Experiment goals
When you create an experiment, you can set a goal that describes the business outcome you want to test, such as higher revenue or lower discount spend. Each goal maps to a primary metric that defines how Talon.One evaluates the results.
For example, imagine you want to lower your discount spend while maintaining your sales volume. You can test a smaller discount against your current one and set a goal to optimize discounts. The experiment results then show which variant keeps sales steady at the lower cost. You can then create a campaign from the winning variant to make the most of your discount budget.
Goal types and primary metrics
You can set or edit a goal on the Goal page any time before you activate the experiment.
To set a goal, select one of the following goal types:
| Goal type | What it tests | Primary metric |
|---|---|---|
| Maximize revenue | Which effect drives a higher order value. | Avg. Session Value |
| Optimize discounts | Which effect maintains sales with lower discounts. | Avg. Discounted Session Value |
| Maximize purchased items | Which effect drives a higher number of items per order. | Avg. Items per Session |
| Other | Any other goal. | None. Talon.One compares all metrics. |
By default, the Maximize revenue goal type is selected.
Hypothesis
The Hypothesis field is optional, and its content does not affect which variant wins. Use it to record the assumption you want to test, so you have a benchmark to measure your results against when you evaluate the experiment.
For example: A 10% discount on orders over $50 increases the average discounted session value among new customers without reducing sales volume.
Goals and experiment results
Your goal decides the winning variant shown in the experiment dashboard. When your goal sets a primary metric, Talon.One determines the outcome from that metric alone:
- If one variant reaches statistical significance on the primary metric, Talon.One declares that variant the winning variant.
- If neither variant reaches statistical significance on the primary metric, the result is inconclusive.
Talon.One does not factor the other metrics into this decision, so a drop in one of them does not make the result inconclusive. The evaluation table still shows all metrics, so you can review them before making a business decision.
When using the Other goal type, Talon.One compares all metrics to determine the winning variant, in the same way as an experiment without a goal.