Variations & traffic split
Add variations, control how much traffic enters the test, and understand how visitors are assigned.
Variations
- Every test starts with a locked Control (your original page).
- Add as many variations as you need, each with its own name and edits.
- Switch between variations in the editor dropdown to edit each one.
Traffic split
The traffic split (1–100%) is the share of eligible visitors who enter the test at all. It is not a split between variations.
- Entry is decided by hashing a per-visitor seed together with the test ID, so the same visitor gets the same answer every time — someone who is out of the test stays out.
- Visitors who enter are then distributed evenly across your variations. With a control and two variations, that's roughly a third each. Per-variation weights are not configurable — if you want an uneven split, use a multivariate feature flag instead.
How long an assignment lasts
Once assigned, a visitor's variation is stored for the session. It survives page navigation and reloads within that session, so nobody sees the page flip mid-visit. A visitor who returns days later may be assigned differently.
This matters when you read results: Spectry reports impressions and unique visitors per variation from what was actually served, so the numbers stay correct — but don't design a test that depends on someone seeing the same variation across weeks.
Device breakpoints
Configure the tablet and mobile max-widths used to apply your device-specific overrides (defaults: tablet ≤ 1024px, mobile ≤ 768px). Match these to your site's own breakpoints, or a variation can look correct in the editor and wrong in the wild.
Split-URL tests
Instead of editing a page, a split-URL test redirects a share of visitors to entirely different URLs — useful for testing two separate page designs. The assigned variation is stored before the redirect so the conversion is still attributed correctly on the destination page.
Lower the traffic split when testing risky changes; raise it to reach a conclusion faster on low-traffic pages. A 10% split on a page with 200 visitors a week will not produce an answer this quarter — check the sample size calculator before you launch.