In six weeks, a B2C app found the cheapest channel, operating system and creative for downloads

In a six-week test to validate a B2C subscription app's entry into a new market, we crossed three variables: channel (Google and Meta), operating system (Android and iOS) and creative line. Google's cost per download was 25% below the account average and more than 4 times lower than Meta's, Android was in one of the weeks more than 15 times cheaper than iOS, and the creative that showed the app itself beat the other lines. The test also showed the limit of cheap downloads: fewer than 1 in 10 downloads became a sign-up.

Profile

B2C app · Seed

Problem

The app was entering a new market, with no media history in the country, a test budget and little time to find out where to invest. And the conversion that mattered most, the start of the free trial, wasn't yet measured in Analytics.

Solution

Test channel, operating system and creative at the same time, read the results every week and quickly cut what didn't perform, without losing sight of the funnel after the download.

What we did

We ran app campaigns on Google and Meta, split by operating system, optimized for the first app open while the free trial start wasn't measured.
We tested five creative lines (product, protagonist, human, typographic and benefits) and paused iOS and Meta when the numbers showed a cost per download far above Android and Google.
We tracked the funnel after the download, from sign-up to free trial, and caught it when sign-ups stopped being measured after an app update.

Result

Google with a cost per download 25% below the account average and more than 4 times lower than Meta; Android more than 15 times cheaper than iOS in one of the weeks; the product creative line with the lowest cost per download on both channels and more than half of Google Android installs. From download to sign-up, conversion stayed below 10%.

Key learnings

Channel, operating system and creative are three different levers. Tested together, they show within weeks where the cheap download is.
The creative that shows the product beat the aspirational ones. Whoever sees the app quickly understands what they'll get when they click.
A cheap download is not a customer. Without measuring the next step, sign-up and free trial, the campaign optimizes for the wrong place.

Do it yourself

Split app campaigns by channel and by operating system from the start, with equal budget per combination, and compare each combination's cost per download every week. Cut the most expensive one from the second or third week.
Test three to five creative lines with different messages, including one that shows the app's interface, and read cost per download by creative line, not by ad.
Before scaling, set up in Analytics the event that's worth money (sign-up, free trial start or purchase) and switch optimization to it as soon as there's volume. Ask an LLM:
I'll paste the weekly results of app campaigns by channel, operating system and creative line, with spend, downloads, sign-ups and trials started. Calculate the cost per download and per sign-up for each combination, tell me which to cut and where the biggest funnel loss is, and suggest 3 actions to increase download-to-sign-up conversion.