How creative testing works
A creative test runs two or more ads that differ in one meaningful way, such as the hook, the format or the offer, under similar budgets and audiences, and compares the results. The goal is to learn what makes the right users install and pay, then produce more of it.
A typical structure:
- Pick a hypothesis. "A gameplay-first hook beats a character intro."
- Change one thing. Keep audience, placement and bid strategy fixed.
- Set the decider. For example, cost per purchase over 7 days.
- Give it volume. Enough spend for each variant to reach a meaningful number of conversions.
- Scale the winner and log the learning.
Judging the result
Click metrics move first, but value metrics decide. A worked example with two creatives at $3,000 spend each:
- Creative A: 1.8% CTR, 2,000 installs ($1.50 CPI), 60 purchases, $1,800 D7 revenue (0.6x)
- Creative B: 1.1% CTR, 1,500 installs ($2.00 CPI), 90 purchases, $2,700 D7 revenue (0.9x)
Creative A wins on CTR and CPI. Creative B wins on cost per purchase ($33 against $50) and ROAS. B is the one to scale. A hook that attracts curious clickers often attracts users who do not buy.
Common mistakes
- Calling it on CTR. It is the easiest number to move and the weakest link to revenue.
- Too little data. Ten purchases per variant rarely settle anything.
- Changing several things at once. You learn that one ad won, not why.
- Ignoring fatigue. A past winner can decay; retest against fresh challengers.
- Testing without attribution down to creative. Network-reported conversions per ad overlap with other networks and differ in method.
Creative testing in Linkrunner
Linkrunner attributes installs, events and revenue down to the creative, with spend synced from Meta, Google and Apple Ads, so each creative has its own CPI, cost per event and ROAS. The Creatives tab shows Meta creative metrics next to those results. Automated creative optimization is coming soon and is in internal testing. See ROAS measurement.