Skip to main content
FHAVI

Free resource

If the event is not logged, it did not happen

Server-side tracking fails silently by design: the request goes out, the platform answers, and nobody reads the answer. One order that never reached the platform looks identical to one that did, until the month closes and the numbers disagree. This is the log that closes that gap.

Fields

One row per event attempt

01

Record

  1. Timestamp, in one timezone, at send — not at queue time.
  2. Event name and the event id you generated. This is the field that makes deduplication provable.
  3. Order or cart reference, so a row can be joined back to the store record.
  4. Value and currency, exactly as sent — the two fields most often wrong on a purchase.
  5. Identifier set present: hashed email, phone, external id, fbp, fbc, client IP, user agent.
  6. HTTP status and the platform's response body, truncated but kept. A 200 with an error body is a failure.
  7. Retry count and why it retried — 429 and 5xx only; retrying a 4xx adds noise and can look like abuse.
  8. Deduplicated outcome: did the platform merge this with the browser event, or count a second purchase.

Reading

The four questions this log answers

02

When the numbers disagree

  1. Did we send it at all? A missing row means the trigger never fired — the problem is in your checkout, not in Meta.
  2. Did they accept it? A 4xx with a validation message names the parameter that is wrong.
  3. Did they count it twice? Two rows with different event ids for one order is your duplicate purchase.
  4. Did the value match? A purchase logged at zero is usually a currency or value-mapping failure, and it quietly wrecks optimisation.

Practice

Keeping the log useful

03

Rules that keep it readable

  1. Log successes as well as failures; a log of errors alone cannot prove an event was sent.
  2. Rotate monthly. Thirty days is enough to diagnose and short enough to read.
  3. Alert on the absence of rows, not on error count — silence is the failure mode you care about.
  4. Never log raw personal data. Hashed identifiers, ids and status codes are enough for every diagnostic here.
  5. Review it weekly alongside the tracking health report, and treat a gap in rows as an incident.

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

Not sure events are reaching Meta?

Send a week of your server logs, redacted. We will tell you whether purchases are failing validation, retrying, or never being sent at all.

Prefer email? Write to md@fhavi.com