Why the Optimizer's recommendations changed (tracking & signal confidence)

Last updated July 19, 2026

If the Optimize audit suddenly stops recommending pauses, ranks ads in a way that doesn't match your ROAS column, or leads with a Tracking finding instead of the usual savings, it's almost always the same cause: the audit decided it can't fully trust your conversion data, and deliberately changed how it judges your ads.

This is intentional. Pausing an ad because a broken pixel recorded zero purchases is the single most expensive mistake an optimizer can make, so the audit degrades gracefully instead of acting confidently on bad numbers.

What the audit measures first

Before scoring a single ad, the audit works out two things about your account.

Which result it should judge by — derived from what your campaigns actually optimize for and report:

ROAS — purchases with a revenue value attached.
Cost per purchase — purchase-optimized, but no value coming back.
Cost per lead — lead-optimized campaigns.
No conversion mode — nothing measurable is being reported at all.

How trustworthy that signal is — read from your pixel and Conversions API health.

The three confidence levels

Full signal. Your pixel is reachable, fired recently, and server-side (Conversions API) events are present. The audit judges ads normally: ROAS, cost per result, purchase counts, the usual pause and scale recommendations.

Partial signal. Conversions are being tracked, but under-counted. This is what you'll see with a browser-only setup (no Conversions API), a pixel that hasn't fired in over 24 hours, or one Meta has flagged as unavailable.

When confidence is partial, the audit stops judging ads on purchase counts and ROAS, because those numbers are unreliable. Instead it falls back on signals that survive broken attribution:

click-through rate and cost per click,
add-to-cart rate,
begin-checkout (initiate checkout) rate.

Two consequences are worth understanding, because they're the ones that surprise people:

An ad showing zero purchases but strong upstream intent is protected — it will not be recommended for pausing on a missing purchase count alone.
Only ads that burned real clicks with no engagement and no intent at all get flagged. Everything in between is genuinely ambiguous, so the audit leaves it alone rather than acting on a broken signal.

That's why a partial-signal audit returns fewer recommendations than usual. It isn't that your account got cleaner — the audit is refusing to guess.

No signal. No conversion tracking at all. Conversion-based detectors switch themselves off entirely and you'll only get engagement-level observations such as click-through rate. Fix tracking before drawing conclusions from anything here.

The findings that tell you which one you're in

The audit surfaces this through two advisory findings under the Tracking check and Tracking & signal health themes. Both are informational — there's no one-click fix, because the cause sits in your pixel setup rather than in your ad account.

"Tracking looks broken — no recent pixel events" — your pixel hasn't recorded events recently (or Meta flagged it unavailable). Highest priority: nothing downstream is trustworthy until this is fixed.
"Conversion tracking looks off" — a meaningful share of your active spend sits on ads that drew real link clicks and recorded zero results, while the rest of the account converts fine. That pattern points at a Conversions API gap, a misconfigured event, or a broken landing page — not weak creative. The finding says it explicitly: verify your pixel fires and the right conversion event is selected before pausing those ads.

Both findings stay silent when everything is healthy, so seeing them at all is the signal.

What to do about it

Check the pixel is firing. In Meta Events Manager, confirm your pixel has recent events and isn't flagged unavailable.
Add the Conversions API. A browser-only setup is the most common cause of partial confidence — browser events are lost to ad blockers and privacy settings, so purchases get under-counted even when nothing is broken.
Confirm the right event is selected on the campaigns in question.
Re-run the audit once tracking is fixed. Confidence is recalculated from scratch on every audit, so recommendations return to normal on the next run.

Frequently asked questions

Why did it stop recommending pauses? Almost certainly partial confidence. It's protecting you from pausing ads whose conversions simply aren't being recorded.
My cart-to-purchase rate is low. Does that alone downgrade me? No. If your pixel and Conversions API are healthy, the audit trusts them regardless of funnel shape — a low cart-to-purchase rate is normal. The funnel is only used as a fallback when pixel health can't be read at all, and then only when conversions are almost entirely absent against a large upstream volume.
No pixel showed up at all — why no warning? An unreachable pixel is usually a permissions gap on the connection rather than a broken setup, so the audit stays silent instead of raising a false alarm. See How to reauthorize or disconnect your Meta account.