Calculating Cumulated Users without counting them twice : CUMSUM

  • 4 August 2022
  • 3 replies

Hello Community !

We lauched a new feature and I want to measure the cumulated number of users who used it since its launch.
I am thinking about using CUMSUM(UNIQUE) on an event, with a Monthly frequency.

Let’s call the event A

I get the following chart

My question is the following : if a user performs the event once on Month 1 (Jul 22) and once on Month 2 (Aug 22)

  • is he counted as user only in Month 1
  • OR is he counted as a user in Month 1 AND in Month 2 ?

What I want is the first option, as I want to calculate the penetration rate of the feature.

Does someone have the answer ?

Thanks for your help !


Best answer by munroek-amp 5 August 2022, 14:35

View original

3 replies

Userlevel 2
Badge +3

Hello @David_B. Your choice of using CUMSUM(UNIQUE, A) will deduplicate the user count, so the user will only be counted once for the data point just like you wanted. Excellent utilization of the formula! 

Userlevel 7
Badge +10

Hi @David_B 

Using CUMSUM, this user will be counted as a user in Month 1 and also in Month 2.
i.e. a running total of uniques who performed the event across your time frame i.e Month1 and Month 2

So in July you had ~4750 users who performed that event, and the cumsum  for August (till date) shows 5825. So from July to August till date, you have 5825 uniques in total who performed the event. And this number includes users who performed the event in July too.

You can also use Active % metric to gauge feature adoption alongside the cumsum approach.

Hello @Saish Redkar & @munroek-amp , thanks a lot for your help.
If I summarize well, the CUMSUM feature will deduplicate properly the users who perform the event in both month.

Am I correct ? IN that case should I use the “Cumulative” option instead ?