Understand optimization before using it

Start with reliable signals

Optimization cannot repair an incorrect conversion event or a misleading landing page. Check that real conversion requests use the configured event, a unique event ID and the saved click identifier. Keep receiver tests separate from live outcomes.

Pacing and creative allocation

Even pacing spreads a daily allowance over the UTC day; ASAP pacing uses eligible inventory as it becomes available. Both remain subject to funds, approvals, caps and placement rules.

Equal allocation is useful for initial creative testing. Performance allocation uses observed clicks or attributed conversions, with continued exploration. This is a statistical rule, not a guaranteed CPA, a validated predictive bidding model or proof that a creative caused a lift.

Use the assistant for a reviewable draft

The planning assistant can help with briefs and copy. The agency reporting assistant can draft commentary from an authorized saved snapshot, after consent to share its aggregates with Groq. Both need human review.

Headline scores in the Tools hub come from an English-copy editorial rubric. They are not CTR forecasts, a policy approval or an indication of likely revenue.

A useful review routine

• Check tracking and date ranges before interpreting a change.
• Compare outcomes and spend, not only clicks.
• Treat small samples as inconclusive.
• Review a proposed change and its budget impact.
• Keep a record of what changed and compare a sensible later period.