GoodDAPP icon indicating copy to clipboard operation
GoodDAPP copied to clipboard

Invite - Adding A/B test to the invite massage

Open Esfir-E opened this issue 4 years ago • 7 comments

Description

  1. We would like to have the ability to A/B test 2-3 types of content for the Social share for invite feature.
  • Once the user is copying our link and share it on social media's the link is shared with content, we would like to test 2-3 options of content for the same Social channels.
  • We should have the ability to monitor the version
  • We should have the ability to switch off the A/B test - feature flag and to stay with the content of our choice.

Implementation

  • useAbTesting - either make it return the random value and decide according to it the variant number, or modify it to work with more than 2 variants.
  • the invite link should contain another param ab=# where # is the variant number
  • add the ab param value to events, call it inviteab
    • inviter - share event
    • invitee - app_open event
  • feature flag to turn off test

Esfir-E avatar Feb 22 '21 09:02 Esfir-E

@Esfir-E how do you intend to measure the effectiveness?

sirpy avatar Feb 22 '21 09:02 sirpy

@sirpy if i will know which option given to whom i will be able to measure that once one the users friends who opened a wallet

Esfir-E avatar Mar 01 '21 20:03 Esfir-E

yes i understand that, i ment which tool are you going to use and how do you plan to configure it bcause but we dont have any tools at the moment to aggregate the data in such a way for meaningful statistics @Esfir-E

sirpy avatar Mar 02 '21 09:03 sirpy

@Esfir-E we have already an mechanism allowing us to show one of two varian of some screen

  • case A or B determined or first screen visit and kept per session (unit app reload)
  • optionally we could cache it forever
  • optionally we could define some event to be sent with parameter indicating the A/B case was applied

I've added estimation based on the current A/B implementation. If only two cases isn't enough for you - we'll need to adjust it, which will increase task complexity

Please update us - would two cases A and B enough for this task or not

serdiukov-o-nordwhale avatar Mar 03 '21 11:03 serdiukov-o-nordwhale

@Esfir-E do we have an effective way to measure success?

omerzam avatar Mar 03 '21 14:03 omerzam

@omerzam we will add another param to the invite link. ab=# where # is the variant number

sirpy avatar Mar 03 '21 14:03 sirpy

@Esfir-E i've added details on implementation. are there other events it is important to have the invite ab variant atttached to?

sirpy avatar Mar 03 '21 14:03 sirpy