FAQ: Event Segmentation

After selecting an event to analyze in a chart, you can use “group by” and "where" clauses to query event and user properties further. “Where” will filter your results for specific property values. "Group by" will group your results by your selected property. 

To see data beyond the last 30 days, use the date picker and date interval options to select different date ranges of data.

For example, if you’d like to see data for the last 90 days, use the datepicker to select Last 90 days instead of Last 30 days. If you’d like to see data for a specific date range in the past, click the Between tab in the datepicker and select your custom dates. If you’d like to see data per month rather than per day, you can select a monthly date interval instead of daily. 

Line charts return values per date interval (eg. per day, per week, per month) while bar charts return values for the entire date range. The date interval selected will not matter when using a bar chart. If a user fires an event on multiple days, the line chart counts a user for each day they fired the event, while the bar chart will only count the user once. In other words, the bar chart will deduplicate the counts. 

For example, if a user fired event A on November 2nd and November 6th, and you have a chart that counts uniques daily between November 1st and November 30th, this user will be counted twice in the line chart but only once in the bar chart. Thus the bar chart will return a count smaller than the sum of each daily data point. 

A new user is a user who has logged an event to Amplitude for the first time (this includes inactive events). The time of when the user is new is the time of when they logged their earliest event. 

In an Event Segmentation chart, when choosing New Users, Amplitude will look for the events fired by new users within each day/week/month (whichever interval was chosen) that the user was new. For example, if the interval is set to daily and a user was new on July 17, only the events that happened on July 17 will appear on the chart, regardless if the user also fired events the next day, on July 18.

More helpful definitions of user properties can be found here.

Active % graphs the percentage of users who fired a specific event, compared to the total number of active users (users who have fired any active event) at each data point.

If you have a cohort or segment filters applied to the chart, the denominator will only include users who have fired an active event and meet your segmentation requirements. 

For more information on how each metric is defined, please see this article. 

In the Measured as section of the Event Segmentation chart, select Average of Property Value from the Properties drop-down shown below to calculate the average value: 

For more information on how each metric is defined, please see this article.

Use the Formula feature to calculate custom ratios. This is found under the Advanced drop-down menu, in the Measured as module.

• Example: I'd like to see how many users downloaded a song or video compared to all the users who favorited a song or video each day. I can use the custom formula UNIQUES(A)/UNIQUES(B) to calculate this ratio/percentage. (Tip: add '%:' to the front of your formula to display results as percentages.)

Customers on Scholarship, Growth, or Enterprise plans can add a where clause to the who did segment definition. This is called an in-line cohort.

You can also define and create a cohort and include this in the segmentation instead. This is helpful if you need to reuse the definition in other charts.

• Microscope: In a chart, click on a data point you're interested in. A popup will appear; click Download users.
• Cohort: Create a cohort definition and click Export CSV.
• Create a chart: Use Any Active Event, grouped by Amplitude or user ID. Then click Export CSV in the breakdown table below the chart.

The best way to create this analysis is by using the ROLLWIN custom formula in the Event Segmentation chart. Rolling windows will help you find the count of unique users over the number of intervals (hours, days, weeks, months) you specify.

(%:Uniques(A) / ROLLWIN(Uniques, A, 30))

See our Help Center article on how custom formulas work in Amplitude Analytics for more information.

Sometimes users instrument event and user properties that have the same naming convention. When choosing the property make sure you are selecting the correct type of property. For example, 'url' could be used as a name for both an event property and user property. If you're filtering or grouping by the 'url' user property instead of the 'url' event property, you may see different results returned in the chart than expected.

Amplitude servers ingest high volumes of data, and there is the potential for some data processing delays. Rest assured that no data is lost. Incomplete data means only that data has been received by Amplitude, but the servers are taking longer to ingest and process the data for you to query.

To stay up to date, please refer to the Amplitude status page: https://status.amplitude.com/

Amplitude will only display the top 100 property values in the breakdown table. If you export the breakdown table to CSV, the file will max out at the first 10,000 values.

For more information on limits, please see here.

An important note regarding the glob match feature is that the asterisk only matches non-"/" characters. To search for strings that contain "/", you will have to use two asterisks, because "**" can match anything. In this case, since the referrer property contains URLs, there are going to be lots of "/" characters.

For more information on glob matching, please see here.

In the Segmentation Module, there is a drop-down menu where you can select accounts instead of users.

Because of the definition of the Active % metric, the denominator changes when you apply the 'group by' in the events and segmentation modules. The Active % metric is derived by:

(number of users who perform event) / (number of active users matching the segment) * 100​

When you apply 'group by' in the events module, the group by is applied only to the event. The formula applied is:

(number of users who perform event where property = AA country) / (number of active users) * 100

When you apply 'group by' in the segmentation module, the Active % metric needs to first take into account the number of active users matching the segments. Therefore, the formula applied here is:

(number of users who perform event where property = AA country) / (number of active users in AA country) * 100

You are asking Amplitude, out of the users who are active in AA country, what % of them performed the event where property = AA country?​

You can see that in both cases, the number of unique users who perform event A where property = AA country remains unchanged. What changes is the denominator, which changes the %.

This is usually because you are querying on an inactive event. The number of users who have performed this event is higher than the number of active users (i.e 110/100, so active % is 110%).

When there are multiple segments, we “split” the group by cap between each segment. This means with 2 segments, each segment only gets 5000 group-by values, hence the limiting seen in the exported CSV. This behavior is similar to how when there are multiple events, we will reduce the number of rows allotted to each event.

Currently, the maximum date range you can query in the Event Segmentation chart is 365 days for daily & weekly. If you change the interval to monthly, you can look up the past 36 months. Another suggestion will be to select the “compare to past year” feature, which will allow you to compare your data this year to the data you have last year.

Amplitude does not currently support creating goal lines or horizontal annotations. However, you can “hack” custom formulas to create horizontal annotations and set “targets” for various metrics. In the example below, the metric is multiplied by 0, then the target value of 90,000 is added. You can rename the labels and y-axis etc. for a more legible chart.