Skip to main content

Hi,
 

Want to ask regarding how to properly group & segment user based on new, first time buyer, & repeat buyer from Amplitude team perspective.


The definition that we use for to our user segment as following: new user is group of user that never purchase, first time buyer is group of user that only buy one time. repeat buyer is group of user that ever purchase twice and above.

Our current approach:

  • new: performing pay=0, within 30days of first use
  • first time buyer: performing pay=1, during exact timeframe that we set (i.e 30 days)
  • repeat buyer: performing pay >=2, during exact timeframe that we set (i.e 30 days)

we use within because we want to know group of user that never purchase during the timeframe but did the event within their 30 days of first time use. we realize there is a intersection of users since when we do funnel analysis we found that when we add new, first time buyer, and repeat buyer, total number of users is not equal to the total all user numbers.

Is there any approach that we can do to get the exact number of group user during exact timeframe that we want? Kindly share what is the best practice to do this kind of analysis

Hi @dedy.kurniadi ,

 

Thanks for writing into Amplitude Support! 

 

Hmm, I think a good option here would be the historical count offered by Amplitude but you would need to be on the Growth or Enterprise plan. I am not sure if you have access to cohorts or not but you could also use them to create these target groups. 

  • new: performing pay=0, within 30days of first use
    You could consider looking for users who never performed new in the last 365 days. It will span a year but this should be enough time. I would use the cohorts or if you don’t have them, the in-line cohorts in segment to define new as anyone who has not done the pay event in the last 365 days. 
     
  • first time buyer: performing pay=1, during exact timeframe that we set (i.e 30 days)
  • repeat buyer: performing pay >=2, during exact timeframe that we set (i.e 30 days)
    I think historical count would be the best option here but it would require an upgrade. Although, which plan are you on? Regardless, otherwise I would use in-line cohorts to segment these users and look if they did the “pay” event. 

I hope this helps for now!

Kind Regards,
Denis

 

Reply


Terms & ConditionsCookie settings