Signal measures and evaluates the association between a correlation event and a goal event and quantifies the correlation between the two. The following article describes how to interpret the results of a report in Signal.
View this article to learn how to build a query in Signal.
Each correlation will be represented by one card in the summary view results. The list-view will display the optimal action, correlation with the goal event, and key findings about the event. Results can be sorted by correlation strength or by Mixpanel opportunity score.
By default, the results will be sorted by “Opportunity.” “Opportunity” is Mixpanel’s proprietary calculation of how important a given correlation might be. Reference this article for more detail on how "Opportunity" is calculated.
The correlation strength is calculated using the phi coefficient. To summarize, the correlation strength is how closely related two events are. Refer to this article to see how Mixpanel calculates phi.
The key findings presented are rarity analysis results and conversion measurements. Rarity defines how common or uncommon it is for users to complete an individual event. The conversion measurements qualifies how likely a given conversion is to be helpful.
Rarity Analysis can be broken down into:
- Rare in user group.
- Never performed in user group.
- Uncommon in user group.
- Common in user group.
- Majority of user group performed.
Conversion Measures can be broken down into:
- No users converted to goal.
- Unlikely to be useful.
- Associated almost perfectly with not converting.
- Associated almost perfectly with converting.
To access the detailed view, click on a card in the summary view.
The detailed view provides the correlation calculations for each event evaluated. This view uses a heat map to show the strength of the correlation with the goal event.
The x-axis represents the velocity, or the number of days it took users to complete the event (up to 15 days) and the y-axis shows the frequency of the event, or the number of times the event was done, up to ten times.
The optimal action will be highlighted in green on the heat map.
The optimal action tells you how many times and in how many days your customers should do the event in order to achieve your optimal correlation. Frequency labels are listed below the action and provide precise information, based on the correlation, about when your users should complete the event. For example, if you are a music streaming app and wanted to see how a “Play Song” event correlates with two week retention, you might see something like, “Play song once within ten days” as an optimal action.
For each combination Mixpanel will provide supporting statistics for further analysis:
Precision: Percent of users who convert to your goal among those who did this event at least x times within y days
False omission rate: Percent of users who convert to your goal among those who did this event fewer than x times within y days
Recall: Percent of converted users who did this event at least x times within y days
Fall-out: Percent of unconverted users who did this event at least x times within y days
Correlation: Association between performing this event at least x times in y days and converting to your goal.
To see the calculations of the optimal correlation data, hover of the ƒx in the upper right-hand corner of the chart at the bottom of the heat map.