Skip to main content

Create experiments

Creating an experiment automatically generates a campaign with the same name. This campaign appears in your main campaign list and contributes to the analytics data related to your campaigns. All campaigns generated by creating an experiment are standard campaigns by default.

Create an experiment​

To create an experiment:

  1. Select the Application where you want to create your experiment.

  2. From the left-side menu, click Experiments.

  3. Click Create Experiment.

  4. In Name, type a name for the experiment.

  5. (Optional) In Description, enter some information to help you identify the experiment.

  6. In Schedule, define a start date, end date, or both:

    • Start date: Date and time when the experiment starts Running.
    • End date: Date and time when the experiment becomes Expired.

    You can modify an experiment schedule at any point.

    note

    For the most reliable results, we recommend running an experiment for a duration of at least two weeks.

  7. In Experiment type, select between two options:

    • Random variant assignment: Select this to allow Talon.One to randomly assign customers to Variant A or Variant B.

    • External variant assignment: Select this if you are using an external platform, such as your own CRM or a third-party testing tool, to decide which variant a customer is assigned to.

      note

      Pass this assignment information to Talon.One with each session update. Sessions must contain the customer profile ID to ensure they are consistently assigned to one variant.

      • In this example, we'll use the following payload to show an external variant assignment session:

        {
          "customerSession": {
            "profileId": "customer456",
            "experimentVariantAllocations": [
              {
                "experimentID": 661,
                "variantID": 2
              },
              {
                "experimentID": 3,
                "variantID": 5
              }
            ],
            "cartItems": [
              {"sku": "PROD002", "quantity": 1}
            ]
          }
        }

  8. Click Next.

  9. (Optional) In Features, select which features you want to enable, for example Coupons or Loyalty.

    note

    You can enable or disable a feature after creating an experiment.

  10. Click Create Experiment.

Create rules for the experiment​

After you've created an experiment, you can test the two variants using the same set of rules. For example, you can test a 10% discount against a $5 discount for all customers whose session total is over $50.

To open the Rule Builder, click an experiment and select Rule Builder. You can then create rules for the experiment as follows:

  1. In Conditions, set a condition that applies to both variants.
  2. In Variant split, choose from the following options:
    • In Variant name, set names for each variant. The experiment dashboard displays these names in the column header of each variant respectively.
    • The Allocation fields appear when you are using the Random variant assignment experiment type.
    • The Variant ID fields appear when you are using the External variant assignment experiment type. You can enter the ID from the CRM or third-party testing tool that references the respective variant in each session.
  3. In Effects, set effects that apply to each variant's customer sessions respectively. Do not set effects for a variant that you want to use as a control group.
  4. In the top-right corner of the page, click Save Rules.

Rules are evaluated as long as the experiment is active.

Activate an experiment​

After you create rules for an experiment, you can activate it at any time.

To activate an experiment:

  1. On the left-side menu of the experiment, click Dashboard.
  2. Click Activate Experiment.

You cannot edit an experiment's conditions, variant split, or effects in the Rule Builder after you activate it. Note that this restriction remains in place if you disable the experiment. This guarantees the validity of the data you are collecting. You can still update an active experiment's schedule, features, and other settings, in the same way as campaign settings.