Solved

User stopped doing VS Not done

  • 12 April 2022
  • 3 replies
  • 125 views

Userlevel 3
Badge +2

I've got a silly question. What is the difference between "User that has stopped doing" and "user who has not done"? I know semantically there is a difference. But in reports, both of them come with a time span. Ex. User who has Not done a Session in 30 Days. And then lower, there is another filter for the time span you want the report for. SO if I wanted to see Users who have not done, say, an "active session" in the past month, which one would I use, with what timespan?

icon

Best answer by Christy H 12 April 2022, 19:50

View original

3 replies

Userlevel 4
Badge +3

Thanks for the post @AnyaLiv  , and please know that this is not a silly question at all! In fact, I think this is one of the more nuanced things in Heap’s analysis capabilities. Explaining this is a bit complex but I will give it a shot.

Users who stopped doing looks at users who have previously completed the event, but have stopped doing so in the time period you select. So in your screenshot it would return users who had a session between 31-60 days ago but have not had a session in the past 30 days.

Users who have not done looks at users who did not do the event in the time range that you selected in relation to the date shown on the graph. So if you were looking at the results for today. It would show you users who have not had a session in the previous 7 days. But when you look at the datapoint for yesterday it would tell you how many users have not had a session 7 days prior to that date and it also takes into account all users it has tracked since Heap has been installed. What's tricky about both of these is that they use the relative date range of the granularity you've selected at the bottom. For example, if you do past 30 days by day as your date range, but say Number of Users -  Users who have not done <event> in past 7 days, for each day in the past 30 day date range, Heap looks at each day in the date range, and looks back 7 days before the start of each day in the date range, to determine how many users qualify for that day in the date range. Heap runs this calc for each of the 30 days in the date range.

To answer your question, if you are simply trying to get a list of users who recently became inactive you would likely want to use something like this:

 

Userlevel 5
Badge +3

@AnyaLiv Along with Christy’s answer, I think this doc will help you understand the difference between the lookback window and the chart date range.

 

The Lookback Window, highlighted above in purple, controls the timeframe over which to aggregate how many users performed [or did not perform] the desired behavior. The Date Range, highlighted above in red, controls the dates over which you want to see the Number of Users calculations. 

It is important to note that the lookback window is calculated from the start of each time bucket seen on the graph. 

 

The semantic difference between the two operators you mentioned boils down how you want to describe your cohorts’ engagement pattern. (You can read the definition of the special active usage cohorts here.)

• “Users who have stopped doing X in past 30 days” means that they stopped doing X in the past 30 days having done X in the 30 days before that. 

• “Users who have not done X in past 30 days” will have some overlap with the above, of course, but it will also include users who have never done X. 

 

To answer your question:

If I wanted to see Users who have not done, say, an "active session" in the past month, which one would I use, with what timespan

Both operators will work, but I would generally opt for the same version as Christy offered (since it implies users becoming inactive). The lookback window is past 30 days, and the date range is today. This tells you the number of users who, as of the start of today, had stopped doing sessions in the past 30 days having had sessions in the 30 days before that. 

 

Of course, if you want to look at this trend over time (e.g. over the past 90 days, is the trend towards inactivity going up or down), you’ll just expand the date range accordingly.

Userlevel 3
Badge +2

@AnyaLiv Along with Christy’s answer, I think this doc will help you understand the difference between the lookback window and the chart date range.

 

The Lookback Window, highlighted above in purple, controls the timeframe over which to aggregate how many users performed [or did not perform] the desired behavior. The Date Range, highlighted above in red, controls the dates over which you want to see the Number of Users calculations. 

It is important to note that the lookback window is calculated from the start of each time bucket seen on the graph. 

 

The semantic difference between the two operators you mentioned boils down how you want to describe your cohorts’ engagement pattern. (You can read the definition of the special active usage cohorts here.)

• “Users who have stopped doing X in past 30 days” means that they stopped doing X in the past 30 days having done X in the 30 days before that. 

• “Users who have not done X in past 30 days” will have some overlap with the above, of course, but it will also include users who have never done X. 

 

To answer your question:

If I wanted to see Users who have not done, say, an "active session" in the past month, which one would I use, with what timespan

Both operators will work, but I would generally opt for the same version as Christy offered (since it implies users becoming inactive). The lookback window is past 30 days, and the date range is today. This tells you the number of users who, as of the start of today, had stopped doing sessions in the past 30 days having had sessions in the 30 days before that. 

 

Of course, if you want to look at this trend over time (e.g. over the past 90 days, is the trend towards inactivity going up or down), you’ll just expand the date range accordingly.

Thank you so much for this addition Jonathan - the look back window explanation is exactly what I was looking for! It helps explain to other users the “time buckets” :) 

Reply