Integrations Overview
Integrations let A vs B send experiment assignment and conversion events to the analytics platforms and tools your team already uses. When a visitor is enrolled in an experiment and when they convert, you can have that data appear automatically in Google Analytics, Mixpanel, Amplitude, FullStory, or any other tool that supports event tracking.
Why connect third-party tools?
A vs B gives you a purpose-built results view with Bayesian statistics, health guardrails, and segment filtering. But your team may already have a central analytics platform where all your data lives, and you may want to analyze experiment results alongside user behavior data, funnels, or session recordings that live in another tool.
Integrations bridge that gap. By sending experiment events to your analytics stack, you can:
- Segment your existing analytics reports by experiment and variation.
- Combine experiment data with behavioral data in your data warehouse.
- Watch session recordings filtered by experiment variation in FullStory or Contentsquare.
- Build custom dashboards in tools your team already knows.
How auto-detection works
Most integrations in A vs B support auto-detection. When the snippet fires an experiment event, it looks for the third-party library on the page by checking the global window object:
window.gtagfor Google Analytics 4window.mixpanelfor Mixpanelwindow.analyticsfor Segmentwindow._satelliteorwindow.sfor Adobe Analyticswindow.FSfor FullStorywindow._uxafor Contentsquarewindow.heapfor Heapwindow.amplitudefor Amplitude
If the library is found, the snippet sends the event to it immediately. If the library is not found, the event is silently skipped — there is no error and no retry.
Where to configure integrations
All integrations are configured in Project Settings > Integrations. Find your project in the sidebar, click Settings, and then click the Integrations tab.
Each integration has:
- An enable toggle — the integration is inactive until this is turned on.
- Optional configuration fields — an account ID, measurement ID, or project token. Leave these blank to rely on auto-detection (the snippet will use the already-initialized library on the page).
What data is sent with every event
Regardless of the integration, every event forwarded by A vs B includes the following properties:
- experiment_id — unique numeric ID of the experiment
- experiment_name — human-readable experiment name
- variation_id — unique ID of the assigned variation
- variation_name — name of the assigned variation
- event_name — the type of A vs B event. The main event is
experiment_viewed, which fires when a visitor is enrolled in a variation. Conversion events may also be forwarded, depending on the integration. - revenue — the revenue amount from a conversion event, if applicable. Only present when a conversion with a revenue value is recorded.