Campaign groups (mutually exclusive tests)
A campaign group makes campaigns mutually exclusive: each visitor is given only one campaign from the group, so tests that run on the same pages or the same audience don't interfere with each other's results. You set a weight for each campaign to control how traffic is divided between them.
Available on: all platforms, for every campaign type.
How it works
- A visitor arrives on your site. Personizely picks one campaign from the group for them, at random, according to the campaigns' weights.
- Only that campaign can run for the visitor. It then assigns the visitor to control or a variation as usual, following its own pages, target and traffic allocation. See How visitors are assigned to variations.
- Other campaigns in the group don't run for that visitor, even on pages where they would apply.
- The visitor stays with their campaign. Once the visitor has entered their campaign (it applied on a page they viewed), they keep getting it on later visits. If they haven't reached a page where it applies yet, the pick may be made again in a later session.
- When their campaign ends, the Ended test behavior setting decides whether the visitor gets another campaign from the group or none.
Without a group, each campaign assigns visitors independently, so one visitor can be in several tests at once.
Set up a group
- Open a campaign and click Group at the top of the campaign editor.
- Choose Create new group, or Add to group to join an existing one.
- Name the group, add campaigns with Add campaign to group, and set each campaign's weight.
- Choose the Ended test behavior, then Save.
From a grouped campaign, the same menu offers Edit group, Change group and Remove from group.
Option | What it does | Values / default |
|---|---|---|
Name | Identifies the group in the campaign editor | Required |
Campaigns | The campaigns that exclude each other | At least 2. A campaign can be in one group at a time |
Weight (per campaign) | Share of the group's visitors picked for that campaign | 0–100%, totaling 100%. Changing one adjusts the others |
Ended test behavior | Exclude ended campaigns: if a visitor's campaign is ended or paused, they don't get any campaign from the group again. Include replacement campaign: if a visitor's campaign has ended, they're given another active campaign from the group | Exclude ended campaigns when you create the group from a campaign |
Behavior details and limitations
- Weights are redistributed when campaigns stop. If some campaigns in the group aren't active, their weights are redistributed proportionally among the active ones, so the total stays 100%. For example, with weights 50/30/20 and the 20% campaign paused, the others get 62.5% and 37.5%.
- Scheduled campaigns wait their turn. A campaign that's scheduled but hasn't started yet is treated as inactive: it takes no share of the group's traffic until it starts.
- Picked visitors see nothing from the group on other pages. A visitor picked for campaign A doesn't see campaign B, even on pages where A doesn't apply. That's the point of a group, but it means grouping campaigns that target different pages lowers each campaign's traffic. The app warns you when grouped campaigns target different pages.
- Keep traffic allocation at 100% for grouped campaigns. A visitor picked for a campaign whose traffic allocation leaves them out still belongs to that campaign and sees nothing from the group. Use group weights, not allocation, to control traffic.
- A single active campaign gets all traffic. If only one campaign in the group is active, every visitor gets it. The app warns you so you can add more.
- Exclude vs include. Exclude ended campaigns keeps a visitor's history clean: once they've been in a test, they never see a different test from the same group. Include replacement campaign gets more traffic into later tests but exposes those visitors to two tests in sequence.
- Pausing in an Exclude group. While a campaign is paused, its visitors see nothing from the group. When it's resumed, they get it again.
- Deploying removes the campaign from the group. When you Scheduling, stopping and deploying winners a grouped campaign, it leaves the group and is deployed on its own. If that would leave only one campaign, the group is deleted.
- Cloning doesn't copy the group. A cloned campaign starts outside any group.
Examples
- Two product page tests at once: a test of the image gallery and a test of the add-to-cart area, both on the product page. Grouped 50/50, so each visitor sees only one and the results don't mix.
- Homepage sequence: two homepage tests in one group with Include replacement campaign. When the first test ends, its visitors are given the second one, which then receives all of the group's traffic.
- Price test isolation: a price test and a free-shipping-threshold test grouped together, so no shopper sees both a new price and a new shipping offer.
Troubleshooting
- A grouped campaign gets far fewer views than its weight suggests → the campaigns target different pages, or the campaign's traffic allocation is below 100%. Visitors picked for it only count when they reach its pages and enter its allocation.
- A newly added campaign fills slowly → visitors who already entered another campaign in the group stay with it. The new campaign gets its share of visitors who haven't been placed yet, mostly new visitors.
- Visitors see nothing from the group after a campaign ended → Exclude ended campaigns is selected. Switch to Include replacement campaign if they should get another test.
Related
- Campaigns