What you'll configure
- Add variants with weighted splits (e.g. 50/50 or 70/30) to a single short link — the original destination becomes the Control.
- Track click counts and conversion rates per variant; a two-proportion z-test flags 95% significance automatically.
- Promote the winning variant when you're ready, archiving the rest without losing analytics history.
Variants split traffic from a single short URL across multiple destinations. Use them for landing-page A/B tests, store-page experiments, or any time you want to compare two destinations head-to-head without standing up a separate experimentation tool.
How a variant works#
When a visitor hits a short link with variants attached:
- The edge picks a variant based on weight (e.g. 50/50, 70/30, or any split that adds to 100).
- The pick is sticky per visitor: we set a long-lived cookie keyed on the link slug so the same person sees the same destination on every subsequent click.
- The redirect happens. The chosen variant is logged with the click event.
Pick-stickiness uses an HTTP-only cookie scoped to your shortener domain. Visitors clearing cookies between clicks will be re-bucketed; for most experiments this is acceptable noise.
Create variants#
- Open a link's detail page → Variants.
- Click Add variant. Set a label (e.g. "Hero v2") and a destination URL.
- Adjust the weight slider. Weights are integers that must total 100 across all active variants — the UI snaps them for you.
- (Optional) Add a hypothesis ("v2 will convert 10% better than v1 because the CTA is bigger") — purely for your records but useful when you come back in two weeks and can't remember what you tested.
- Save. The new split is live within seconds across all regions.
The link's original destination automatically becomes one of the variants — it's labelled Control and gets the leftover weight.
Measuring results#
The variants panel shows live counts: clicks per variant, plus the conversion rate if you've wired conversion tracking. Statistical significance is calculated using a two-proportion z-test against the control; a green check appears once the test reaches 95% confidence.
The dashboard does NOT auto-stop tests when significance is reached — you decide when to call it.
Promoting a winner#
Once you've decided, click Promote next to the winning variant. The promoted variant becomes the link's new primary destination, the other variants are archived, and the cookie is cleared on next visit. Visitors who were in the losing bucket are now routed to the winner.
If you want to keep the losing variant around for a future test, click Archive instead — it stays in the variants list but with weight 0, so it gets no traffic.
Limits#
- 5 active variants per link (Pro / Business).
- Conversion tracking is required if you want significance calculations; without it you only see click splits, not outcomes.
Variants vs smart-link rules#
Variants are random; smart-link rules are deterministic. If you want to send German visitors to a German page, use a rule. If you want to randomly split all visitors 50/50 between two German pages, use variants. They compose: rules pick a destination bucket, variants then split within that bucket.
Troubleshooting#
Click counts don't sum to total clicks. Bots and bad UA detection are filtered out of the variant counts but counted in total. The discrepancy is usually 1–3%.
Same visitor reports a different variant on different days. They likely cleared cookies, or they switched browser / device. Cookie stickiness is per-browser, not per-user-account.
Significance bar stuck at 60% even with thousands of visits. The two variants are too close — there's no real winner. Consider trying a bolder change rather than fiddling with weights.