How to test conversion-signal filtering
A practical audit plan for checking on-site tag controls, conversion definitions and independent outcomes.
Check both eligible tag events and mature external business outcomes. Lower pixel counts alone do not demonstrate improvement.
Before changing what reaches an advertising platform, write down what you intend to improve. Fewer pixel events is an observable technical outcome; better leads or sales requires separate evidence.
1. Map the conversion
Document the event name, the trigger, the page and the tag responsible. Check which tags load outside Button's supported managed loader. Identify what counts as a genuine customer action: a completed transaction, a validated form submission or a qualified CRM stage.
2. Establish the baseline
Collect tagged visits and inspect classification reasons, journey coverage and observed goal paths. Record the observation window and any configuration changes. Compare cohorts with similar dates and sufficient sample sizes.
3. Configure narrowly
Start in observation mode. If the evidence supports filtering, test a supported tag and one clearly defined page-path rule. Verify that the managed tag loads for eligible visits and is withheld when your rule specifies it. Check for duplicate loaders that might fire independently.
4. Evaluate consequences
Compare measured tag events and external outcomes after enough time has passed for leads or orders to mature. Look for false positives, collection gaps, seasonal changes and ordinary campaign edits. A before/after difference can be caused by more than the filtering rule.
5. Decide deliberately
Keep the control only if the result supports your measurement goals. Otherwise revise it or return to observation.
Research and expert guidance
Google's conversion-adjustment guidance separates corrected business outcomes from simple changes in event counts. Check what tags fired, what genuine leads completed and what was lost to false positives before calling a test successful. Source.
Next step: Should I turn on signal suppression immediately? · Related guide: Test mitigation without losing genuine customers