Is It You or Is Meta Broken? A 15-Minute Triage

Every week, the big advertising communities fill up with the same thread. "Zero sales since yesterday. Spend is normal. Clicks are normal. Is Meta broken?" Hundreds of replies pile up. Half say the platform is cooked. Half say check your own account. Nobody actually knows, because almost nobody runs the checks that separate the three possible causes. This post is that triage. It takes about 15 minutes, and at the end you will know which of three very different problems you have.

The three suspects

When performance falls off a cliff overnight, there are only three places the problem can live:

• Measurement broke. Sales are still happening, but your tracking stopped recording them. The business is fine. The dashboard is lying.
• Delivery changed. Meta started sending you different, worse traffic. Sales genuinely stopped, and the cause is on the platform side or in how your account interacts with it.
• Demand changed. The traffic is the same quality as always, but people stopped buying. Seasonality, a competitor, a price problem, a broken checkout.

These three need opposite responses. If measurement broke and you "fix" it by rebuilding campaigns, you destroy weeks of learning for nothing. If delivery changed and you spend the day debugging your pixel, you burn budget while the real problem continues. And if demand changed, nothing inside Ads Manager will fix it no matter how many knobs you turn. The entire point of the triage is to avoid treating the wrong disease, because the treatments themselves have side effects.

Check 1: Did sales stop, or did reported sales stop? (3 minutes)

This is the single highest-value check in advertising, and almost nobody does it first. Open your backend, meaning Shopify orders, Stripe payments, your CRM, whatever records real money. Compare yesterday's actual order count against what Ads Manager claims for the same day.

• Backend has sales, Meta shows zero: your tracking broke. Stop here. You do not have an ads problem, you have a measurement problem. Head to how to tell if your Meta pixel is broken and work the checklist. Do not touch your campaigns while tracking is down, and when it comes back, remember the platform optimized on missing data the whole time.
• Backend is also at zero: the sales drop is real. Move to check 2.

One warning while you are here: match the date ranges and time zones exactly. Ads Manager reports in your ad account's time zone, your store may report in another. A sloppy comparison creates fake mismatches and sends you chasing ghosts. Also remember attribution lag: Meta keeps assigning conversions to past days for up to a week, so yesterday's reported number will grow. For a broken-versus-working verdict that lag rarely matters, a real outage reads as zero against a backend full of orders, but for anything subtler, compare completed days, not this morning.

Check 2: Did traffic quality collapse? (4 minutes)

If real sales stopped, the next question is whether the visitors changed. Open GA4 (or whatever analytics you trust) and look at engagement metrics for paid traffic only, comparing the bad days against the good days:

• Average session duration. If it fell from 90 seconds to 10, the clicks you are buying are not the same humans you were buying last week. During bot-traffic waves, advertisers report session times collapsing to a few seconds while click-through rate looks normal or even improves.
• Pages per session and add-to-cart rate. Junk traffic bounces off the landing page. Real buyers browse.
• Geography and device mix. A sudden surge from countries or device types you never saw before is a delivery-quality problem, not a creative problem.

If engagement collapsed, you are in delivery-change territory. Check your placement breakdown, especially Audience Network, review any automatic enhancements or recommendations that recently switched on, and check whether the drop lines up with a budget change on your side. Scaling budget quickly pushes Meta into wider, colder inventory, which looks exactly like "the algorithm broke" but is actually the auction responding to what you asked for.

Check 3: Is it everyone, or just you? (3 minutes)

Now, and only now, is the moment for the community check. Search the advertising subreddits and X for posts from the last 24 hours. You are not looking for one other person with problems, because on any given day someone somewhere is having a bad day. You are looking for a cluster: many advertisers in different niches reporting the same symptom starting at the same time.

Here is the honest limit of this check, and it is important. True platform-wide outages are rare. Most of the time, the people confidently posting "Meta is broken" are experiencing one of the two problems above and have not run the checks. A crowd of unverified reports is weak evidence. Treat the community signal as a tiebreaker, never as the diagnosis.

Check 4: The boring suspects (5 minutes)

If sales are truly down, traffic quality looks normal, and no platform-wide wave exists, work through the mundane list. In our experience this list finds the culprit far more often than anything exotic:

1. Checkout actually works. Place a real test order on mobile, in the Instagram in-app browser if that is where your traffic comes from. A payment provider hiccup or a broken button on one browser can zero out sales while every dashboard looks healthy.
2. Billing and account status. A failed card, a spending limit reached, an ad set silently paused by a rule.
3. Recent changes, by anyone.Theme updates, new apps, consent banner changes, a developer "cleaning up" the tag manager. Most tracking and site breakage traces back to a change someone made without telling the person who runs the ads. Our guide on what breaks tracking after a website migration covers the usual damage.
4. The calendar. Holidays, paydays, seasonality in your niche, a competitor running a huge sale. Demand problems are real problems, they are just not fixable inside Ads Manager.

