After Meta removed the 7-day-view and 28-day-view attribution windows from its Ads Insights API on January 12, 2026, a good Meta audit is no longer about creative volume or audience precision. It is about whether your measurement plumbing can still see conversions Meta now refuses to model on weak signal.
To audit a Meta Ads account in 2026, work down a causal chain rather than a feature list: account and campaign Structure feeds Signal quality, which feeds the learning phase, which feeds automated delivery, which is only worth trusting if Measurement is configured to credit the right conversions. Audit those four layers in order, because you cannot trust a single delivery number until you have verified that signal quality and attribution settings are sound. This 25-point checklist walks each check with what to look for, why it matters, and the Meta documentation that governs it.
A Meta Ads audit is not about hunting for one broken toggle. It is about pressure-testing every layer of the account against the question Meta itself now forces: can the system still see and credit the conversions that matter, given how much of the auction it has automated and how much of the old measurement surface it has removed. The checklist below is ordered the way Cortex runs a real audit, and it deliberately skips knobs Meta has retired, like the customer budget cap and the long view-through windows, in favor of the four layers that survived consolidation.
Most 2024-era Meta checklists still audit controls that no longer exist. This one flags the specific 2024 to 2026 platform changes that silently break legacy setups: Campaign Budget Optimization renamed to Advantage Campaign Budget, Advantage+ Shopping folded into Advantage+ Sales, and the removal of the 7d_view and 28d_view attribution parameters announced October 13, 2025 and enforced January 12, 2026. Tick each item as you complete it; your progress saves in this browser, so you can close the tab and resume where you left off.
If structure is wrong, every fix below gets harder. These seven checks confirm you own the assets, can read the account at a glance, and have consolidated spend enough to clear the learning phase.
The ad account, Pixel or dataset, Pages, and catalog are owned by one Business Manager, not a personal profile.
Assets stranded on a personal account or an agency's Business Manager can be lost on offboarding and break signal sharing.
Consolidate ownership under the client's Business Manager and grant agency access via partner sharing, not transfer.
Names encode objective, audience, and creative theme so the account is legible without opening each level.
Without a convention, reporting becomes archaeology and you cannot reliably compare like with like over time.
Adopt a fixed pattern such as Objective_Audience_Theme_Date and rename going forward; do not rename live ads mid-learning.
Each campaign uses the right post-ODAX objective (Sales, Leads, Engagement, Traffic, Awareness, App promotion) for its goal.
A mismatched objective optimizes toward the wrong event, so Meta spends to win cheap clicks instead of profitable purchases.
Confirm Sales campaigns optimize for a purchase conversion event, not Add to Cart or Landing Page View.
The campaign uses one budget at the campaign level (Advantage Campaign Budget, formerly CBO) or per-ad-set budgets on purpose.
Advantage Campaign Budget lets Meta allocate spend across ad sets, which helps when ad sets are comparable but can starve a strategic one.
Use campaign budget when ad sets share a goal and audience scale; use ad-set budgets when you must protect a specific segment.
Each active ad set can realistically reach about 50 optimization results in the week after its last significant edit.
Below roughly 50 results per week an ad set tends toward Learning Limited, where delivery is less stable and CPA is typically higher.
Merge overlapping ad sets and pick an optimization event frequent enough to accumulate 50 results within the week.
Cold and warm audiences sit in distinct campaigns or ad sets, but the account is not splintered into dozens of tiny segments.
Mixing prospecting and retargeting muddies attribution, while over-segmentation starves every ad set below the learning threshold.
Keep a small number of broad prospecting ad sets and one consolidated retargeting effort rather than many micro-audiences.
Any housing, employment, credit, or social-issue campaign is declared as a Special Ad Category.
These categories remove lookalikes and detailed targeting expansion, restrict age to 18 to 65+, and prohibit ZIP-code targeting and audience exclusions.
Flag the category at campaign creation and rebuild targeting within the allowed surface, since misdeclaration risks account-level enforcement.
Structure is the layer almost nobody audits, because it feels like plumbing. But you cannot steer spend you cannot read, and you cannot trust a cost-per-result that came from an ad set Meta never let out of the learning phase. Cortex, Capconvert paid media desk
This is the new center of gravity. After the attribution change, weak signal is not just inaccurate reporting; it is delivery Meta will not model. These seven checks verify the account is feeding clean, well-matched server events.
