What rolled-out variants bring in, added up metric by metric and corrected for the winner’s curse. Each gain is cross-checked during the rollout steps, then against the workspace’s holdout. Disagreements between these sources are flagged below.
Gains only add up across variants measured on the same metric, in its unit. A rollout in progress counts for its exposed share, a return to the original counts for nothing, and a rollout that its steps contradict is shown separately, outside the total.
Its announced gain (+€180,309 per month) is not part of the counted gain because the rollout steps contradict it.
A winning test overstates its effect on average, so its gain is adjusted toward what the workspace’s tests usually bring in.
A winning test overstates its effect on average, so its gain is adjusted toward what the workspace’s tests usually bring in.
A winning test overstates its effect on average, so its gain is adjusted toward what the workspace’s tests usually bring in.
A winning test overstates its effect on average, so its gain is adjusted toward what the workspace’s tests usually bring in.
The holdout group also measures running tests and 5 tracked rolled-out variants on other metrics (“Product click”, “Conversion”, “Add to cart”), so the agreement is an order of magnitude, not line-by-line proof.
The holdout group measures only the workspace’s main metrics with enough data. “Product click” does not have a full reading yet.
The holdout group also measures running tests and 5 tracked rolled-out variants on other metrics (“Revenue per visitor”, “Product click”, “Add to cart”), so the agreement is an order of magnitude, not line-by-line proof.
The holdout group measures only the workspace’s main metrics with enough data. “Add to cart” does not have a full reading yet.
| Test | Measured | Corrected | Gain per month | During rollout |
|---|---|---|---|---|
| Free shipping from €60B · Free shipping from €60 · “Revenue per visitor” · since Oct 5, 2026 · 50% exposed | +5.8% | +4.8% | +€44,126+13,668 to +€74,584 | Consistent, not proven yetsteps +4.2% · 31,130 visitors |
| Free returns highlightedB · Free returns highlighted · “Revenue per visitor” · since Sep 14, 2026 | +3.4% | +3.2% | +€160,945+39,195 to +€282,694 | Consistent, not proven yetsteps +2.8% · 258,840 visitors |
| Simplified filters on mobileB · Simplified filters on mobile · “Product click” · since Aug 17, 2026 | +4.0% | +3.7% | +704 converted visitors+251 to +1,158 converted visitors | Consistent, not proven yetsteps +3.6% · 16,340 visitors |
| Shipping costs shown on the product pageB · Shipping costs shown on the product page · “Revenue per visitor” · since Jul 13, 2026 | +2.1% | +2.1% | +€180,309+23,167 to +€337,452excluded from the total | Contradictedsteps -4.5% · 484,690 visitors |
| Guest checkoutB · Guest checkout · “Conversion” · since Jun 22, 2026 | +7.3% | +6.4% | +567 converted visitors+346 to +789 converted visitors | Confirmedsteps +6.1% · 8,020 visitors |
| Promo code field collapsedB · Promo code field collapsed · “Conversion” · since Apr 9, 2026 | +1.8% | +1.8% | +1,306 converted visitors+194 to +2,417 converted visitors | Rolled out outside Probant |
| Simplified order summaryB · Simplified order summary · “Conversion” · since Mar 16, 2026 | +4.1% | +3.8% | +702 converted visitors+257 to +1,147 converted visitors | Consistent, not proven yetsteps +3.2% · 47,340 visitors |
| Sticky buy button on mobileB · Sticky buy button on mobile · “Add to cart” · since Jan 20, 2026 | +2.5% | +2.4% | +986 converted visitors+146 to +1,826 converted visitors | Rolled out outside Probant |
Each gain starts from the difference measured by its test, pulled toward what the workspace’s finished tests usually bring in (winner’s curse), then scaled to a month of traffic in its zone. During a rollout’s steps, some visitors still see the original, so the difference is measured again on other visitors. The holdout (5% of visitors, who see no tests and no rollouts) measures what all other visitors gain together.
Report fingerprint 98a1d96e…8cb8f6: SHA‑256 of its numbers. An export or shared link with the same numbers carries the same fingerprint.