Meta Ads Diagnostic: How the Model Works
Most Meta ads advice tells you what to check. The diagnostic engine tells you what order to check it in and why. The triage sequence matters because fixing the wrong problem first wastes time and often makes things worse.
Why most diagnostics fail
The standard Meta troubleshooting approach is a checklist: check your pixel, check your creative, check your targeting, check your budget. The problem is that a checklist treats all problems as equally likely. They are not. Some root causes are foundational — fixing them unlocks everything downstream. Others are symptoms that look like causes but will recur until the underlying problem is resolved.
The diagnostic engine is built around a triage order. Before showing you a ranked list of root causes, it evaluates your symptoms against five categories in a fixed sequence. The sequence is not arbitrary. It reflects the dependency structure of a Meta account: later-stage problems cannot be properly diagnosed until earlier-stage problems are ruled out.
The five root causes
Why triage order matters
Consider a common scenario: CPA has been rising for three weeks and the advertiser suspects creative fatigue. They refresh the creative. CPA stays high. They refresh again. Still high. Two months later, someone checks the pixel and discovers the purchase event was misfiring — it was firing on the add-to-cart page instead of the confirmation page. The algorithm had been optimising toward cart adds, not purchases, the entire time. The creative refreshes were irrelevant.
This is why tracking is checked first. A misfiring pixel produces symptoms that look like creative problems, economics problems, and audience problems simultaneously. Until the signal is clean, you cannot trust any other diagnostic data.
Economics is checked second for the same reason. If daily budget divided by target CPA is below 1.4, the algorithm is mathematically incapable of hitting the CPA target regardless of creative or targeting quality. Diagnosing creative problems in an underfunded campaign produces misleading conclusions.
How symptoms map to root causes
Each symptom you select in the diagnostic — high CPA, performance dropping, stuck in learning, low CTR, clicks but no conversions — maps to one or more root causes. The engine scores each root cause based on how many of your selected symptoms point to it, then applies the triage order to rank them.
Two symptoms that individually suggest different causes can together point clearly to one. "Was working, now dropping" plus "CPM very high" plus "frequency above 3" produces a different diagnosis than "was working, now dropping" plus "clicks but zero conversions." The combination matters. The engine is designed to use the full symptom set rather than treating each symptom in isolation.
What the output tells you
The output has three layers. The verdict at the top commits to the most likely cause based on your dominant symptom pattern. It is intentionally specific because "check everything" is not actionable. Below that, the ranked root cause blocks show the mechanism behind each issue and include your actual numbers where possible. The symptom cards below that give you the specific Ads Manager column path, the threshold to check against, and the exact action to take.
The "not the problem" callouts are as important as the fixes. Under pressure — a CMO asking questions, a client demanding changes — the instinct is to change everything at once. The diagnostic explicitly tells you what not to touch, because simultaneous changes make it impossible to know what worked and frequently reset the learning phase.
Each issue has an escalation trigger. If CPA is 2x target after 100 conversions and 30 days, that is no longer an optimisation problem. It is a strategic problem — offer, vertical competitiveness, or audience exhaustion — that requires a different kind of conversation. The escalation trigger defines that boundary so you know when to stop optimising and start rethinking.
What the diagnostic cannot do
The diagnostic covers the 12 most common Meta symptoms and five root cause categories. It handles the majority of performance problems in accounts spending under €30k/month. What it cannot do is account for account history, competitor behaviour, seasonal factors, or the specific interaction between your offer and your audience that only someone who has looked at your account can see.
The Troubleshoot section covers 41 specific situations in more depth, including scenarios the diagnostic does not reach. If the diagnosis gives you a root cause but the fix in Troubleshoot is not enough — or if your situation does not fit the model — that is where a direct conversation makes sense.