Key events fire both from the browser Pixel and the server-side Conversions API, sharing an event ID for deduplication.
Browser signal alone is increasingly lost to ad blockers and tracking restrictions, so CAPI recovers conversions Meta would otherwise miss.
Confirm matching event_id and event_name across both channels in Events Manager so the same purchase is not double-counted.
Each important server event has an Event Match Quality score reviewed in Events Manager, where EMQ is rated 0 to 10.
EMQ reflects how well customer-information parameters match events to Meta accounts, which directly governs targeting and attribution accuracy.
Raise low scores by sending more and better-quality parameters; practitioners treat 8 and above as strong, though Meta publishes only the 0 to 10 scale.
High-value parameters are present first: Email and Click ID are High priority; Facebook Login ID, birthdate, country, phone, external ID, and browser ID are Medium.
Meta ranks parameters by matching priority, so missing the High-priority fields caps how high your EMQ can climb.
Add hashed email and the click ID first, then layer in Medium fields; lead, name, city, and ZIP are Low priority and help least.
The Conversions API connection has sent matched website events within the last 48 hours, the window Meta uses to calculate EMQ.
EMQ is calculated only on the last 48 hours of data and is available only for Conversions API website events, so a stalled feed shows no score.
Check the dataset for recent server events and fix any broken integration before judging delivery, since a dead feed silently degrades targeting.
Every conversion domain is verified, and the eight Aggregated Event Measurement events are ranked with the primary purchase event first.
If the priority order is wrong, an opted-out user's most valuable conversion may not be the one Meta records.
Verify the domain in Business Settings and set purchase as the top-priority event in the Aggregated Event Measurement configuration.
Custom audiences draw on uploaded customer lists and server events rather than relying solely on browser-side pixel activity.
First-party signal is now the foundation of durable audiences, since browser-based retargeting pools shrink as tracking restrictions tighten.
Sync hashed customer lists and CAPI-based events to feed Custom Audiences and the lookalike or Advantage+ seed.
The consent management platform passes consent state correctly to both the Pixel and the Conversions API.
A misconfigured CMP can drop consented events or send non-consented ones, either starving signal or creating compliance risk.
Verify consent flags reach both channels and that event volume tracks expected consent rates rather than collapsing.
With targeting increasingly automated, creative and the few audience constraints that remain are where human judgment still moves results. These six checks separate helpful automation from quiet damage.
Each creative enhancement (such as text improvements, image expansion, or music) is reviewed and toggled with intent, not left fully on by default.
Some enhancements lift performance, but others distort brand-controlled creative or overlay text in ways a regulated brand cannot accept.
Enable the enhancements that preserve message fidelity and disable any that alter claims, pricing, or required disclosures.
Creative is supplied in the right aspect ratios for every active placement, including vertical for Reels and Stories.
A single horizontal asset stretched across all placements wastes impressions in feeds where it looks broken or gets cropped.
Upload placement-specific assets or use placement customization so each surface receives a correctly framed version.
The number of active ads per ad set is high enough to test variety but low enough that each can gather meaningful results.
Too many simultaneous ads split the roughly 50 weekly results across creatives, slowing every one out of the learning phase.
Keep a focused set of strong creatives live and retire underperformers rather than stacking dozens that never accumulate signal.
Video creative opens with a clear hook in the first seconds and reads as native to the feed rather than a repurposed TV spot.
Most scroll-past decisions happen in the opening moments, so a weak hook caps performance no matter how good the targeting is.
Lead with the value or pattern interrupt up front, add captions, and frame for sound-off vertical viewing.
Frequency per audience is tracked over a 7-day window so saturation is caught before performance decays.
Note that Meta limited frequency breakdowns in the Ads Insights API to 6 months of history, so long-range frequency trend analysis is constrained.
When frequency climbs and results soften, refresh creative or widen the audience rather than simply raising the budget.
Where Advantage+ Audience is used, hard constraints are limited to the essentials such as location and a minimum age, with suggestions used as a seed.
Over-constraining automated audiences forces Meta back into a small pool, undoing the discovery benefit of the automation.
