The Weekly Meta Account-Health Audit Every Operator Runs

Most account problems don’t announce themselves. A pixel quietly stops deduplicating, a card declines on a Sunday, one ad gets a soft policy flag, and delivery starves for four days before anyone notices the revenue dent. By then the fault isn’t a five-minute fix — it’s a recovery project. A weekly meta ads account health audit exists to catch those faults while they’re still cheap, in the window where a single toggle solves what a week of compounding will not.

The discipline that makes this work is order. You don’t scan the account randomly — you walk it top-down, from the layers that can silently void everything beneath them to the layers that merely need tuning. Five layers, same sequence every week: integrity, payments, delivery, signal, learning. An agentic operator runs this pass continuously; a human operator runs it every Monday in twenty minutes.

For the surrounding account decisions, compare Bach.ai vs Triple Whale: Which Tool Actually Runs Your Meta Ads? and use Account Health Before Blame: The Meta Ads Diagnostic Tree as the next diagnostic.

Why order beats coverage

Each layer sits upstream of the next. A restricted asset makes your delivery analysis meaningless. A failed payment throttles the spend you’re about to “optimize.” A broken signal layer corrupts the very numbers you’d use to judge learning. If you start at the bottom — staring at ROAS and CPA — you’ll misread a plumbing failure as a performance failure and “fix” it by killing healthy campaigns. Most panic budget cuts are misdiagnosed upstream faults. Walking the layers in causal order means every check you run sits on a foundation you’ve already verified.

Layer 1: Integrity

Start where the platform can shut you down without warning.

• Account and asset status. Confirm the ad account, Business asset, and Page are all active and unrestricted. A restriction at the account level halts everything; one at the asset level can starve a single campaign while the rest run fine, which is harder to spot.
• Ad-level disapprovals and soft flags. Filter for rejected and limited ads. A “limited” status isn’t a full stop, but it suppresses reach quietly — the campaign keeps spending while delivery narrows.
• Recently edited assets. Anything touched in the last week is the plausibly thing to have broken. New creative, a swapped landing URL, an edited Page — these are where fresh policy flags hide.

The cheap-fault signature here is a single disapproved ad inside an otherwise healthy ad set. Left alone, it drags the ad set’s delivery; caught early, it’s a resubmit.

Layer 2: Payments

This is the most underrated layer because it has nothing to do with marketing skill and everything to do with revenue.

• Funding instrument health. Check that the primary payment method is valid and not near expiry. An expired card doesn’t generate a campaign alert you’ll naturally see — it generates a delivery cliff.
• Spend cap and threshold posture. If you run on a billing threshold, a recent threshold reset or an account-spend-limit you set months ago can quietly cap delivery mid-period.
• Failed-charge history. A declined charge can pause delivery account-wide until it clears. The campaigns look “paused for no reason” — but the reason is in billing, not in the ad sets.

Payments faults are the purest example of cheap-while-early. A card you update on Monday costs nothing. The same card discovered Thursday cost you three days of your best-performing window and a learning reset on top.

Layer 3: Delivery

Now — and only now — you look at whether spend is actually flowing the way you intend.

• Spend pacing. Compare planned versus actual spend per campaign. Persistent under-delivery points upstream (integrity or payments) or to an auction problem: bid caps too tight, audiences too narrow, budgets stranded in fragmented ad sets.
• Budget concentration. Check that money is sitting where intent is. Fragmentation across many small ad sets splits your conversion signal and keeps everything perpetually under-optimized.
• Frequency. Rising frequency on prospecting is an early fatigue signal long before CTR collapses. Treat a creeping frequency as a leading indicator and rotate creative before performance decays, not after.
• Auction overlap. Multiple ad sets chasing the same audience bid against you. If pacing is fine but efficiency is soft, self-competition is a common, invisible culprit.

The discipline: delivery is where most operators want to start. By arriving here third, every anomaly you find is genuinely a delivery problem, not a symptom of something you skipped.

Layer 4: Signal

Optimization is only as good as the events you feed the auction. This layer audits the data, not the ads.

• Event freshness and volume. Confirm your key optimization events are still firing at expected volume. A silent drop — a tag removed in a site deploy, a checkout flow change — looks identical to “demand fell.” It isn’t.
• Server-side and browser deduplication. If you run both browser and server events, verify they’re deduplicating cleanly. Double-counting inflates reported results and quietly mistrains delivery toward the wrong people.
• Event match quality. Weak match quality erodes attribution and optimization precision. It seldom breaks loudly; it just slowly makes every campaign dumber.
• Attribution-window drift. Make sure you’re reading results on a consistent window. A window change makes a flat account look like a swing.

Signal faults are insidious because the dashboard still shows numbers — just wrong ones. This is the layer where Bach earns its keep, watching event integrity continuously so a broken tag surfaces as a flagged anomaly instead of a month of corrupted decisions. (It surfaces the finding for your approval; it doesn’t act on its own.)

Layer 5: Learning

Last, judge performance — because now you can trust the inputs.

• Learning-phase status. Note which ad sets are still learning versus stable. Meta needs enough recent optimization-event signal to exit the learning phase; as a planning range, think on the order of dozens of conversions per ad set per week, not a precise quota. Ad sets that never accumulate enough signal stay volatile forever.
• Edit-triggered resets. Significant edits re-enter learning. If someone’s been tweaking budgets daily, you’re paying a perpetual learning tax. Check the recent-edit log against unstable ad sets.
• Efficiency against margin, not vanity ROAS. Read CPA against contribution margin and account-level efficiency (MER), not platform-reported ROAS in isolation. A campaign can post a flattering on-platform number while losing money once you account for blended efficiency.
• Patience guardrail. Resist judging anything that hasn’t cleared learning and a full attribution window. Killing an ad set mid-learning is the most expensive “optimization” there is.

Run it as a checklist

The point of a fixed sequence is that it becomes muscle memory and survives a busy week. Keep it to one pass:

Layer	You’re confirming	Cheap-fault signature
Integrity	Nothing is restricted or disapproved	One limited ad starving an ad set
Payments	Money can actually flow	Expired card → delivery cliff
Delivery	Spend pacing matches intent	Auction overlap eroding efficiency
Signal	Events fire and dedupe correctly	Silent tag drop read as lost demand
Learning	Performance judged on trusted inputs	Daily edits causing endless resets

The takeaway

A weekly meta ads account health audit is not a performance review — it’s preventive maintenance. The faults that quietly bleed accounts are almost never sophisticated; they’re an expired card, a disapproved ad, a dropped event, a budget split too thin. They’re cheap on Monday and expensive on Friday. Walk the five layers in order, top-down, so every check stands on verified ground beneath it. Twenty minutes a week, the same sequence every time, and you stop discovering problems in your revenue and start catching them in your audit.