Interval Analysis
Interval Analysis analyzes the time interval between two events. You can use it to understand how often users perform a core action, or as a complement to Funnel Analysis: besides the conversion rate between steps, you can also learn the conversion time in more detail.
With Interval Analysis, you can answer questions like these:
- On average, how long does it take users to make their first payment after registering?
- The time from opening the app to actually entering it is abnormal. Is there a problem with the installation package of a particular channel?
- How long do users need to stay in chapter 1 before they can move on to later chapters?
Interval calculation logic
Before you analyze event intervals, you need to understand how intervals are calculated.
1. If you want to analyze the interval between different events, such as from registration to first payment, the model follows the shortest-interval principle: for consecutive starting events, the interval is calculated only from the last one. Likewise, a starting event is paired only with the first ending event that follows it.
For example, if a user's behavior path is A1 -> A2 -> B1 -> B2, only one interval A2 -> B1 is counted. If the behavior path is A1 -> B1 -> A2 -> B2, two intervals are produced: A1 -> B1 and A2 -> B2.
2. If you want to analyze the interval between occurrences of the same event, such as the median interval between a user's two consecutive payment events, any two adjacent events produce an interval. That is, N events produce N-1 intervals.
For example, if a user's behavior path is A1 -> A2 -> A3 -> A4, three intervals are produced: A1 -> A2, A2 -> A3, and A3 -> A4.
Configure interval events
In Interval Analysis, you need to select the entity, the starting event, and the ending event. An interval is counted only when the same entity completes the starting event and then the ending event in order.
You also need to set the interval limit, which can be at most 180 days (180*24 hours) and at least 1 minute. After calculating the data of all intervals, Interval Analysis excludes all interval data that exceeds the limit. If you're not sure what limit to set, you can keep adjusting it while you view the interval statistics to find the best interval limit.
To get intervals that better match your criteria, you can also use Hold Property Constant to select a property for the starting event and the ending event respectively. The two properties must be of the same type. After you configure it, users must not only complete the starting event and the ending event in order, but the property values of these events must also be identical. Events whose property value is null aren't used in the calculation.
If you select a numeric property, you can also set how much the property values differ. For example, to analyze how long users stay at each level, set both the starting and ending events to passing a level, and use the associated property to require that the ending event's property value is 1 higher than the starting event's. You then get the conversion time of level 1 -> level 2 -> ..., excluding the effect of clearing the same level repeatedly.
You can also filter data or compare with group by.
View interval statistics
Interval Analysis offers two chart types. Choose one based on your analysis goal.
- Box-plot
If you want to see aggregate statistics of event intervals, such as the maximum, upper quartile, median, lower quartile, and minimum, select the box plot. If the time granularity you select isn't Total, the box plot shows how the data changes by date.
Besides the metrics above, the table also shows the number of users, the number of times, and the average. You can click the number of users who completed the interval to open the user list and view the details of these entities.
You can also use the Metrics control in the upper-right corner of the table to choose the metric types to show in the table. Here you can select the metrics shown in the box plot and the quantile metrics.
- Histogram
If you want to see the distribution of event intervals, select the histogram.
By default, the model divides the configured interval limit into 12 equal ranges, and the histogram shows the number of users or the number of intervals in each range. To adjust this, click the interval settings control in the upper-right corner of the chart.
To change the number of equal ranges, select the number you need from the drop-down list. The range is 2~30.
Besides the default equal ranges, you can use custom ranges to configure unequal ranges. First, select the unit of the ranges: days, hours, minutes, or seconds. Next, set the boundaries of each range; you can check the range of each interval in the preview area at the bottom of the window. Finally, you can use Batch add to quickly create multiple ranges, or click the history button in the lower-left corner of the window to view and reuse custom ranges you used before.
Note that custom range settings aren't affected by the interval limit. You can set ranges beyond the interval limit, but the data is affected: these ranges won't contain interval data that exceeds the interval limit. Adjusting the interval limit also doesn't clear your custom ranges
As with the box plot, you can quickly view the data of different groups in the table, or view the details of the entities in an abnormal range.