Meta signal health checklist: what to check before you blame the algorithm

When Meta performance drifts, the first instinct is to blame targeting or creative. Often the real problem is upstream: the signal telling Meta who converted is arriving late, deduplicated wrong, or stripped of the parameters that let Meta match a conversion to a person. Optimisation is only ever as good as the events feeding it.

This is the checklist the Dispatch runs before touching a single campaign. It is a diagnostic, not a fix: it tells you where your signal is leaking, in the order worth checking. Work down it with Events Manager open.

Before you start

• Open Events Manager for the pixel in question, and have your Conversions API setup to hand (or the developer who owns it).
• Know which events actually matter to you. A perfect signal on a page-view you do not optimise for is not worth an afternoon; the purchase or lead event is.

Event match quality

• Find your EMQ score: Events Manager, select the event, open its details. Meta rates each event's match quality on a scale to ten. Roughly, below five is poor, five to seven is okay, above seven is good and worth the effort to reach.
• Read EMQ per event, not as an account average. A strong page-view score can hide a weak purchase score, and the purchase event is the one optimisation leans on.
• Treat a low score as a prompt to send more and better customer parameters (see advanced matching below), not as a number to chase for its own sake.

Event deduplication

• Confirm the SAME conversion sent by both the browser pixel and the Conversions API carries a shared event_id. That is the key Meta uses to recognise the two hits as one event and drop the duplicate.
• Also match event_name across both channels for the same action; a mismatch defeats deduplication even when the id is present.
• Check the event's diagnostics for a deduplication figure. If browser and server events are NOT being deduplicated, your conversion counts inflate and optimisation learns from phantom events.

Advanced matching parameters

• Send hashed customer information with your events: email and phone are the highest-value, then first and last name, city, state, postcode, country, and the external ID.
• Prioritise email and phone. They move EMQ more than the address fields and are usually the parameters you already hold at the point of conversion.
• Confirm the parameters are being HASHED (SHA-256) before they leave the browser or server. Meta expects normalised, hashed values; sending raw PII is both wrong and a compliance problem.

Pixel and CAPI parity

• Open the Test Events tool in Events Manager and fire a real conversion. Both the pixel event and the server event should appear for the same action.
• Check they carry the same parameters, not just the same name. A server event missing the customer parameters the browser sent is a parity gap that drags EMQ down.
• Watch for server events arriving with no matching browser event (or the reverse). One-legged events are the usual sign of a broken event_id or a CAPI payload that lost its matching data.

The consent-mode interaction (the quiet one)

• For EU and UK traffic under consent mode, a visitor who has not granted consent generates events with the customer parameters stripped or the event withheld entirely. Advanced matching depends on exactly those parameters, so unconsented EU traffic can quietly pull EMQ towards zero on that segment.
• This is expected behaviour, not a bug to fix. Read EMQ with your consent rate in mind: a "poor" EU score on a site with low consent may simply be the signal you are lawfully allowed to send.
• If EMQ fell off a cliff on a date, check whether a consent-banner or consent-mode change landed then, before you rebuild the pixel.

How to read this, and the honest limit

Run the checklist top to bottom; the earlier items gate the later ones (deduplication is meaningless if half your events never arrive). But keep EMQ in proportion. It is a proxy, not a KPI. It measures how well-matched your events are, not how many sales you made, and a campaign can perform perfectly well at a middling score. Chase EMQ only until the signal is genuinely healthy, then stop and let the account's results, cost per result and volume, be the thing you actually optimise. A signal you have made technically pristine at the cost of a week you could have spent on offers or creative is a poor trade. Fix the leaks this checklist finds, then get back to the campaign.