Stack Broad on Broad

Year: 3, Month: 9, Week: 1, Day: 2

In the middle of winter 2023, a client from our recurring management projects gave us a boss-level task: they wanted 3,000 subscribers per month instead of 1,000, while keeping the same price we had been hitting for four months straight: RUB 70 per subscriber on average.

We wanted to argue. We accepted the challenge anyway. It was a real growth point, and it turned into a strong case. In that moment, though… we were mildly panicking.

We were working in the legacy VK ads cabinet (older VK ads interface), and after some suffering and thinking we found a set of solutions inspired by a Russian meme: “Stack broad on broad.”

The old cabinet offers several ways to work with lookalike audiences (algorithmic similar-users segments):

1. Collect 1,000 reactions and build a lookalike from 25,000 to 2,000,000 people.
2. Use the client’s community as the base, exclude that community, and move the similarity slider from 10% to 100%.
3. Use keyword queries of different sizes and apply different “query freshness” windows — from 1–3 days up to 30 days max.

Here’s how we played with that LEGO:

1. We built the broadest click-based lookalike and intersected it with broad keywords with up to 30 days of freshness.
2. We built the broadest percent-based lookalike from the group and intersected it with keywords.
3. Same approach, then intersected it with a 2,000,000 click audience.

And a few more variations in the same spirit.

That’s how a 2,000,000 audience could shrink to ~30,000.

This is how we stayed within the target price the client wanted — all the way until early summer.

Result you can apply: When you need scale without losing CPA, go wide first, then narrow via intersections with intent (keywords, recency, behaviour). Broad seed plus smart filters keeps volume and protects cost.