Skip to main content
FHAVI

Free resource

Event Match Quality, tracked instead of glanced at

Meta scores how well each event can be tied to a person. The score is a diagnostic, not a KPI — but a score that falls after a change tells you something broke, and a score that never moves tells you nothing is being improved. This is the table that makes the difference visible.

What it measures

Match quality is about identifiers, not spend

The score reflects which customer information arrived with the event: email, phone, name, city, country, external id, browser and device signals. Server events can carry hashed email and phone; browser events can carry the cookie. Neither alone is as strong as both together on the same event id.

Two things follow. First, a low score is usually a missing-parameter problem, fixable in code. Second, raising the score without raising event volume does nothing for optimisation — matching is a prerequisite, not a result.

01

The four inputs worth logging per event

  1. Score, and the date it was read — a single number in a slide is not a trend.
  2. Parameters present, not 'good or bad': email, phone, external id, fbp, fbc, client IP and user agent.
  3. Browser and server volume for the same event, so deduplication is visible rather than assumed.
  4. The last change made, with a date. An unlabelled score series cannot be attributed to anything.

The tracker

One row per event per week

02

Columns

  1. Week ending.
  2. Event name — PageView, ViewContent, AddToCart, InitiateCheckout, Purchase. Track the ones that drive optimisation; a full export is noise.
  3. Match score.
  4. Browser count and server count separately.
  5. Deduplicated count — the number the platform actually used. If this does not equal one of the two columns, check event_id pairing.
  6. Parameters present — a short code list, e.g. em/ph/ext/fbp/fbc.
  7. Changed this week — free text, one line.
  8. Verdict — improved, unchanged, degraded, broken. Degraded and broken get an owner and a date.

Reading it

What good looks like

03

Reasonable expectations

  1. Purchase should score highest — it is the event where you have the most identifiers, because you just took an order.
  2. Upper-funnel events will score lower. PageView from an anonymous visitor cannot be matched well, and chasing that is wasted effort.
  3. A score that drops the week after a theme or checkout update points at the update, not at Meta.
  4. Two events with identical parameters but different scores usually differ in event volume and reporting lag, not in data quality.
04

What not to do

  1. Do not send parameters you cannot lawfully hold, and do not send placeholder values — a hashed empty string is worse than absence.
  2. Do not judge a change within 24 hours; scores update on a lag.
  3. Do not treat a score improvement as proof of better attribution. Confirm it against the order count in your store.

Keep going

The rest of the library

Every checklist here is free and ungated. The research pages carry the reasoning behind them.

All resources · Research library

Scores stuck in the middle?

Send the current parameters for one purchase event. We will tell you which identifiers are missing and whether they are worth the implementation.

Prefer email? Write to md@fhavi.com