Is it good or bad? What does an optimal development path look like?

I think everything depends on the implementation of UAC-like mechanisms in a particular product, as well as the tasks and level of expertise of the advertiser.

To be sure, you need to compare UAC with manual results, examine the recommendations of the ad system for the test period, and gradually scale its application. An exception, perhaps, is display campaigns, which use videos, images, and other media. In these ads, the creative component is the key element, and therefore, it is important to know how it looks in the site where it is placed.

Like autopilots in cars, UAC is being introduced gradually, and little by little, it starts to solve more and more cases without human assistance. For large advertisers, they might even defer the UAC input or add manual settings.

GP: So ad systems are gradually taking the job of managing ad campaigns from marketers? Something like how autopilot will eventually take jobs from car drivers?

Yes, that’s right.

So far, every case where a person manages advertising better than UAC means for the ad system that there is still work to be done. But automation will inevitably prevail. Marketers don’t have to worry though; automation isn’t going to be complete anytime soon. A marketer is a specialist with a broader understanding of how to attract a target customer and influence business growth.

A world without attribution. How to live without IDFA, GAID and cookies

GP: Advertising systems are gradually taking jobs from marketers, and device and browser manufacturers are taking away access to unique identifiers from both ad systems and marketers. How to live without IDFA, GAID and cookies?

It is important to understand that there is no going back, and there will never be. Google is already switching to cookieless browsers, iOS is removing IDFA, and Android has also announced plans to abandon GAID. This is a new reality, and the guys at Google and Apple needed courage to make those decisions.

Some believe that the actions of Apple and Google were aimed at extracting profits from the advertising money that now passes by the app stores (more precisely, they end up on Facebook – about $80B a year), and that user privacy protection is just a facade. We will find out the truth in a couple of years when we see the updated ad revenue metrics. In any case, this doesn’t negate the courage and audacity of the decisions made, which change one of the fundamental and infrastructural parts of modern Internet.

Advertisers can be intimidated, but it is better to go straight to the adaptation phase and learn to work in new conditions. Any revolution is a chance for new players to hit the market. (Note by GoPractice: You can read more about refusal from IDFA here.)

All this is not the end of the world. For example, here’s an unexpected AppsFlyer report that shows how the total number of installations is growing at the expense of unattributed users.

GP: Where should marketers look to adapt to this new reality?

Marketers should develop expertise in developing experiments for the new conditions: how to get the most effective advertising campaigns on the market without accurate conversions. At the very least, look at this Forbes post and think along the following lines:

  • Collect all your own first-party data – phone numbers, segmentation data (gender, age, geography, income level, etc.) and upload it to advertising sites;
  • Integrate with marketplaces, e-commerce, payment gateways or banks as soon as possible;
  • Experiment with large reach campaigns instead of highly targeted ad campaigns;

GP: How will advertising systems themselves feel in a situation where there is an audience, but there is no data on the effectiveness of the advertising campaigns shown to it?

The transition period won’t be easy. Then the big players will agree on the exchange of conversion data by hard identifiers (for example, by phone number). The possibility of such an exchange is an important advantage of advertising systems with their own platform and audience.

Moving away from traditional attribution models will be painful and almost blind. But I am optimistic because it will level the ground for everyone. It will be inconvenient to work with, but the inconvenience will apply to everyone.

GP: Would you dare to look even further? Who will benefit from these changes and who will lose?

It’s hard to say. My prediction is that the prices for attracting users will first grow, but will subsequently decrease, and eventually plateau above the current levels. The technical limitations introduced by Apple and Google are quite noticeable, so it is impossible to completely avoid the impact on business performance.

I think that advertisers will have a less accurate understanding of the effectiveness of the money spent on advertising. The approximate cost of users and sales for the channel will be clear, but in-depth analytics at the level of campaigns, banners, and segments will become virtually impossible.

The sites that own identifiers from the real world—first of all, phone numbers—will benefit the most. Attribution based on mobile numbers seems to be the most promising replacement at the moment.

Ad networks without their own platforms and audience will suffer the most. In fact, they will be left without data and tools to optimize for advertisers’ goals. Unless, of course, something unexpected happens. Narrowly targeted advertising campaigns will be tough.

For sites and apps that have so far hosted third-party ad systems that have audience data, this is, in a sense, a revenge. The power is likely to shift again to those who own the audience directly. Native ad formats for a narrow specialized audience of specific sites can get an impetus for development.

(Note by GoPractice: If the previous assumption is correct, then this will be a positive change for advertising agencies. Their added value has been decreasing with the development of the Internet: why contact an advertising agency if there are two or three advertising systems with which you can work directly. If we go back to the previous version of the world, where we need to launch campaigns on dozens of different resources, then the value of the agency’s work will increase again. Read more here.)

In general, for ads, opting out of IDFA, GAID and cookies is a bad change. Advertising will become less accurate to “hit” the user, and it will become less interesting. The balance of the three sides will be upset and everyone will suffer.

In doing so, this change will strengthen the position of the current largest ad systems in the market. It was already focused in the hands of a few players, but now new companies will have almost no chance to secure a spot in the market.

The largest ad systems won’t be affected in the long term. Global competition will force advertisers to keep growing their budgets. If you advertise a game on iOS, you need to advertise it there via Facebook and Google anyway – IDFA/GAID or not. This budget cannot be transferred to other channels.

One might get the impression that, due to the decrease in advertising efficiency, many companies will become more conservative (and this will lead to a decrease in business growth, a deterioration in the quality of advertising). But in reality, most of the money is spent by industry leaders, that is, those who value growth at any cost. Therefore, I believe that if in a certain sector at least one of the players continues to flood the audience with an advertising budget, then the rest will catch up, despite the decrease in efficiency.

In conclusion, I will say that advertising specialists need to expand their arsenal every day. Now the most important thing is to try working with SKAdNetwork—which replaces IDFA—as early as possible.


You have read the material prepared by GoPractice and Andrey Novoselsky, who managed the VK advertising system from 2014 to 2021.

→ Test your product management and data skills with this free Growth Skills Assessment Test.

Learn data-driven product management in Simulator by GoPractice.

Learn growth and realize the maximum potential of your product in Product Growth Simulator.

→ Learn to apply generative AI to create products and automate processes in Generative AI for Product Managers – Mini Simulator.

→ Learn AI/ML through practice by completing four projects around the most common AI problems in AI/ML Simulator for Product Managers.