Conversions API

Also known as: CAPI

On this page 8

A Conversions API (CAPI) is a server-to-server interface through which an advertiser sends conversion events, such as purchases, leads and sign-ups, directly from its own systems to an advertising platform. In performance marketing, it complements or replaces the browser pixel so that attribution, optimization and audience building rest on data that browsers, ad blockers and consent tools do not silently discard.

The term is most closely associated with the Meta Conversions API, but comparable interfaces exist elsewhere, including the TikTok Events API, the Pinterest API for Conversions, the Snap Conversions API and Google Enhanced Conversions. They share one principle: the advertiser controls the data pipeline, decides which fields to send and remains responsible for consent and data minimization.

How a Conversions API transmits events

A Conversions API works by having the advertiser’s server, not the visitor’s browser, post an event payload to the platform’s endpoint over HTTPS. Each payload typically contains an event name, a timestamp, an action source such as website or app, custom data like order value and currency, and customer information parameters that let the platform match the event to one of its users.

Matching relies on identifiers such as email address, phone number, click identifiers and IP address. Personal identifiers are normalized and hashed with SHA-256 before sending, so the platform compares hashes rather than reading raw values. The platform then attributes the matched event to an ad exposure or click within the configured attribution window and feeds it into reporting and bidding.

The table below lists the main components of a typical event payload and what each one influences.

FieldExample contentWhat it affects
Event namePurchase, Lead, CompleteRegistrationWhich conversion action is optimized
Event timeUnix timestamp in secondsAttribution window and freshness
Event IDOrder number or unique stringDeduplication with the pixel
User dataHashed email, phone, click ID, IPMatch quality and attribution
Custom dataValue, currency, content IDsValue-based bidding and ROAS
Action sourceWebsite, app, physical storeReporting category and eligibility
Core fields in a server event payload

Why server-side tracking became necessary

Server-side tracking became necessary because browser-based measurement lost reliability after 2017. Safari’s Intelligent Tracking Prevention restricted cookie lifetimes, Firefox blocked known trackers by default, ad blockers suppressed pixel requests, and Apple’s App Tracking Transparency in iOS 14.5 cut identifier access for apps. Consent requirements under the GDPR and the ePrivacy Directive added further gaps when visitors declined tracking.

The result was systematic under-reporting of conversions, which distorts more than dashboards. Automated bidding systems learn from the conversions they observe, so missing events make campaigns look less efficient, push algorithms toward the wrong audiences and inflate apparent cost per acquisition. A Conversions API restores part of that signal for users who have consented, although it does not recover data from users who have refused.

Conversions API compared with the browser pixel

A Conversions API and a browser pixel are complementary rather than competing methods, and platforms recommend running both in a redundant setup. The pixel captures rich browser context instantly, while the server sends events that the browser cannot or should not carry, including offline conversions and events confirmed only after payment.

The following comparison summarizes the practical differences.

AspectBrowser pixelConversions API
Execution pointVisitor’s browserAdvertiser’s server or cloud function
Exposure to ad blockersHighLow, if the endpoint is not blocked
Cookie and identifier accessLimited by browser policiesFirst-party data from the advertiser
Offline and CRM eventsNot possibleSupported
Implementation effortLow, tag manager snippetModerate to high, engineering work
Data controlPlatform script collects broadlyAdvertiser selects every field
Browser pixel and server-side events

When both channels report the same action, the platform must avoid counting it twice. Deduplication works by sending an identical event name and event ID from the pixel and the server; the platform keeps one record and merges the parameters. Missing or mismatched event IDs are the most common cause of double-counted conversions and inflated ROAS.

Implementation options and their trade-offs

Advertisers can implement a Conversions API through direct integration, server-side tag management or partner connectors, and the right choice depends on engineering capacity and data complexity. Direct integration offers the most control, while managed options trade flexibility for speed.

The decision table below matches common situations with a suitable approach.

SituationRecommended routeReason
Hosted ecommerce platformNative partner integrationFast setup with maintained mappings
Several ad platforms in parallelServer-side tag managerOne data stream feeds many endpoints
Long sales cycle with CRM stagesDirect API from the CRMSends qualified leads and closed deals
Limited engineering resourcesManaged gateway serviceLittle code, recurring hosting cost
Strict data governance needsDirect custom integrationFull control over fields and logging
Choosing an implementation route

Lead-generation businesses gain particular value by sending downstream CRM events such as qualified lead or closed deal. Optimizing toward these deeper stages lets bidding favor prospects who actually buy, which usually lowers effective cost per lead for valuable leads even if raw lead volume falls.

Measuring match quality and data completeness

The quality of a Conversions API setup is judged mainly by how many events the platform can match to its users and how quickly they arrive. Meta reports an Event Match Quality score on a scale of 0 to 10, and other platforms publish similar diagnostics.

  • Match rate improves with more identifiers per event, especially hashed email and click identifiers.
  • Event latency should stay low; events sent hours late weaken real-time bidding signals.
  • Deduplication rate shows whether pixel and server events are merged correctly.
  • Parameter coverage tracks how often value, currency and content IDs are present.

A rising match score does not prove better business outcomes. Teams should validate results against an independent source of truth, such as order databases or incrementality tests, because platform-reported conversions remain modeled and attributed by the platform itself.

Common mistakes and compliance risks

Most Conversions API failures come from configuration errors and weak consent handling, not from the interface itself. The table lists frequent problems, their symptoms and the usual fix.

MistakeSymptomFix
No shared event IDConversions roughly doubledPass the same ID from pixel and server
Unhashed or unnormalized dataLow match quality, policy warningsTrim, lowercase and hash with SHA-256
Ignoring consent statusRegulatory exposure under GDPRSend events only with valid consent
Wrong timestamps or time zoneEvents rejected or misattributedUse Unix time in UTC seconds
Missing value and currencyValue-based bidding underperformsInclude order value on every purchase
Typical errors in server-side conversion tracking

Legally, server-side delivery does not bypass consent rules. Hashing is pseudonymization, not anonymization, so hashed emails remain personal data under the GDPR. Advertisers need a lawful basis, a data processing agreement with the platform and a consent management platform whose signals the server respects.

Stripped to the essentials

Online shops and other advertisers want to tell ad platforms when an ad led to a sale. Traditionally, a small script in the visitor’s browser sent that message, but browsers and privacy tools now often block it, so many sales go unreported.

A Conversions API solves this by letting the company’s own computer send the message directly. Think of it as mailing the receipt yourself instead of hoping a courier picks it up. The ad platform gets a more complete picture, so it can show ads to people more likely to buy. The company still has to respect each customer’s privacy choices.

Where it is applied

An online furniture retailer spends a large share of its budget on social media campaigns optimized for purchases. After a browser privacy update, its ads manager shows about 30 percent fewer purchases than the order database, and automated bidding starts favoring cheap accessory buyers over sofa buyers.

The team deploys a server-side tag manager that sends Purchase events with order value, currency, hashed email and the click identifier, using the order number as the shared event ID with the pixel. Events are sent only for visitors who accepted marketing cookies.

Within several weeks, the gap between reported and actual purchases narrows to under ten percent, Event Match Quality rises from 5 to 8, and value-based bidding again prioritizes high-basket customers. The result is a measurably higher ROAS on the same budget, confirmed against the retailer’s own sales records.