There are two versions of Optimizely
. If you're using Optimizely X, check this article out instead.
What version do you have?
This is what the Optimizely Classic user interface looks like.
This is what the Optimizely X user interface looks like.
This article will help you:
- Determine what percentage of your total traffic is eligible to be in your experiment
- Determine what percentage of your experiment traffic sees each variation
- Pause a variation
- Auto-allocate traffic to the "winning" variation for an experiment
By default, Optimizely allocates 100% of visitors to experiments and distributes traffic equally among variations. Read on to learn how to control traffic allocation and distribution in Optimizely classic.
Here's a short video about how to set your traffic allocation.
To change your traffic allocation:
In Visual Editor, click Options > Traffic Allocation.
Or, from the Home page, choose an experiment to open the Experiment Details sidebar. Next to Traffic Allocation, click Edit.
Drag the slider to increase or decrease the percentage of traffic to include in the experiment.
If you choose 50%, half of the visitors who land on your page and meet your audience conditions will enter the experiment and be tracked in results.
Click each box to change the percentage of visitors who should see a given variation. The total of your percentages must always equal 100%.
Changes made to the overall experiment traffic allocation only affect new visitors. Existing visitors (whether or not they were bucketed in a variation) will continue to see the same variation, even after you change the traffic allocation. Visitors who are excluded from the test will always be excluded, even if you change the overall traffic allocation to 100%.
Changing the allocation between variations after you've started your experiment will affect your results. To preserve your data integrity, we don't recommend changing traffic allocation after you've started an experiment. When you change a variation’s traffic allocation mid-experiment, all new users will be allocated accordingly from then on. However, all users that entered your experiment before the change will be bucketed into the same variation they entered previously, altering the results and making it difficult to interpret the conversion rate. For this reason, we recommend that you do not change individual traffic allocation to a single variation.
Pause a variation
You can pause a variation during an experiment to send 0% of new traffic to the variation. In the Traffic Allocation dialog for a variation, click Pause.
Pausing a variation means new visitors will no longer be bucketed into that variation. Visitors who've already seen the variation will continue to see it for as long as the experiment is running. After the experiment is archived, no visitors (existing or new) will see it. You might pause 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.
To prevent all visitors from seeing the variation, including visitors who've seen it before, pause the experiment and duplicate it. Then, allocate traffic to variations as you like.
Traffic auto-allocation automatically adjusts your traffic allocation over time to maximize the number of conversions for your primary goal. This feature is designed for dynamic environments, where the best-performing variation is likely to change or fluctuate over a relatively short period.
To turn on traffic auto-allocation, click Options > Traffic Allocation, then choose Enable auto-allocation for this experiment.
Rather than declaring a final winner for your experiment, traffic auto-allocation continually tests which variation is best at converting visitors for your primary goal. As the experiment runs, Optimizely's algorithm continually observes the conversion rate for each variation in terms of the primary goal and the statistical significance associated with the variation.
As the algorithm becomes increasingly certain that one variation is outperforming the others, it progressively serves the "winning" variation to more and more new visitors to the experiment. If a different variation begins to perform better in terms of the primary goal, the algorithm adjusts the traffic allocation so that more visitors see the stronger variation instead.
To learn more about how these kinds of algorithms work, check out this popular statistics problem called the “multi-armed bandit."