Viewing Experiment statistics
The objective of an Experiment is to allow users to test whether one piece of content (Content Variant) is more successful than others in converting visitors. In the Experiment Statistics view, the performance of individual Content Variants is indicated using Conversion Rates.
About this task
- to ensure that sufficient data has been collected, each Content Variant must have at least 100 conversions;
- there must be a significant difference in Conversion Rate between the Content Variants;
- the difference in Conversion Rate between the Content Variant with the highest Conversion Rate and the next best Content Variant must exceed their combined error margins.
In the following example, Content Variant 1 has a conversion rate of 60% and an error margin of 3%, meaning there is a 95% likelihood that the true conversion rate for the variant lies between 57% and 63%. Content Variant 2 has a conversion rate of 56% and an error margin of 2% so the true conversion rate for the variant likely lies between 54% and 58%. Since the error margin ranges overlap, a winner cannot be reliably declared and so no Best Performing Variant will be displayed:
| Conversion Rate (Error Margin) | Conversion Rate Error Margin Range | |
|---|---|---|
| Content Variant 1 | 60%(±3) | 57% - 63% |
| Content Variant 2 | 56%(±2) | 54% - 58% |