What to verify at the reporting level first
Optimization starts with trust in the numbers. If the conversion data feeding the account is incomplete, every decision built on it is unreliable. Pixel-only tracking misses a significant share of conversions to browser restrictions, ad blockers, and cookie loss. The reviewer should confirm server-side tracking is active, the Event Match Quality score is above 6.0, and browser and server events are properly deduplicated before any lower-level recommendation.
A reporting gap does not need to be perfect to be actionable, but it needs to be named. If the reviewer cannot confirm which conversions are observed versus modeled, or whether key events like purchases are arriving with complete customer parameters, every recommendation that follows should carry a caveat. Reporting confidence is the foundation. Nothing below it is reliable until it is confirmed.
- Confirm server-side tracking with EMQ above 6.0. Below that, the algorithm bids on partial data
- Check pixel and server event deduplication. Double-counted conversions inflate ROAS
- Verify purchase events carry full parameters: email, phone, and external ID for matching
- Name the gap if reporting is incomplete. A caveat is better than optimization on broken numbers
How to diagnose a campaign structure problem
The most common structural error is conflating what belongs at each level of the campaign hierarchy. Campaign level sets the objective. Ad set level controls audience, placement, and budget. Ad level delivers creative. When budget is set at ad-set level but the algorithm needs campaign-level freedom, or when creative variation is crammed into too few ad sets that starve the learning phase, the account underperforms for structural reasons no amount of bid adjustment can fix.
Campaign consolidation consistently outperforms fragmentation. Fewer ad sets per campaign mean faster learning, lower CPMs, and more predictable delivery. The reviewer should check whether the campaign objective matches the actual conversion event the business cares about, whether the budget mode is right for the spend level, and whether ad sets are fragmented across overlapping audiences that bid against each other in the auction. A structural fix at the campaign level compounds every optimization beneath it.
- Match the campaign objective to the real conversion event. Optimizing for clicks when you need purchases trains the algorithm wrong
- Audit ad set fragmentation. More than 3 to 5 ad sets per campaign under CBO dilutes learning
- Check for audience overlap. Overlapping ad sets bid against each other and inflate CPM
- Confirm CBO vs. ABO choice. CBO needs roughly $50 per day per ad set to work cleanly
What to check at the ad set level when delivery stalls
Ad set constraints sit at the middle layer where audience size, placement selection, budget, and learning-phase status all interact. When cost per result rises without a creative change, the reviewer should check whether the ad set is constrained by audience size, frequency, or a learning phase that never completed. Vertical scaling on the same audience hits saturation fast, typically above a 3x frequency on the core segment.
The learning phase requires roughly 50 conversion events per ad set per week to exit. If the budget is too low to generate that volume, the ad set will never optimize and performance will remain unstable regardless of creative quality. Advantage+ placements should be the default. Manual placement restrictions limit the algorithm's delivery paths and raise CPM without improving conversion quality. An ad set stuck in learning is not an optimization target. It needs a budget or audience fix, not a creative one.
- Check learning phase status. If the ad set has not exited, performance data is unreliable
- Confirm 50-plus conversions per week. Below that, the ad set cannot complete the learning phase
- Watch frequency above 3.0 in prospecting. Rising CPA with steady creative signals audience saturation
- Default to Advantage+ placements. Manual restrictions raise CPM without improving conversion quality
How to read ad-level creative quality and fatigue
Creative quality is not a one-time check. It is a moving target. The reviewer should confirm whether each creative in the account has a distinct concept, or whether the ad pool is repeating the same angle in slightly different packaging. Creative iteration means changing the hook. Creative variation means changing the entire concept. An account full of iterations on one concept will fatigue against the same audience pool regardless of how many ads are running.
The first sign of fatigue is not CPA. It is a drop in hook rate and 3-second video view rate, followed by CTR decline, then CPM inflation. By the time CPA visibly spikes, the creative has already been decaying for several days. The reviewer should set fatigue triggers on frequency above 3.0 for prospecting and a 15 percent CTR drop from the 7-day peak. A creative showing both signals should be flagged for replacement, not re-optimized.
- Audit creative diversity. If every ad uses the same concept with different hooks, the account is iterating, not testing
- Watch hook rate and 3-second view rate first. Both decay before CTR and give the earliest warning
- Flag creative when frequency crosses 3.0 and CTR drops 15 percent from peak. Both signals together mean it is done
- Refresh creatives before fatigue hits. A replacement launched during decay costs more than one launched proactively
When to hold versus approve a scaling recommendation
Vertical scaling means increasing budget on an existing proven ad set. It works until the audience pool is saturated. Horizontal scaling means adding new winning ad sets with fresh creative angles and audience segments. It compounds without the same ceiling. A recommendation to vertically scale without confirming the current ad set still has headroom on frequency and CPM is not a scale decision. It is a spend increase with unknown efficiency.
Even a campaign with stable CPA may be capturing demand that would have converted organically. Before approving scaling, the reviewer should confirm the recommendation names the diagnosed level, the specific constraint, the proposed change, the expected consequence, the observation window, the success criteria, and the rollback plan. If any of these are missing, the memo stays on hold. A scale recommendation without a rollback plan is a budget decision without a circuit breaker.
- Confirm vertical scaling has headroom. Frequency below 3.0 and stable CPM mean room remains
- Shift to horizontal scaling when vertical headroom runs out. New angles on fresh segments compound without saturation
- Run a conversion lift test before large budget increases. Stable CPA may be captured demand, not incremental growth
- Require a rollback plan. The observation window, success criteria, and rollback trigger must be documented
Sample Review Note
The reviewer confirms reporting confidence: server-side tracking is active with EMQ above 6.0, events are deduplicated, and missing conversion parameters are caveated. Campaign objectives match real conversion events, ad set fragmentation is within 3 to 5 per campaign, and audience overlap is checked. Every ad set has exited the learning phase with 50-plus weekly conversions. Creative diversity is audited across concepts, and fatigue triggers are active on frequency above 3.0 and 15 percent CTR decline.
The recommendation names the diagnosed level, the specific constraint, the proposed change, the expected consequence, the observation window, the success criteria, and the rollback plan. If any reporting gap, structural constraint, learning-phase block, fatigue signal, or scaling condition is modified after this review, the memo is gated for recheck. The next action stays approval-gated until the media lead accepts the evidence. An optimization built on incomplete data is not optimization. A scale recommendation without a rollback plan is a spend increase without a safety net.