Skip to main content

Hi,

 

new user to Amplitude and love it already. I’m trying to evaluate a KPI we’ve defined:

“How many users have opened the app on at least 5 different days during the last 14 days.”

I’ve tried this and that:

  • A cohort with amount of app openings during the last 14 days
    This doesn’t show me the number of days though, only the number of events (e.g. someone could open the app 5 times a day and not at all in the rest of the 14d)
  • A stickyness chart
    This seems to show me however the amount of days per week, not in a specific time range

Am I missing something? How would I achieve such a chart / cohort?

Hi @konradm,

Thank you for posting your use case.

The closest workaround I can suggest is to “count in interval” feature in cohorts. Using our Demo data, the cohort will look like this: 

The “in any quarter” might slightly affect the result if the 5 times of event firing spans over two quarters in the last 14 days. Will this work for you? 

Best,

@Yuanyuan Zhang 

Thanks for sharing this. The “count in interval” looks like a fairly new clause in cohort definitions. Somewhat replicating the stickiness chart aspect inside the cohort, which is pretty cool.

Hey @Yuanyuan Zhang, amazing, that looks pretty much what I’m looking for. However it doesn’t seem to work: This should basically show all of our users (to my understanding) but the result is 0. What am I missing?

 

 

Hi @konradm This is an interesting finding!

I could reproduce what you are seeing in the Demo, but if I switch to in “any week” or change the date picker to “last 90 days”, I get results. As this is a quite new feature, I will need to clarify with the engineering team how the computation is done. I will update you here once I know more!

Hey @Yuanyuan Zhang, right, this works:

 

 

However it isn’t yet what I need as this checks only in one week (I suppose) as opposed to the whole last 14d

Hi @konradm, 

You are correct that the screenshot above checks only in one week, not the last 14 days. 

The engineering team has confirmed that issue with the example cohort I provided is that the “start of quarter” falls outside of the query time range (last 14 days). “count in interval” cohorts are powered by the same backend logic as stickiness charts, which ignores the most recent incomplete interval. In this example, “last 30 days” = 7/21/22 - 8/20/22 and we’re looking at quarterly interval. Since Q3 (July '22 - Sept '22) hasn’t completed yet, we don’t include users during that date range, hence there’s 0 users for 7/21 - 8/20. 

I have asked the engineering team if we have any plan for “improvement” on this feature but I am afraid that I do not have a better workaround for your use case at the moment. 

Thank you for bringing this issue to our attention! In case of further updates, I will post it here, too.

Hi @konradm, 

I just heard back from the engineering team that we currently do not have plans for changing this feature. I have submitted a feature request for having the possibility to specify >= 5 days in a 14-day period. Apologies for the limitation at the moment!

Best, 

Reply


Terms & ConditionsCookie settings