Provide audience suggestions as a hint, keep mandatory exclusions minimal, and let delivery expand beyond the seed.
The final checks make sure the numbers driving every decision above are honest after the January 2026 attribution change. The old view-through cushion is gone, and reporting that still assumes it will overstate Meta's contribution.
Reporting uses surviving windows only: 1-day click, 7-day click, 28-day click, 1-day engaged-view, and 1-day view.
Meta removed 7d_view and 28d_view from the Ads Insights API on January 12, 2026, so any dashboard built on those parameters is now broken.
Rebuild reports on the surviving windows and re-baseline historical comparisons, since pre- and post-change numbers are not like for like.
Meta's reported conversions and ROAS are compared against a blended view, marketing mix modeling, or analytics such as GA4.
Each platform claims credit on its own terms, so a single in-platform ROAS rarely matches the cash reality of the business.
Trend Meta-reported revenue against an independent source and decide spend on the blended picture, not the platform's self-report.
You know roughly what portion of reported conversions Meta is estimating rather than directly observing.
As observed signal weakens, modeling fills more of the gap, so a great-looking number can rest on estimation you cannot bank.
Improve signal quality first (Checks 8 to 14) so Meta has more observed data to model from, then weight decisions toward observed results.
Performance judgments lean on click-based windows, with view-through treated as directional rather than a primary KPI.
With 7d_view and 28d_view no longer queryable in the Ads Insights API and reporting standardizing on shorter windows, long view-through credit can no longer be pulled to flatter results.
Set primary KPIs on click windows and use the 1-day view window only as a secondary, clearly labeled signal.
Long-range reports account for Meta limiting unique-count fields and hourly breakdowns to 13 months and frequency breakdowns to 6 months.
Aggregate metrics still support up to 37 months, but anyone pulling unique or hourly history beyond 13 months will now hit a wall.
Archive key metrics on a fixed cadence so year-over-year analysis survives the retention limits rather than relying on the API to remember.
All 25 checks condensed to one line each, grouped by the four causal layers for quick hand-off to a client or a junior auditor.
Budget two to four hours for a full 25-point pass. Start with signal, not delivery: verify that the Conversions API is live, deduplicated against the Pixel, and producing a healthy Event Match Quality score within the last 48 hours. You cannot trust any cost-per-result until you know Meta can actually see the conversions, which is why the audit runs Structure, then Signal, then Creative, then Measurement.
Only with reconciliation. On January 12, 2026 Meta removed the 7-day-view and 28-day-view windows from the Ads Insights API, leaving 1-day click, 7-day click, 28-day click, 1-day engaged-view, and 1-day view. Reporting built on the old view windows is now broken and pre-change numbers are not directly comparable. Treat Meta's in-platform ROAS as one input and reconcile it against a blended source like marketing mix modeling or GA4.
Meta scores EMQ from 0 to 10 based on the quality of customer-information parameters and the share of events matched to Meta accounts, using only the last 48 hours of Conversions API website data. Practitioners commonly treat 8 and above as strong, but those named tiers are a convention, not a Meta-published threshold. To improve it, send the High-priority parameters first, hashed email and the click ID, then layer in Medium-priority fields such as phone, external ID, and country.
A Google Ads audit centers on search terms, keywords, match types, and negative lists, because intent is expressed as a query. A Meta audit has no query; the center of gravity is signal quality and the learning phase, because Meta infers intent from your conversion events. So a Meta audit spends most of its time on the Conversions API, Event Match Quality, and attribution settings, where a Google audit spends it on the search term report. Many agencies run both checklists for the same client.
Not wholesale. Audit them one by one. Some enhancements genuinely lift performance, while others alter your creative in ways a regulated brand cannot accept, such as overlaying text or changing how a claim or price reads. Keep the enhancements that preserve message fidelity and disable any that touch claims, pricing, or required disclosures.
As few as it takes for each active ad set to reach roughly 50 optimization results in the week after its last significant edit. Below that, an ad set tends toward Learning Limited, where delivery is less stable and CPA is typically higher. There is no fixed count; the rule is to consolidate overlapping ad sets and choose an optimization event frequent enough to clear 50 results, rather than splintering spend across many micro-segments.