How Experiment Traffic Is Assigned

Understand how Jurni assigns visitors to experiment variants, applies traffic weights, and keeps experiment results consistent over time.

When traffic enters a Jurni experiment, Jurni routes the visitor to one of the configured destinations according to the experiment’s traffic allocation. The goal is to keep the test stable enough that differences in results reflect the variants rather than inconsistent routing.

Traffic allocation

Each experiment has a traffic distribution across its variants. A simple two-variant test may use a 50/50 split, while other tests can use different weights when the experiment setup requires it.

The configured percentages describe the expected allocation over meaningful traffic volume. They do not guarantee that every small group of visitors will be split perfectly.

A 50/50 experiment can temporarily look like 47/53 or 52/48 at low traffic. Random variation is normal. Evaluate the split over a larger number of sessions before treating a small imbalance as a routing problem.

Assignment order

Jurni does not simply randomize every request. The current assignment order is:

1. Routing Rules first. If the incoming URL matches a configured Routing Rule, that rule’s variant wins.
2. Existing assignment next. If no rule matches and the visitor already has a valid assignment for the experiment, Jurni keeps that variant.
3. Traffic weights last. If there is no matching rule and no valid previous assignment, Jurni chooses a variant according to the configured weights.

Jurni currently remembers an experiment assignment for up to 30 days. A later matching Routing Rule can override that remembered assignment and persist the newly routed variant.

If you are manually testing different variants, use controlled test URLs or change the relevant parameter rather than repeatedly refreshing the same production Smart Link and expecting a new random variant.

Changing traffic allocation

If you change the allocation after the experiment has already collected traffic:

• Existing results remain attributed to the variant that received the traffic.
• The new allocation affects traffic going forward.
• Historical results are not rewritten to simulate the new split.

Large allocation changes in the middle of a test can make the overall experiment harder to interpret. Record why the change was made and consider whether the test should be analyzed in separate time periods.

External URL variants

Experiments can include supported external destinations, such as an existing Shopify page, when you want to compare a Jurni experience against the current website. Make sure tracking is configured so the external destination can report the events required for your decision metric.

Troubleshooting an uneven split

Before escalating an apparent traffic-split issue, check:

1. Total session volume for the experiment.
2. The configured traffic weights.
3. Whether visitors are repeatedly testing from the same browser.
4. Whether all variants are published and reachable.
5. Whether campaign traffic is actually entering through the intended Smart Link.

Related guides

• What are experiments?
• Create and run an A/B test
• Troubleshoot experiment traffic splitting
• Test Jurni against an existing website page
• Reading experiment results