Google Analytics (GA4) Integration

Send Featureflow flag exposures to Google Analytics 4 as a custom event plus a user property per flag, so you can build audiences and compare conversions by variant in GA4's own reports and explorations. See A/B testing with your analytics tool for the whole loop.

Setup

GA4 has no built-in experiment event, so the recipe does two things: a flag_exposure event for the moment of exposure, and a user property so every later event and session from that user carries the variant.

import Featureflow, { exposureIntegration } from 'featureflow-client';

// gtag('config', 'G-XXXXXXX') is already on the page

const featureflow = await Featureflow.init('your-js-client-key', {
  integrations: [
    exposureIntegration(({ key, variant }) => {
      gtag('event', 'flag_exposure', { flag_key: key, variant });
      gtag('set', 'user_properties', { [`ff_${key}`]: variant });
    })
  ]
});

Exposures go through your gtag on the page, so they carry your GA4 client id and user id — no user-id mapping needed — and are deduplicated per user, flag and variant for the page's lifetime, so calling evaluate() on every render will not inflate your GA4 event volume.

Send only the flags you're experimenting with

GA4 limits each property to 25 custom user properties, and user property names are limited to 24 characters (values to 36). Flag keys often exceed that, so limit exposures to the flags you analyse and give them short names:

const experiments = { 'checkout-v2': 'ff_checkout', 'pricing-test': 'ff_pricing' };

exposureIntegration(({ key, variant }) => {
  gtag('event', 'flag_exposure', { flag_key: key, variant });
  gtag('set', 'user_properties', { [experiments[key]]: variant });
}, { flags: Object.keys(experiments) })

In GA4

1. Admin → Custom definitions → Create custom dimension with scope User and the user property name you send (for example ff_checkout). Do the same with scope Event for variant if you want the exposure event itself in reports.
2. Build a comparison or an audience on the user-scoped dimension — ff_checkout = b versus ff_checkout = a — and apply it to any conversion report or exploration.
3. GA4 does not compute significance; export to BigQuery or use your own test if you need a p-value.

Notes

• User properties set with gtag('set', 'user_properties', …) apply to events sent after the call on that page, and persist for the user in GA4's reporting.
• Available from featureflow-client 2.6.0 / react-featureflow-client 2.6.0. On older versions, listen to the raw EVALUATION event and dedupe yourself.
• Server-side, the Node SDK emits a raw evaluation event you can forward to the GA4 Measurement Protocol with your own client id — see the Node SDK README.

---

Any problems? Reach us via the in-app chat or at support@featureflow.io.