Skip to main content
cancel
Showing results for 
Search instead for 
Did you mean: 
jcs_2015
Contributor III
Contributor III

Comparing Cohort datasets across a standard manufactured time dimension

I've created a cohort report that tracks performance of each cohort population across calendar months (DIM).  My Selected Cohort time frame is based on a dynamic range selected by the user (2 manufactured DIM pick lists), limited to a Month-Year combination.  In this example, Jan-2015-Mar-2015.  Using variables, I then calculate the previous and future cohorts populations based on start/end of the selected cohort and number of months in that selected range.  In this example, Cohort-1 is manufactured to be Oct-2014 thru Dec-2014 and Cohort+1 is Apr-2015 thru Jun-2015.

As seen below, Cohort-1 has 26 time based metrics, Cohort has 23, and Cohort+1 has 20.

I'd like to time shift the graphic to a generic time definition of Month1, Month2, Month3 etc for each cohort.  If I can make this happen, the user can then see their performance by cohort population against all 3 cohorts without the need to visually time-shift each line to the left. 


The desired view is the snapshot below but left justified across a generic Month1, Month2, Month3... axis and a value for all 3 cohorts in the first manufactured dim of Month1 thru Month20, then it would be 2 populations for another 3 months, and lastly a single population for the remaining 3 months (Month26).

Thoughts on how I might make this happen would be greatly appreciated


2017-01-03_14-12-51.png

My Set Expression for Cohort:

// Count of completed metrics

Count(

{<

Eligibility = {'E'},

[Met Status] = {'M'},

[Report Date] = {"$(='>=' & Date($(vCS),'YYYY-MM'))"},

[Assessment Date] = {">=$(=$(vCS)) <=$(=$(vCE))"}

>}

MET_CNT)

/

// Count of total possible metrics to be completed

Count(

{<

Eligibility = {'E'},

[Report Date] = {"$(='>=' & Date($(vCS),'YYYY-MM'))"},

[Assessment Date] = {">=$(=$(vCS)) <=$(=$(vCE))"}

>}

CNT)

//$(vCS) Cohort Start variable based on picklist value chosen by user

//$(vCE) Cohort End variable based on picklist value chosen by user

//[Report Date] is current DIM used to render x-axis

//[Assessment Date] is used to ID the cohort population in conjunction with the selected cohort range

Attaching limited data App

1 Solution

Accepted Solutions
sunny_talwar

Something like this?

Capture.PNG

View solution in original post

8 Replies
sunny_talwar

So you won't be using Month as your dimension in the case where you want to shift to generic time, right? I am thinking of using ValueLoop() function to create a dimension and some create a measure based on that. I am not 100% sure how this can work, but if you can provide a sample, we can play around with it to see if we can make it work.

Best,

Sunny

jcs_2015
Contributor III
Contributor III
Author

Yes Sunny, I will not/can not use native Month/Year values as a visual time dimension given the lag in each cohort population.

There will never be more than 27 distinct Month/Year periods in the data model.  But the span of the cohort range could be anything from a single month to 27 months in theory.  Likely cohort populations will be quarterly however.

sunny_talwar

Something like this?

Capture.PNG

jcs_2015
Contributor III
Contributor III
Author

That looks exactly like what I was hoping to deliver!!!

What a simple solution as well, a calculated DIM using Rank.  You ROCK!!!

sunny_talwar

Not sure if you saw the expression or not, but expressions are sort of messy

jcs_2015
Contributor III
Contributor III
Author

As a result of what they are doing or just my syntax?

sunny_talwar

Well, I changed them for the second and third expression.... So because of what you needed, I had to make them quite messy. Original expressions were okay

jcs_2015
Contributor III
Contributor III
Author

Sorry, Only looked at the 1st expression.  I'll dive in deeper as I migrate this to my live App. 

Given what it is trying to do, define the laws of time, I'd suspect "messy" should be expected anyway

Regardless, Fantastic solution to a problem many who are challenged with Cohort Reporting, intertwined with Analytics, will likely be benefited.