Skip to content
For the complete documentation index, see llms.txt. This page is also available as Markdown.

Traffic Allocation: Split Traffic Between Variations

How Elevate A/B Testing splits visitors between control and variations: custom percentages, changing the split mid-test, and why counts vary.


Traffic allocation decides how the visitors who enter an Elevate A/B Testing experiment are split between the control and each variation. You set it in the Traffic Allocation step when you create an experiment, and you can change it while the experiment runs.

The step works the same way on every experiment type: Page, Product, Price, Subscription, Split URL, Theme, Shipping, Content, Checkout, Post-purchase, Custom Code, and Product Image. It always comes after you set up your variations. Personalizations have no traffic allocation step, because a personalization shows its experience to every visitor who matches its audience.

How do I set the traffic split?

In the Traffic Allocation step, find the card How much traffic should each variation receive?

  • Type a percentage for a variation to change only that variation.

  • Drag the slider to move the split between the two variations on either side of the handle.

  • Percentages can use one decimal place, such as 33.3%.

  • The total must be exactly 100%. If it is not, you cannot save, and the step shows how far over or under you are.

The split starts even, such as 50/50 for two variations. It resets to even whenever you change the number of variations.

Can I split traffic 70/30?

Yes. You can give each variation any share, as long as the total is 100%.

An uneven split costs you time. The result can only be as reliable as the variation with the least traffic, so a 70/30 split takes longer to reach significance than 50/50 with the same visitors. Use an uneven split when you want to limit how many visitors see a risky variation, and accept that the experiment will run longer.

Can I set a variation to 0%?

Yes. The step shows a warning, but you can still save. New visitors are not assigned to a variation at 0%. Visitors who were already assigned to it keep seeing it, so it can still collect some data.

Can I change the split while an experiment is running?

Yes. Open the experiment and change the split in the Overview tab. You can change it at any time until the experiment is Done.

Visitors who were already assigned keep their variation. Elevate A/B Testing remembers each visitor's variation in a cookie that lasts 365 days, so only new visitors follow the new split.

Does Elevate A/B Testing move traffic to the winning variation automatically?

No. The split stays exactly as you set it for the whole experiment. Elevate A/B Testing does not shift traffic toward a leading variation, and it does not stop an experiment or apply a winner on its own.

Why are my visitor counts not exactly 50/50?

Each new visitor is assigned at random, weighted by your split. A 50/50 split means each visitor has an even chance of either version, so the counts come out close to 50/50 but never exact. With few visitors, the gap can look large. It shrinks as more visitors arrive.

Some visitors are never counted in the split:

  • Bots and search engine crawlers

  • Shopify theme previews and Elevate A/B Testing previews

  • Visitors who do not match the experiment's audience targeting

How is traffic allocation different from traffic isolation and audience targeting?

Three settings decide who sees what, and they apply in this order.

Setting
Question it answers
Where you set it

Which visitors are eligible?

The Audiences step

What share of eligible visitors enters only this experiment?

Isolate traffic for this experiment (optional), in the Traffic Allocation step

Traffic allocation

How are the visitors who enter split between control and variations?

How much traffic should each variation receive?, in the Traffic Allocation step

Isolation decides whether a visitor is in the experiment. Allocation decides which version they see.

Next steps