Exclusion Groups

Exclusion groups divide your traffic into non-overlapping slices so no visitor is in two experiments within the same group. This prevents experiments from contaminating each other's results when they run on the same pages.

When to use exclusion groups

Use exclusion groups when two or more experiments target the same pages and you need to ensure their results are independent. Without exclusion groups, a visitor could be enrolled in both experiments simultaneously. If one experiment increases conversions and the other decreases them, the measured results for each experiment will be distorted.

Common scenarios:

Multiple experiments on the homepage (hero banner test + navigation test)
Several experiments on the checkout flow (CTA color + form layout + pricing)
Any situation where you want to guarantee that each visitor only sees one experiment

When NOT to use exclusion groups

You do not need exclusion groups when experiments target completely different pages. If experiment A only runs on the homepage and experiment B only runs on the pricing page, there is no overlap and no risk of contamination.

Experiments that are not assigned to any exclusion group continue to work exactly as they always have. Exclusion groups are entirely opt-in.

How traffic slicing works

Each exclusion group represents 100% of your project's traffic. When you add experiments to a group, you assign each experiment a percentage of the group's traffic. The system automatically assigns each experiment a contiguous range of slots on a 0–9999 scale.

Every visitor gets a deterministic number (0–9999) based on their visitor ID and the group ID. This number determines which "slice" of the group they fall into. If the visitor's number falls in an experiment's range, they are eligible for that experiment. If not, they are excluded entirely.

This check happens before any other targeting (audiences, URL rules, traffic allocation). The exclusion group is the first gate after basic scheduling and targeting rules.

Creating an exclusion group

Go to the Audiences page

From your project, click Audiences in the left navigation.

Switch to the Exclusion Groups tab

Click the Exclusion Groups tab at the top of the page.

Click New Group

Click New Group. Enter a name (e.g., "Homepage Tests") and an optional description.

Save the group

Click Save. The group is created with no experiments and 100% of traffic unallocated.

Adding experiments to a group

Open the group detail page

Click the group name in the list to open the detail page. You will see a traffic map showing allocated and unallocated traffic.

Click Add Experiment

Click the Add Experiment button. A modal appears showing eligible experiments.

Select an experiment and set traffic

Search for and select an experiment. Enter the percentage of traffic to allocate (e.g., 30%). The modal shows the remaining capacity so you know how much space is available.

Save

Click Add Experiment. The experiment is assigned to the group and appears in the traffic map.

Only experiments in Draft, Scheduled, or Paused status can be added to a group. Running experiments cannot be added or removed — you must pause the experiment first to make changes.

Understanding the traffic map

The traffic map is a horizontal bar on the group detail page. Each experiment is shown as a colored segment with its percentage. Unallocated traffic appears in gray. Completed or paused experiments appear in a muted style.

Below the bar, a legend lists each experiment with its assigned percentage and color.

How it works with traffic allocation

Exclusion groups and traffic allocation are multiplicative. If an experiment gets 30% of a group's traffic and has 50% traffic allocation, it effectively reaches 15% of total traffic (30% × 50%).

The group slice determines which visitors are eligible. Traffic allocation then further narrows the eligible visitors down to those who will actually enter the experiment.

Lifecycle rules

Draft, Scheduled, Paused experiments can be freely added to or removed from a group.
Running experiments cannot be added or removed. Pause the experiment first, make changes, then resume.
Completed experiments can be removed from a group (to free the traffic slot) but cannot be added to a new group.

Removing experiments and freeing traffic

When an experiment is stopped or completed, its traffic slot is not automatically freed. The slot stays reserved until you explicitly remove the experiment from the group on the detail page. This prevents accidental traffic redistribution.

To free a slot, click the remove button next to the experiment in the group detail page. The traffic becomes available for new experiments.

After removing completed experiments, new experiments will fill gaps in the traffic map first, then append to the end. If you see a fragmentation error when adding a large experiment, remove some completed experiments to create enough contiguous space.

One group per experiment

An experiment can belong to at most one exclusion group at a time. If you need to move an experiment to a different group, remove it from the current group first, then add it to the new one.

Duplicating experiments

When you duplicate an experiment that belongs to an exclusion group, the copy is not assigned to any group. It starts as an independent experiment. You must manually assign the copy to a group if desired.