Traffic allocation

Send more visitors to the version that wins

Run a fixed split, or let epsilon-greedy or Thompson sampling move new visitors toward the leading group while the test runs.

Share of new visitors per group

Thompson sampling

Day 1Day 6

Three ways to split traffic

Choose how traffic is shared

Set each group's share, up to eight groups, and keep it for the whole test.

TraditionallyA spreadsheet of splits

With Growth Dept CROA number under each ring

Most new visitors go to the leader, a set share keeps exploring.

TraditionallyWaiting weeks to act

With Growth Dept CROTraffic follows the leader

Shares follow each group's chance of being best, updated by a job and published to the storefront.

TraditionallyManual reweighting

With Growth Dept CROWeights that update themselves

Sticky

Weights change only who new visitors get. A visitor who already has a group keeps it.

Exclusion groups

Put tests in an exclusion group so a visitor enters only one of them.

FAQ

Does an adaptive split change the statistics?

Results account for the weights a test used. The engine's golden tests include a Thompson-sampling test whose weights moved.

Book a 45-minute call with the Growth Dept team. We look at your store and your traffic, and tell you which tests to run first.