Smart TV in Performance Marketing: How to Cut Cost Per Opened Deposit by 32%Article · digest for November 18
Smart TV in Performance Marketing: How to Cut Cost Per Opened Deposit by 32%
Article4 sections≈ 7 min

Smart TV in Performance Marketing: How to Cut Cost Per Opened Deposit by 32%

Seventy-eight percent of viewers take action after seeing an ad on a smart TV: they install apps and go on to make purchases. Those figures come from researchers at Visible and Data Insight, while Smart TVs are already in 68% of urban households in Russia. Even so, the big screen in the living room was treated for years as an expensive brand-awareness billboard where teams ran pretty videos with no clear financial return. A hard ceiling in classic channels changed everything.

In March 2026, the financial marketplace Finuslugi, the agency E-Promo, and the programmatic platform Hybe tested the CTV-to-Mobile format in a live campaign, which they detailed on Sostav.ru. The teams linked TV video views to actual bank deposit openings inside a mobile app. By the end of the campaign, cost per sale dropped 32% below the target KPI, though getting there meant surviving a launch with more than double the planned cost. Let's look under the hood of this setup and see what made it work.

Why Traditional Channels Hit a Ceiling

To understand why a financial marketplace needed television in the first place, look at E-Promo's study "Media Inflation in Digital." During 2025, the cost per target action in mobile marketing rose 8–34% depending on the niche. In fintech, the squeeze was even tighter: in-app ads and Yandex buys hit a physical audience ceiling.

The moment a business tried to scale and raise spend by just 20–30%, customer acquisition cost jumped 1.5 to 2 times above plan. Ad auctions had become so overheated that every new depositor cost far too much. Against this backdrop, internet-connected TVs looked like an almost empty field. According to Hybe business development manager Marat Agaev, Smart TV penetration across the country already reaches 78% of households, yet auction competition there remains calm with no price panic.

E-Promo Group head of mobile marketing Evgeny Lyubimov sums up the goal of the test simply: reach fresh audiences in a brand-safe environment and hit a target cost per sale that can be scaled smoothly later.

How the TV Screen Connects to the Smartphone

Let's see how a tracker connects two different screens in the same apartment. Until May 2026, the CTV-to-Mobile format mostly ran on foreign trackers like AppsFlyer and Adjust, until the local tracker AppMetrica launched cross-platform attribution. Finuslugi, E-Promo, and Hybe were among the first to test it on live traffic.

The setup relies on two bridges between the couch and the phone. On web platforms and TV apps, a short 10–15 second video ad plays on the television screen. At the end, a closing frame pops up with a dynamic QR code that leads straight to the app store. The most impatient viewer points their phone camera at it, and the click is tracked directly.

However, few people jump off the couch for their phone while a movie is playing. More often, a viewer remembers the service name and installs the app an hour later or the next morning. That's where probabilistic IP-matching comes in, as described in AppMetrica's documentation. When the ad plays on TV, an Impression URL fires—a special view tag that records the home router's external IP address.

If a Finuslugi app install happens on a smartphone over the same home Wi-Fi network within 24 hours of that ad view, the system links the TV impression to the mobile conversion. From there, the tracker sees the entire in-app funnel as usual, all the way to registration and funding a deposit.

A Two-Week Learning Dip and Hybe's Algorithm Ramp-Up

Next came the most nerve-wracking part of the project. E-Promo and Finuslugi produced the videos and set up analytics and integration, while Hybe handled inventory buying, ad labeling, and algorithm tuning. They immediately pushed traffic through SSP ad networks: Between Digital, Bidvol, Solta, Buzzoola, UMG, Videohead, Yandex ORTB, and other platforms.

In the first phase, the algorithms simply learned and gathered audience data. On the budget sheet, this scouting phase looked harsh: cost per opened deposit (CPS) was 2.37 times higher than the target KPI. At a start like that, it's tempting to pull the plug, so the teams checked how users behaved after installing the app. Install-to-registration conversion stood at 41.25%, and deposit-opening conversion held at 4.43%. That proved TV was bringing relevant, solvent users; the system was just buying too many wasted impressions.

In the second phase, Hybe's ML algorithms kicked in. They sifted incoming traffic in real time and adjusted bids based on install cost, gradually cutting out dead placements. At the same time, budget shifted toward sources where people completed target actions best: ViNTERA.TV, START, PlutoTV, and ProgTV Android rose to the top. To lock in volume without losing quality, the teams opened direct deals with Zona.tube TV and Lime HD TV apps, as well as Videohead and UMG SSPs, while testing both web and in-app TV placements.

Returns showed up in just a couple of weeks. Cost per sale dropped 30% compared to the previous week. Registration-to-deposit conversion grew 2.6 times over the learning phase, and total sales volume rose 2.14 times. By the finish, overall install-to-registration conversion settled at 40.85%, slightly below the initial 41.25% because algorithms shifted away from casual sign-ups toward future depositors. Meanwhile, the share of registrations that turned into actual opened deposits climbed to 7.64%.

The main financial result looks like this: in the final week of the campaign, cost per sale came in 32% below Finuslugi's target benchmark. Measured across the full month, the cost per acquired deposit matched Yandex Advertising Network (RSYa) and beat VK Ads and several OEM sources like smartphone lock-screen ads, all in a fully controlled licensed environment. Finuslugi marketing director Yulia Mayorova noted after the test that Smart TV moved from a one-off experiment into the marketplace's permanent media mix.

Cost vs. Return: Where This Channel Works and Where It Burns Cash

The report numbers look appetizing, so let's take off the rose-colored glasses and weigh the real price of this entry. First, you pay upfront with time and an initial loss. Running a two- or three-day Smart TV test makes no sense: during learning, cost per deal can fly more than twice above target, and your budget must comfortably survive those first weeks before ML models find working placements. For low-ticket products with tiny budgets, that ramp-up is usually unaffordable.

Second, probabilistic screen matching relies on one technical condition: the TV and the phone must sit behind the same home router. If someone watches an evening movie on Wi-Fi while their smartphone uses mobile data, the external IP addresses won't match. In that case, only a direct scan of the dynamic QR code on screen saves the attribution link.

Third, financial services have a long consideration window. To open a deposit, a user has to pass identity verification and transfer funds between accounts. AppMetrica's standard 24-hour attribution window catches quick installs, yet evaluating delayed demand over a 14–30 day window gives a truer picture of long-term return. Finally, Evgeny Lyubimov warns specifically about inventory quality. The market has plenty of junk placements where it's easy to fake pretty impression numbers with zero return, so without direct deals with trusted video services and strict SSP cleanup based on in-app actions, growth on paper quickly turns into fiction.

Want a deep diveVotes help us choose what to break down in detail. This material has 0 votes.
Sources · 2