Integrations Overview
Integrations send your experiment data to the analytics tools your team already uses. When a visitor is shown a variation, and when they convert, that can appear automatically in Google Analytics 4, Google Tag Manager, Mixpanel, Segment, Amplitude, Heap, Fullstory, Contentsquare, and Adobe Analytics. Anything else is covered by the event bus.
Why connect your analytics tools
A vs B gives you a purpose-built results view with proper statistics, health checks, and segment filtering. That is where you decide whether a variation won. But your team probably has a central analytics platform where everything else lives, and there are questions only that platform can answer.
Connecting the two lets you:
- Break your existing reports down by experiment and variation.
- Watch session recordings filtered by the variation someone saw.
- Combine experiment data with behaviour data in your warehouse.
- Build dashboards in the tool your team already knows.
How it works
We talk to the library your page already has. We never load an analytics tool for you, and we never inject a third-party script. That means there is no measurement ID, project token, write key, or app ID to find: if the tool is on the page, switching the card on is the whole setup.
We look for the tool at the moment we have something to send. Analytics libraries load at different speeds, so if a tool is not there yet we hold the events and keep checking for up to 30 seconds. As soon as the library appears, everything we are still holding is sent. We hold up to 20 events at a time, and if more pile up while we wait, the oldest one makes way for the newest. If the library never appears, the held events are dropped rather than queued forever, and nothing on your page breaks.
If events are not reaching a tool, the usual cause is that its library is not on that page at all, or loads more than 30 seconds in. Most analytics snippets load early, so this is rare, but it is the first thing to check.
Everything starts off. No tool receives anything until you switch its card on, and no experiment data leaves your page for a tool you have not enabled.
Changes reach your site within about a minute. Save on the Integrations tab, wait a minute, and your live site picks the change up.
What is sent, and when
Experiment views are always sent to a tool that is switched on. That is the point of switching it on.
Everything else is opt in, per tool, under Also send:
- Conversions: someone completed one of the goals you are measuring.
- Purchases, with revenue: the order total, currency, and order ID.
- Recommendation views and clicks: which recipe was shown, which items, and which one was clicked.
Two rules apply to everything:
- Your own preview and internal traffic is never forwarded. Sessions you open in preview, and traffic you have tagged as internal, stay out of your analytics tools completely.
- Any single experiment can opt out. On the experiment's review step there is an Analytics forwarding switch. Turn it off and that one experiment stops being forwarded, while every other experiment carries on.
If a visitor is in three live experiments at once and they convert, each tool receives three conversion events, one per experiment. That is what lets the same action be measured against each experiment separately. It also means these events must not be added up as if they were separate conversions. See why counts differ.
Event names you can change
Each card has a box for every event name, and each box starts empty with that tool's standard name showing inside it as a hint. Leave the box empty and we send that standard name, which is chosen so the tool's own built-in experiment reports work with no extra setup. Type your own name to override it.
The card warns you where a rename costs something real, such as Amplitude not charging for its own exposure event. Each provider page lists the standard names for that tool.
Every event carries an ID
Every event we forward carries a unique event ID, so you can line individual events up between your analytics tool and A vs B rather than comparing totals and hoping.
An experiment view is recorded once and forwarded under that same ID, so the row in your analytics tool and the row in your A vs B data are provably the same event. Conversions and purchases are sent once per live experiment, so each of those copies gets its own ID. Why counts differ covers what that means when you compare numbers.
Checking your setup
The Integrations tab has a Send a test event link. It opens your site in a new tab and sends one clearly labelled test event to every tool you have switched on, so you can watch it arrive in that tool's live or debug view. Contentsquare is the one exception: it only accepts experiment variables, so no test event is sent to it.
Test events never create an experiment view and never appear in your A vs B results.
Each card also previews the exact call we will make, under See exactly what will be sent.
Where to configure everything
All of this lives in Project Settings > Integrations. Find your project in the sidebar, click Settings, then open the Integrations tab.
Analysis guides
The tools with a card
Everything else in this section
The Integrations section also covers installing the SDK in your framework, and installing the snippet plus recording purchase revenue on the major commerce platforms.