Using Universal assignment method in a web experiment
The Universal assignment method enables you to use the same assignment for a single guest across multiple classic A/B experiments. This method is ideal for experiments requiring precision and more control over guest assignment, such as when running parallel tests on the same audience or creating exclusivity between experiments. It ensures each guest receives the same assignment in every classic A/B experiment that utilizes this method. Alternatively, if you don't need this level of control and precision, you can use the Unique assignment method, which automates guest assignment and is the default option.
An experiment that runs with optimized testing prevents you from selecting the Unique and Universal assignment methods or adjusting how you allocate traffic. This option will automatically allocate 100% of the traffic to your experiment, and dynamically assign guests to the highest performing variants.
Using buckets to assign guests
The core of the Universal assignment method is the use of buckets for assigning guests to experiment variants. In Sitecore Personalize, a bucket is a number between 1 and 120, randomly associated to a guest using their unique guest reference. With the Universal assignment method, this bucket number stays the same for the guest across all experiments.
When using this method, you need to set the traffic allocation by defining a bucket range for guests entering the experiment. Selecting all 120 buckets includes 100% of website traffic, while a smaller range, like buckets 1 to 60, means only guests whose bucket numbers are within this range can enter the experiment. The following diagram illustrates this concept, with purple buckets representing guests included in the experiment (buckets 1 to 60), and clear buckets representing those that are not included (buckets 61 to 120).
After the bucket range is allocated, you can assign traffic among the variants. This ensures that of those who enter the experiment, a certain portion of guests will be randomly assigned to the control, and the remainder will be assigned to the other variants. For example, in an experiment aimed at testing headline effectiveness, guests in buckets 1 to 40 would see the current headline (Control), and guests in buckets 41 to 60 would see the version with a shorter, emotional headline (Variant 1). The image below shows the buckets assigned to the control and those assigned to the variant.
The following sections describe two different scenarios for running simultaneous experiments using the Universal assignment method.
Running mutually exclusive experiments
In the first scenario, a website is running three mutually exclusive experiments, A, B, and C, each using the Universal assignment method. As shown in the following diagram, these experiments use non-overlapping bucket ranges for traffic allocation, which isolate the experiments from each other. Experiment A has buckets 1 to 30 for traffic allocation, Experiment B has buckets 31 to 80, and Experiment C has buckets 81 to 120. Each experiment has two variants, consisting of a control and a variant. The allocated traffic is assigned to each variant using distinct buckets. For instance, in Experiment A, the control is assigned buckets 1 to 15 and the variant is assigned buckets 16 to 30.
With the Universal assignment method, an eligible guest visiting the website is randomly assigned a bucket number, such as 65, which stays the same across all experiments. As a result, the same guest is excluded from both Experiments A and C, and assigned to the variant of Experiment B. The Universal assignment method ensures that the guest uses the same bucket number in all three experiments. This enables you to run mutually exclusive experiments, where a guest is exposed to only one variant across the different experiments.
Running parallel experiments
In the next scenario, a website is running three parallel experiments, A, B, and C, each using the Universal assignment method. As shown in the following diagram, these experiments use identical bucket ranges to allocate traffic, ensuring that each experiment runs on the same guest audience. All experiments use buckets 1 to 120 for allocation, which represent 100% of website traffic. Each experiment has two variants, consisting of a control and a variant. The allocated traffic is equally distributed between the variants, with the control always assigned buckets 1 to 60 and the variant assigned buckets 61 to 120.
With the Universal assignment method, an eligible guest visiting the website is randomly assigned a bucket number, such as 65, which stays the same across all experiments. As a result, the same guest enters all three experiments and is always assigned to the same variant in each. The Universal assignment method ensures that the guest consistently maps to the same variant across all experiments. This enables you to synchronize tests on variants that are intentionally linked in the user journey, such as in multi-page tests.
Example: Including and excluding guests in simultaneous experiments
Let's look at an example of how the Universal assignment method works for two experiments, one to test newsletter sign-ups and another to test Black Friday conversions.
The first experiment includes three variants: the original page currently in use (Control), a variant featuring a pop-up window (Variant 1), and another variant with a top bar banner (Variant 2). The second experiment includes two variants: the original Black Friday page currently in use (Control) and a variant featuring an alert bar (Variant).
The following tables give an overview of the variants in each experiment, including how two guests might be assigned.
Experiment 1: Newsletter sign-ups
|
Variant |
Feature to test |
Variant traffic assignment |
Guest assignment example |
|---|---|---|---|
|
Control |
Original page |
Buckets 1-20 |
Guest 1 (Bucket number 14) |
|
Variant 1 |
Pop-up window |
Buckets 21-40 | |
|
Variant 2 |
Top bar banner |
Buckets 41-60 |
Experiment 2: Black Friday sales
|
Variant |
Feature to test |
Variant traffic assignment |
Guest assignment example |
|---|---|---|---|
|
Control |
Original page |
Buckets 61-90 | |
|
Variant 1 |
Alert bar |
Buckets 91-120 |
Guest 2 (Bucket number 99) |
Due to universal assignment, Guests 1 and 2 are shown distinct variants across the two experiments. Guest 1, with bucket number 14, is assigned to the original page (Control) in the newsletter experiment, and excluded from the Black Friday experiment. Conversely, Guest 2, with bucket number 99, is excluded from the newsletter experiment, but is assigned to the variant with an alert bar (Variant 1) in the Black Friday experiment. This example demonstrates how you can use the Universal assignment method to include or exclude a guest from variants across mutually exclusive experiments, helping to prevent guests from being overwhelmed by multiple tests at the same time.