Creative testing

Creative testing, end to end

Screen larger batches on install signal when useful, validate on purchases, then scale. Smaller batches can start with purchase testing. Part 1 is live, the rest follow.

Stage 1Read

Elimination

Install campaign, new
  • Close versions of the same concept compete here
  • Signal arrives within days: CPI and early behavior
  • Goal: nominate variants for purchase validation
the ads that pull away advance, usually 1 to 2
Stage 2Coming

Validation

Purchase campaign
  • Screened candidates or a small batch of new ads enter
  • Set a budget, evaluation window and purchase cost target
  • Judge evidence against the decision, not a universal event count
sufficient purchase evidence + cost below your target
Stage 3Coming

Scaling

Main campaign
  • Ads below the cost line move to the main campaign
  • Set the cost line separately for your campaign

What this guide is for

This is the method I use to test ad creative for apps and games on Meta, written out in three parts. Each stage answers one question and passes only a few ads on, so most of the budget goes to ads that have already earned it.

How the three stages fit together

Elimination is optional. Close versions of one concept compete in an install campaign, and CPI and early behavior show up within days. It is for teams with more variants than they can afford to test on purchases. In my accounts a version needs about 10 purchases to show potential and about 50 for a verdict, which adds up fast across many variants. Usually 1 or 2 versions pull away and move on.

Validation runs those candidates, or a small batch of new ads, in a purchase campaign with its own budget, evaluation window and purchase cost target.

Scaling moves the ads that land below your cost line into the main campaign.

What a round costs

The worked example in part 1, 17 creatives in 4 ad sets at a $1 CPI, comes to about $240 for a three day round. A first purchase test costs more: $2,000 to $3,000 for first results, as the first paid test article sets out.

When to skip elimination

A small batch can go straight to purchase testing. Elimination earns its cost when there are more versions than purchases to judge them by.

Sources and operating assumptions are distinguished in the guide.