What's a Good EMQ Score? (And the 9/10 Trap)

What EMQ actually measures (and doesn't)

Event Match Quality scores, per event, how well the customer information on your server events lets Meta match events to real accounts — more matched events, better attribution and delivery signal. What it does NOT measure: your tracking's correctness (dedup can be broken at EMQ 9), your ads' performance, or some pass/fail threshold Meta enforces. It's a coverage metric, and it earned its 2025–26 obsession status honestly — matching IS the post-cookie currency — but the obsession outran the mechanics: the score is a means (matched volume) wearing the costume of a grade, and media buyers grinding from 6.2 to 6.8 for a week are usually polishing the costume.

The benchmarks nobody publishes plainly

Bands apply to identity-bearing events (Purchase, Lead). Upper-funnel events score lower by nature — that's anatomy, not failure.

Per-event ceilings differ by anatomy — Purchase carries identity, PageView is anonymous strangers. Score accordingly.

The part that saves you a week: volume beats polish

Here is the mechanic under the contrarian claim. Meta's delivery learns from matched conversions — the absolute count of events it could tie to accounts. EMQ is the match rate's proxy; the optimization fuel is the matched volume. Raising EMQ from 6 to 9 on the same traffic might convert a 1,000-conversion month into 1,010 matched — real, marginal — while the same week of engineering pointed at conversion rate, creative testing or (if signal is genuinely thin) capturing email earlier in the funnel moves matched volume by whole percentage points. That's also why the score visibly stops paying above ~6: the heavy keys (email, phone) do the matching; everything after is decimals. The posture that wins: fix 2–4 scores urgently, nudge 4–6 scores in an afternoon, and treat 6+ as a solved problem — then go do something delivery can feel.

Raising it, in leverage order

The keys in weight order — and the placement rule that matters more than any single key.

The afternoon version: send hashed email and phone on every server event that can carry them — purchase obviously, but also signups, lead forms and account creation, because every capture point is a chance to teach Meta who this browser belongs to before the purchase moment. Add fbp/fbc pass-through (knowing Safari now eats fbc for a third of B2C traffic — the industry-wide EMQ dip of fall 2025 in one sentence), then the cheap adders: name, zip, city, country, all hashed per spec. Verify in Events Manager's per-event EMQ details, which literally list the keys received and their coverage — the rare Meta screen that tells you exactly what's missing. And mind the two classic own-goals: malformed hashing (wrong casing/whitespace before the hash → valid-looking keys that match nothing) and consent-blocked parameters (EU setups stripping the very fields EMQ needs — a compliance choice to make knowingly, not a bug to discover in a score).

"My EMQ dropped overnight" — the triage

In order of likelihood: (1) Calculation noise — the score runs on a ~48-hour window; small accounts see it wobble a point on traffic mix alone. Wait two days before touching anything. (2) Traffic-mix shift — a new campaign flooding anonymous cold traffic dilutes match rates mechanically; the score fell because your funnel widened, which is fine. (3) Platform changes — the fall-2025 industry dip was Safari stripping fbclid; check your Safari share before blaming your stack. (4) An actual break — a checkout update dropping parameters, a plugin regression, consent misconfiguration: the Events Manager key-coverage list pinpoints which key vanished, and the fix is that key's plumbing. Note what's NOT on the list: panic re-platforming, $2k/mo tracking suites bought at midnight, and grinding a 6.4 toward 7 while your creative pipeline sits empty.

Fix what's broken, nudge what's cheap, and let 6+ be done — matched volume pays; decimals don't.

Frequently asked questions