How AEM ranks events when consent is missing

What AEM constrains#

When visitors opt out of tracking on Apple devices and similar surfaces, per-person attribution ends and aggregation begins. AEM is the rulebook for that world: each verified domain configures a ranked set of conversion events, and opted-out visits are measured against that set in aggregate, with delays and modelled gaps where data cannot be observed directly. Start at the Meta Ads hub for the family view.

The practical effect is scarcity: measurement that once tracked everything now rations attention to eight ranked slots. Every inclusion is a choice and every exclusion a blind spot, so configuration becomes strategy rather than admin. Meta own AEM overview states the current rules.

The 8 slots and the unsafe default#

Each domain gets 8 slots, and the out-of-the-box order is unsafe for most advertisers: informational events such as page views and content views sit above revenue events, so opted-out visits count browsing while purchases go unrecorded. Optimisation then trains on the top slot, which means it trains on the wrong behaviour. Assume the default is wrong until you have ranked it yourself.

Changing priority pauses opted-out measurement for roughly three days while the new configuration propagates, so batch ranking decisions rather than dripping them. Document the intended order, apply it once, and leave it alone through the propagation window. Domain verification must be in place before any of this applies.

Ranking for revenue#

Rank by commercial value, not by volume: Purchase first for shops, qualified Lead or CompleteRegistration first for lead businesses, then the supporting funnel steps in descending money order. Page views and generic content views belong last or nowhere, because a slot spent on browsing is a slot stolen from buying. Review the ranking against margin quarterly, not just revenue.

Keep the list short on purpose: fewer configured events concentrate opted-out signal into the actions that matter, while a full shelf of marginal events dilutes it. Each slot should answer which budget decision it informs; slots that inform nothing are decoration. Event vocabulary is set in standard events.

Value sets and optimisation#

Value optimisation under AEM needs value sets configured on the ranked events, so delivery can distinguish a large order from a small one even in aggregated traffic. Without value sets, bidding sees counts where it should see money, and ROAS targets steer by volume. Configure value ranges that reflect your real order distribution, then confirm reporting shows value arriving.

Match the campaign optimisation event to a ranked slot, or delivery optimises toward an event AEM cannot measure on opted-out visits. The campaign, the ranking and the value set form one chain; audit all three together. Developer configuration detail is at AEM developer docs.

Reporting under AEM#

Expect delays, gaps and modelled fills: opted-out conversions arrive late, some never resolve to person level, and modelled conversions estimate the unobservable remainder. Read performance in weekly cohorts rather than daily spikes, and judge tests on directional consistency across weeks instead of single-day wins. Modelled conversions explains the estimates.

Reconcile platform counts against your order book monthly and track the gap as a metric of its own; a widening gap names a configuration or consent fault before it names a creative one. Our mismatch guide reconciles platform and analytics counts.

AEM maintenance#

Revisit configuration quarterly and after every range, pricing or funnel change: new products need new thinking about which events earn slots, and retired offers should release theirs. Freeze changes through peak trading, since propagation pauses plus volatility make December the worst month to re-rank. Log every change with dates and reasons.

Pair AEM discipline with full-funnel signal: CAPI cover for observed traffic, clean dedup for honest counts, and our team for a pre-peak configuration review.

Frequently asked questions#