“The ads are performing poorly” is too broad to diagnose. Define the target CPA, revenue, measurement window and conversion event, then find the funnel step that loses users. This guide adapts the framework in Xiaowei's 2022 article; its old interface examples are not current platform instructions.
Confirm three inputs before launch
- Business target: markets, CPA/ROAS target, conversion quality and test budget.
- Product experience: complete the install, signup and purchase flow yourself.
- Measurement: events, MMP and platform postbacks, naming, attribution window and owner.
A low CPA can still attract poor-quality users. See our event quality analysis.
Diagnose each funnel step
| Step | Metrics | First checks |
|---|---|---|
| Impression to click | CTR, CPC, frequency | Creative message, audience and placement |
| Click to store or landing page | Arrival rate, load time | Links, redirects and promise consistency |
| Page to install and first open | Install rate, CPI | Store page, compatibility and first-open loss |
| First open to purchase | Purchase CVR, CPA, ROAS | Product flow, tracking and user quality |
High CTR with low purchases can reflect a mismatch between the ad promise and the product, not simply weak creative.
Run a test that can answer one question
Write one hypothesis, such as whether a product demo improves store-to-install rate versus an abstract claim. Hold the optimization event, attribution window and main market steady, then change a creative concept or page. Record impressions, clicks, installs, purchases and downstream value. Reduce parallel tests when event volume is sparse. The source's three-to-seven-day period is an example, not a rule; use your own conversion volume and reporting delay.
A practical troubleshooting order
- Validate tracking, filters and date definitions.
- Locate the funnel step where performance changed.
- Review recent budget, bid, audience, creative, store-page and product changes.
- Test the most plausible cause at limited scale and review on the same basis.
The source's “Restricted” Meta attribution example came from a 2022 account. Use AppsFlyer's current Meta integration guide for today's fields and settings. See also our Meta app ads guide. Adapted from Xiaowei's original article.
FAQ
What should I fix first when CTR is high but purchases are low?
Locate the loss across arrival, install, first open and purchase, then check whether the ad promise matches the actual product experience.
How many days should an ad test run?
There is no universal duration. It depends on conversion volume, the business cycle and attribution delay; set decision criteria before launch.
Why can CPA improve while ROAS worsens?
The event may bring cheaper but lower-value users, or the revenue window may be immature. Compare costs and revenue for the same cohort.