From 80% to 19%.
Ten months of MediaNation work on the CoolClever mobile app: from a measured 80.5% DRR to 18.7%. AppMetrica, postbacks and Unified Performance Campaigns — in that order.
Below is a walkthrough of the MediaNation case. Full version on the agency site: CoolClever case ↗.
The brand and the starting point
CoolClever is a natural-foods retail chain in Moscow and Nizhny Novgorod: 250+ stores, an online shop and an app used by more than a million shoppers. The assortment rests on in-house Myasnov production and direct beverage imports. MediaNation started on the app in autumn 2025.
On the surface everything looked fine: impressions, installs, spend. But there was not enough purchase data to optimise campaigns with intent. Over ten months we first measured true DRR — 80.5% — and brought it to 18.7%. Purchases rose while ad spend nearly halved.
A brief for developers and AppMetrica
The audit was clear: without in-app events and e-commerce in AppMetrica, Direct learns blind. We wrote a brief for cart, checkout-start and purchase postbacks, revenue and transaction IDs, and cross-platform attribution — and stayed with the client’s developers until the critical pieces worked. Then we let the new statistics accumulate.
Once the numbers were honest, the client set a KPI: DRR no higher than 35%. The optimisation plan used purchase conversion: cut high-CPI / low-conversion segments, moved campaigns to purchase strategies, focused on iOS (higher yield) and on users who added to cart but did not order. By December 2025 DRR was 61%, ROAS 164%.
UPCs as the scaling tool
In January–April 2026 we scaled through Unified Performance Campaigns: Search, the Yandex Advertising Network and smart banners under one strategy. Budget went where it paid back; weak segments were cut. SKAN postbacks added iOS signal. By April DRR was 31%, ROAS 322.6%, CPO 54% below start on lower spend. May hit 28.7% DRR. June spiked to ~55% while new YAN campaigns trained; by August we were back under the 35% target.
The takeaway
- Measure the economics before you twist bids. Without purchase postbacks, “optimisation” is guessing.
- UPCs amplify what is already measured. One strategy gathers more training data — only if the account sees user value.
- Scale without matching spend growth is possible. Here purchases rose while spend nearly halved.
Thanks to the CoolClever team for shipping the brief quickly — and to MediaNation for ten months of keeping DRR under 35%.