What not to do while you diagnose

The most expensive mistakes in these situations come from acting before diagnosing. Three specific things to resist:

• Do not rebuild campaigns on day one. Duplicating and relaunching resets learning. If the real problem was measurement or a temporary delivery wobble, you just converted a two-day problem into a two-week one.
• Do not turn knobs while tracking is broken. Every change you make while the platform cannot see conversions is a change made blind, and the algorithm learns from the broken period unless you exclude it later.
• Do not accept "wait and see" as a diagnosis. Waiting is sometimes the right treatment, but only after you know which problem you are treating. Waiting on a broken pixel costs you the entire waiting period.

A worked example: the triage on a real case

Here is the shape this takes in practice, based on a real pattern we see constantly. A solo operator runs lead gen for a high-ticket service. Thirteen days in, things are great: 19 booked calls, a 14x return on spend. Then leads stop dead for three days. Ads still spending, clicks still arriving. The forums tell him it is an algorithm update. His gut tells him he scaled budget too fast. Both feel plausible, and neither is checkable from Ads Manager.

Check 1: his calendar bookings, the backend truth, also show zero. So the drought is real, not a reporting gap. Check 2: session times on the landing page look normal, geography looks normal, so the traffic is still human. Check 3: no platform-wide wave in the communities that day. Check 4, the boring list, is where it cracks open. Two findings: Meta support had just fixed a domain block on his pixel, meaning the system had been optimizing with degraded signal for days, and one real user reported the landing page CTA doing nothing on their phone. Neither finding was visible in any ads dashboard. Fifteen minutes of triage replaced a week of guessing, and more importantly, it stopped him from rebuilding a campaign structure that was never the problem.

Notice the general lesson in that story: the answer lived in the gap between systems. The ad platform, the analytics, the backend, and the actual website each held one piece. No single dashboard would ever have assembled it.

When it really is Meta

For completeness: yes, sometimes it genuinely is the platform. Delivery systems have rough days, reporting pipelines lag by hours, and entire regions occasionally see auction weirdness that resolves on its own. When your triage comes back clean on all four checks and a credible cluster of other advertisers shows the same timeline, the right play is deliberately boring: document the dates and the evidence, make no structural changes, and give it 48 to 72 hours. Platform wobbles pass. Rebuilt campaigns do not un-rebuild. If the incident affected your conversion reporting rather than delivery, note the window so you can discount that data later. The discipline of doing nothing, on purpose, with evidence, is completely different from waiting because you have no idea what is wrong. One is a strategy. The other is a coin flip that costs your daily budget per flip.

The 5-minute daily habit that makes the triage unnecessary

Advertisers who never post the panic thread tend to share one habit: a short daily glance at the numbers that diverge first when something breaks. If you want to build it manually, it looks like this:

• Backend orders vs platform-reported conversions, yesterday. The ratio between these two is your early-warning system. It is never exactly 1, and it does not need to be. What matters is that it is stable. A ratio that has hovered around 0.8 for a month and suddenly reads 0.3 means measurement broke, hours or days before performance reporting will make that obvious.
• Event Match Quality on your purchase event. EMQ dropping a full point is a leading indicator that something upstream changed, a consent tool, an app update, a form change, even when event volume looks untouched.
• Delivery sanity. Spend pacing normally, frequency not spiking, no new placement suddenly eating budget.

Five minutes, every morning, and the "is it me or is Meta" question gets answered before it is ever asked. The honest problem with the manual version is that nobody sustains it. It survives about two weeks of enthusiasm, then a busy Tuesday happens, and the one morning you skip is the morning after something broke.

Why this keeps happening to you

If you have run this triage more than once, notice what it actually consists of: comparing platform numbers to backend numbers, watching signal quality, and checking whether events are still flowing. None of that is analysis you should have to do reactively at 11pm while budget burns. It is monitoring, and machines are better at it than panicked humans.

That is the reason we built Taglert. It connects to your Meta and Google accounts read-only, watches your pixels, conversion events, and match quality around the clock, and emails you the moment something stops looking right. The 15-minute triage still works. Getting the answer before you had to ask the question works better. The math on what silent breakage costs is in the hidden cost of broken tracking.