Update traffic allocation and distribution

Traffic allocation is the fraction of your total traffic to include in the experiment, specified as a percentage. By default, Optimizely Experimentation allocates 100% of visitors to experiments and distributes traffic equally among variations.

Re-allocate and redistribute traffic

To change traffic allocation and distribution for variations in Optimizely Web Experimentation and Optimizely Performance Edge:

1. Go to Experiments and select your experiment.
2. Click Traffic Allocation and adjust the values as desired.

   You can also change your traffic allocation and distribution directly from the Results page. Click Edit Experiment > Traffic Allocation.

   If you are using Optimizely Performance Edge, a change to a single variation’s traffic distribution could result in all visitors being re-bucketed. For that reason, you should not change the traffic for a live experiment in Optimizely Performance Edge.
3. In the Traffic Allocation screen, under Experiment Traffic Allocation, change the percentage of eligible visitors who enter your experiment. If you enter 50%, half of the visitors who land on your page and meet your audience conditions enter the experiment and are tracked in the results.

   Under Variation Traffic Distribution, you can adjust the percentage of traffic that is bucketed into a particular variation. For example, if you have four variations and the traffic is distributed equally, each new visitor has a 25% chance of being placed into each variation.

   Changes to traffic distribution affect only new visitors. Existing visitors (whether they were bucketed in a variation) keep seeing the same variation, even after you change traffic distribution. Visitors who are excluded from the experiment are always excluded.

   Stopping variations completely, however, can potentially affect all visitors. If, for example, you stopped a variation in an Optimizely Web experiment, returning visitors who were previously bucketed into an experiment cannot see the variation to which they had been assigned and instead get the same experience as everyone else.

   Unlike Optimizely Web Experimentation, Optimizely Performance Edge does not use sticky bucketing. If you set traffic allocation of a particular variation to zero in an Optimizely Performance Edge experiment, returning visitors who were previously bucketed into an experiment would not see the variation they were originally assigned; instead, they are re-bucketed into another variation.

4. Click Save.

If you are changing the traffic allocation from a previously published traffic allocation, Optimizely Experimentation adds a yellow Changed label indicating that this traffic allocation is not published yet.

Click Publish Experiment for the new traffic allocation to take effect.

Stop a variation

You can stop sending traffic to a specific variation during an experiment. In the Traffic Allocation view for a variation, click Stop next to the variations you would like to stop.

If you stop a variation (as opposed to setting traffic distribution to 0%), it has the following effects:

• In Optimizely Web Experimentation, any visitor bucketed into the stopped variation is not bucketed into any other variation in the same experiment. Stopping a variation removes visitors who are bucketed into that variation from the same experiment.
  In Optimizely Performance Edge, visitors who were bucketed into the stopped variation are re-bucketed into another variation in the experiment.
• New visitors do not see the stopped variation.
• Visitors who were in the variation before it was stopped are tracked in your results until they visit a page targeted by the experiment. After that point, their conversions are no longer tracked.

Stopping a variation removes visitors bucketed to that variation from the experiment cohort.

You might stop variations for a few days before archiving when a longer conversion window is expected, so you can gather the full set of results after an experiment ends.

You can also duplicate the experiment, pause the original experiment, and run the newly duplicated experiment to force previous visitors to be redistributed into a new experiment. Before running the new duplicated experiment, allocate traffic to variations as you like.