Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Glossary term

Conversions API

Server-side event sharing with Meta.
CAPIAnalyticsplatform featureUpdated

Definition

Meta's developer documentation describes the Conversions API as a connection between an advertiser's marketing data, such as website events, app events and offline conversions, held on the advertiser's server, website platform or CRM, and the Meta systems that optimise ad targeting, reduce cost per result and measure outcomes. The pixel does the same job from inside a web page; the Conversions API does it from a system the business controls. The registry's one-line summary is the practical version: it complements the pixel and recovers signals that the browser path loses, and it needs deduplication to avoid double counting.

The word "server" is broad. It covers a shop platform's own integration, a tag manager running in a server container (GTM), a CRM that reports qualified leads, and a custom back end posting events directly. All of these send the same payload shape: an event name, a time, an action source saying where the event happened, customer information parameters, and an event identifier.

How it works

A visitor performs an action. If the pixel is present and allowed to run, the browser sends the event to Meta with the browser identifiers it holds. Independently, the business's system records the same action and sends it through the Conversions API with the customer information it holds, such as a hashed email address or phone number, plus whatever browser identifiers were captured when the visitor arrived. Meta receives two copies. Because both carry the same event name and the same event identifier, Meta's documented deduplication treats them as one event and keeps one.

The server copy survives the failures the browser copy does not: declined consent, content blockers, shortened cookie lifetimes, abandoned page loads and offline actions. That is the recovered signal. The quality of the recovery depends on matching: Meta's documentation describes customer information parameters as the data used to match an event to an account, and Events Manager reports a match quality score per event. A server event with no usable identifiers is received but cannot be attributed to anyone, so it adds nothing to optimisation or reporting.

PathWhat it seesWhere it fails
Pixel onlyBrowser actions with browser identifiersBlocked scripts, declined consent, cleared storage, offline events
Conversions API onlyServer-recorded actions with customer informationMissing browser identifiers lower match rates; browser-only actions are never recorded
Both, deduplicatedThe union of the twoMismatched event identifiers or names cause double counting

In practice

Most websites run both paths, which Meta's guidance calls a redundant setup, and the engineering effort goes into three things: generating one event identifier per action and passing it to both the pixel call and the server payload; sending the customer information the business legitimately holds, hashed as documented; and checking Events Manager after launch to confirm that deduplication is happening and that match quality is acceptable. The Conversions API guide walks through setup options and verification, and the server-side tagging basics page covers the server container route.

Illustrative example — not a real result

A lead form with and without the server path

A services company runs a lead campaign and counts a completed form as its conversion. Over a month the pixel alone reports 180 leads. The CRM, which records every submission regardless of the browser, holds 260. The company adds a Conversions API integration from the CRM, sending each submission with the hashed email the visitor typed, the browser identifiers captured on arrival and the same event identifier the pixel used. After deduplication, Events Manager shows 245 leads for the next month on similar spend: not the CRM total, because some submissions lack identifiers Meta can match, but far more than the browser alone saw. Reported cost per lead falls from about $55 to about $40 without any change to the ads, and delivery now optimises on a fuller set of outcomes.

What it is not

It is not a replacement for the pixel on a website; Meta recommends the redundant setup because the browser path carries identifiers that match well. It is not an attribution model: it changes which events Meta sees, and the attribution settings decide how those events are credited (Attribution). It is not a consent workaround: events must still be sent in line with the consent and the legal basis the business holds, and Meta documents a data processing options parameter for restricting use. And it is not a guarantee of more conversions. It produces a more complete count, which is a measurement change; whether reported cost per acquisition (CPA) falls depends on how much the browser path was missing.

The pixel is the browser-side half of the pair, and the Meta pixel setup guide is where the event identifier generation that deduplication depends on is usually implemented. Tag management (GTM) matters because a server container is one of the common ways to send Conversions API events without custom code. Attribution (Attribution) is the next step after the event arrives, deciding which ad interaction receives the credit. Cost per acquisition (CPA) is the metric most visibly affected when a more complete event stream replaces a partial one. Comparable server-side channels exist on other platforms, and the same principles of deduplication and match quality apply, but their details belong to their own documentation and to the platform guides on this site.

Common mistakes

  • Sending both copies without a shared event identifier, so every conversion is counted twice and delivery optimises on inflated numbers that look like success in the dashboard.
  • Omitting customer information parameters because they seem optional; events arrive but match nobody, and the match quality score in Events Manager shows it.
  • Treating the server path as a way to bypass consent, when the obligation attaches to the data, not the route it travels.
  • Dropping the pixel after adding the API, which usually lowers match quality rather than simplifying measurement, because the browser identifiers go with it.
  • Judging the integration by conversion volume alone instead of checking deduplication and match quality in Events Manager, where a rise in volume may simply be double counting.

More terms to look up

Where this term is used

Sources

  1. Conversions API — Meta for Developers

Related