A/B Testing
One link, several landing pages. Attri splits the traffic for you, keeps each visitor on the same variant, and shows you which destination actually converts. No separate testing tool, no query-string hacks.
A/B testing is available on Business plans and up.
Setting up a test
Section titled “Setting up a test”Open the link builder (new or existing link) and switch the destination control from Single to A/B test. You get a variants list instead of a single URL:
- Label is for you. “Long-form page”, “Short page”, whatever makes the results readable later.
- Destination URL is where that variant sends people.
- Weight is the share of traffic. Two variants default to 50/50. Three default to 34/33/33. Change any weight and the split becomes proportional, so 70/30 and 7/3 mean the same thing.
You can run 2 to 5 variants per link. Your UTM parameters ride along on every variant, so attribution stays consistent across the test.
One thing to know: geo-targeting and A/B testing don’t combine. A link in A/B mode ignores its geo rules. Pick one or the other per link.
How visitors get assigned
Section titled “How visitors get assigned”Each new visitor is assigned a variant by weighted random pick, then stays on it. That’s the whole point. If someone lands on variant B on Monday and clicks the same link on Thursday, they see B again. Otherwise your conversion numbers would be comparing shuffled audiences.
How “stays on it” works depends on your workspace’s consent mode:
| Consent mode | Stickiness |
|---|---|
| Full | A small cookie on your short-link domain remembers the variant for a year. |
| Anonymous | No cookies, so Attri derives the pick from the visitor’s daily rotating identity. Same visitor, same variant, for the whole day. It can re-assign at midnight UTC, the same boundary anonymous mode already accepts for visitor counts. |
Bots and crawlers are filtered out of the results either way.
Reading the results
Section titled “Reading the results”Edit a link that’s running a test and open the Results tab. For the last 90 days you see, per variant:
- Clicks, and each variant’s share of the total
- Sessions attributed to that variant on your site
- Conversions, and the conversion rate
Sessions and conversions need the tracking script on the destination site. Without it you still get clicks and shares, which is enough to check the split is working but not enough to pick a winner.
Results show the moment clicks arrive. Under 50 clicks per variant you’ll see a low-traffic note, because a 3-for-4 conversion rate isn’t a result, it’s a coincidence. Attri doesn’t compute statistical significance for you yet. Look at the numbers, use judgment.
Promoting a winner
Section titled “Promoting a winner”Click Promote next to the winning variant. The link goes back to single-destination mode with that URL, and everyone, including visitors who were stuck on a losing variant, gets the winner on their next click.
Historical results aren’t deleted. Variant clicks stay in your analytics with their original labels, so you can still look back at what the test showed.
Removing a variant mid-test
Section titled “Removing a variant mid-test”Delete it from the list and save. Visitors who were assigned to it get re-bucketed across the remaining variants on their next click.