Help with Cohort Analysis - By Week

[jackj]

I'm wondering if anyone can help with this problem - I've been trying to solve it for days with no progress.  I have a data table similar to this:

Service Date	Week Ending	Customer ID	Service
1/1/2021	1/3/2021	123	Daycare
1/8/2021	1/10/2021	123	Boarding
1/14/2021	1/17/2021	123	Daycare
1/1/2021	1/3/2021	124	Daycare
1/15/2021	1/17/2021	124	Daycare
1/1/2021	1/3/2021	125	Boarding
1/1/2021	1/3/2021	126	Boarding
1/8/2021	1/10/2021	126	Daycare
1/13/2021	1/17/2021	126	Daycare
1/2/2021	1/3/2021	127	Boarding
1/5/2021	1/10/2021	127	Daycare
1/1/2021	1/3/2021	128	Boarding
1/6/2021	1/10/2021	128	Boarding
1/2/2021	1/3/2021	129	Daycare
1/8/2021	1/10/2021	130	Daycare
1/15/2021	1/17/2021	130	Daycare
1/1/2021	1/3/2021	131	Daycare
1/9/2031	1/10/2021	131	Daycare
1/15/2021	1/17/2021	131	Daycare

Basically, I am looking to do a cohort analysis by week, showing the total # of users who visited during the week ending 1/3/21, and how many of those users returned the next week, the following week, etc.

There are lots of helpful examples for doing this by month, but I want to create this by week.  Any suggestions?

[richbenmintz]

Hi @jackj,

Using the Data you provided I came up with the following Measure that creates a table variable to store all of the prior customer ids, then returns the distinct count of customerid's that existed in the prior periods.

Prior Purchase Customer = 
    var prior_cust_ids = CALCULATETABLE(VALUES(Cohort[Customer ID]), FILTER(all('Cohort'), 'Cohort'[Customer ID] in FILTER(all(Cohort[Customer ID]), 
            'Cohort'[Week Ending]  < SELECTEDVALUE(Cohort[Week Ending]))))
    return 
    CALCULATE(DISTINCTCOUNT(Cohort[Customer ID]), 'Cohort'[Customer ID] in prior_cust_ids)

Super simple pbix also attached

I hope that helps you out,

Richard 

[jackj]

Thank you!  This definitely helps!  One question - this will show each week the number of customers who have previously purchased at any point in time.  How would I restrict this to just showing the distinct count of customer ID's who purchased *last week* only, instead of looking back to the beginning of time?

Thank you so much!

[jackj]

Thank you!!  This works!  One final question - how could I modify this to show, for each week's cohort, how many of those users returned the following week?

I very much appreciate your help!  This is tremendous.

[richbenmintz · Accepted solution]

Hi @jackj,

See the measure below, should do the trick

Purchase Customer Next week = 
    var prior_cust_ids = CALCULATETABLE(VALUES(Cohort[Customer ID]), TOPN(1, FILTER(all('Cohort'), 'Cohort'[Customer ID] in FILTER(all(Cohort[Customer ID]), 
            'Cohort'[Week Ending]  > SELECTEDVALUE(Cohort[Week Ending]))), Cohort[Week Ending], ASC))
    return 
    CALCULATE(DISTINCTCOUNT(Cohort[Customer ID]), 'Cohort'[Customer ID] in prior_cust_ids)​

[richbenmintz]

Hi @jackj ,

give this a try

Prior Purchase Customer Last week = 
    var prior_cust_ids = CALCULATETABLE(VALUES(Cohort[Customer ID]), TOPN(1, FILTER(all('Cohort'), 'Cohort'[Customer ID] in FILTER(all(Cohort[Customer ID]), 
            'Cohort'[Week Ending]  < SELECTEDVALUE(Cohort[Week Ending]))), Cohort[Week Ending], DESC))
    return 
    CALCULATE(DISTINCTCOUNT(Cohort[Customer ID]), 'Cohort'[Customer ID] in prior_cust_ids)

[jackj]

Perfect!  This gives me the numbers I need.  One final, final question - thank you again for all this help!  How can I show a list of the Customer Id's that meet these criteria - those who visited the following week and those who did not?  Really struggling to wrap my head around all of this and it is so much easier in Excel.  Thanks so much for all of your assistance here.