Learn from your experiment

You’ve designed your experiment, rolled it out to your users, and given them enough time to interact with your new variants. Now it’s time to see if your hypothesis was correct.

In the Analysis card, you’ll be able to tell at a glance whether your experiment has yielded statistically-significant results, as well as what those results actually are. Amplitude Experiment takes the information you gave it during the design and rollout phases and plugs them in for you automatically, so there’s no repetition of effort. It breaks the results out by variant, and provides you with a convenient, detailed tabular breakdown.

Amplitude will not generate p-values or confidence intervals for experiments using binary metrics (i.e., unique conversions) until each variant has 100 users and 25 conversions. Experiments using non-binary metrics need only to reach 100 users per variant.

View results

To generate and view experimental results, follow these steps:

1. In your experiment, the Activity page includes two sections to view your results. The Summary section and the Analysis card. The Summary section will describe your experiment's hypothesis and note whether it has or has not reached statistical significance.

   An experiment is said to be statistically significant when we can confidently say that the results are highly unlikely to have occurred due to random chance. (More technically, it’s when we reject the null hypothesis.) That might sound pretty subjective, but it’s grounded solidly in statistics. Stat sig relies on a variant’s p-value, which is the probability of observing the data we see, assuming there is no difference between the variant and the control. If this probability drops below a certain threshold (statisticians refer to this threshold as the alpha), then we consider our experiment to have achieved statistical significance.

   The Summary section will display a badge labeled Significant if stat sig was met, and a badge labeled Not Significant if stat sig was not met.

   The Summary section may include multiple badges simultaneously:

   • Inconclusive: the test was inconclusive for the primary metric.
   • Above Goal or Below Goal: the primary metric's mean was either above or below its goal depending on the direction of the test (increase = above, decrease = below).
   • Above Control or Below Control: the primary metric's mean was either above or below the control's mean, depending on the direction of the test (increase = above, decrease = below). These badges are only relevant to stat sig results.

   

2. At the top of the Analysis section is an overview of how your experiment performed, broken down by metric and variant. Below that is the experiment's exposure definition: how many variants were shown, what the primary metric was, and what the exposure event was. This is the event users will have to fire before being included in an experiment.

   Note

   The exposure event is not the same thing as the assignment event. If, for example, you’re running an experiment on your pricing page, a user might be evaluated on the home page for the experiment—but if they don’t visit the pricing page, they'll never actually be exposed to it. For that reason, this user should not be considered to be part of the experiment.

   To learn more about exposure events, see this article in the Amplitude Developer Center.

   Click Chart Controls to see the chart definition.

   You can also create a chart in Amplitude Analytics from this experiment by clicking Open in Chart.

   Note

   If you are running an A/B/n test, Amplitude Experiment displays the confidence interval / p-value for the control against each treatment individually. To instead see the comparison between two non-control treatments, either change the control variant, or open the test in Analytics and create a chart using the two treatments you're interested in.

3. If desired, adjust the experiment’s confidence level. The default is 95%. You can also choose between a sequential test and a T-test. 

   Note

   Lowering your experiment’s confidence level will make it more likely that your experiment achieves statistical significance, but the trade-off is that doing so increases the likelihood of a false positive.

4. Set the time frame for your experiment analysis, either from the selection of pre-set durations, or by opening the date picker and choosing a custom date range.

   The tables, graphs, and charts shown in the Analysis section are explained in depth in the articles on understanding the Experiment Analysis view and interpreting the cumulative exposures graph in Amplitude Experiment.

   Note

   Amplitude Experiment needs something to compare your control to in order to generate results. If you neglect to include both the control and at least one variant, your chart will not display anything.

Interpret notifications

Your preferred notification settings allow you to receive experiment updates by email or Slack. 

Click the check box next to the desired notification:

• Experiment end reached: You will receive this notification when your experiment is complete.
• SRM detected: You will receive this notification if a sample ratio mismatch issue is identified.
• Long-running experiments: You will receive this notification when your long-running experiment is complete.
• Statsig for the primary metric is reached: You will receive this notification when your experiment's recommendation metric has reached stat sig.

What’s next?

It’s important to remember that no experiment is a failure. Even if you didn’t get the results you were hoping for, you can still learn something from the process—even if your test didn’t reach stat sig. Use your results as a springboard to asking hard questions about the changes you made, the outcomes you saw, what your customers expect from your product, and how you can deliver that.

In general, the next step should be deciding whether to conduct another experiment that supports your hypothesis to gather more evidence, or to go ahead and implement the variant that delivered the best results. You can also export your experiment to the Experiment Analysis in Amplitude Analytics and conduct a deeper dive there, where you can segment your users there and hopefully generate more useful insights.