Hi everyone, here for a query.
We want to do multi...
# announcementsb
brief-hair-56764
08/22/2022, 9:19 AMHi everyone, here for a query.
We want to do multiple experiments and instead of using strings,numbers or JSON , we use a nested ON/OFF to get it working but the traffic split seems to be skewed.
Can someone help out ?
Eg - To run test A,B,C ( 20%,20%,20% ) along with Prod Build ( 40% ) - We have setup 3 exp
exp-a, exp-b and exp-c and logic goes like
if ( exp-a )
{
if ( exp-b )
{
if ( exp-c ) { log ( exp-c); }
else{
log ( exp-b );
}
}
else
{
log ( exp-a );
}
}
else
{
log ( “prod”);
}
}
f
future-teacher-7046
08/22/2022, 10:15 AM
what traffic splits did you choose for the different experiments? I don't think nesting like that will work very well.
future-teacher-7046
08/22/2022, 10:19 AM
I think it could theoretically work if you do a 60/40 split for A, a 66.7/33.3 split for B, and a 50/50 split for C. But you would have to do custom assignment logging and not use the built-in trackingCallback.
b
brief-hair-56764
08/22/2022, 10:47 AM@future-teacher-7046 This is exactly how we have done. 60/40 for A, 66/34 for B and 50/50 for C but haven’t done custom assignment logging. Where can i read more on that?
brief-hair-56764
08/22/2022, 10:30 PM@future-teacher-7046?
brief-hair-56764
08/23/2022, 8:50 AM@fresh-football-47124 Can you help here?
f
fresh-football-47124
08/23/2022, 8:50 AM
I can try
fresh-football-47124
08/23/2022, 8:51 AM
I think jeremy means that you would have to check the user attributes and fire the tracking event yourself
b
brief-hair-56764
08/23/2022, 8:51 AMIs there a document for this?
f
fresh-football-47124
08/23/2022, 8:52 AM
I'm not sure why you don't just use a numeric or string experiment
fresh-football-47124
08/23/2022, 8:52 AM
and just map those values to the sections of your experiment
fresh-football-47124
08/23/2022, 8:52 AM
would be a lot easier
b
brief-hair-56764
08/23/2022, 8:52 AMWe had a code running and it worked for us so expanded on it
brief-hair-56764
08/23/2022, 8:53 AMBut could this be because of Growthbook’s deterministic hashing that nested booleans don’t work as expected?
f
fresh-football-47124
08/23/2022, 8:55 AM
we can help you out in a few hours, if you want to meet - we can perhaps help you find a more scalable way
b
brief-hair-56764
08/23/2022, 9:01 AMThat helps. We are from India and we couldn’t find best slots in the next two days. If any ad-hoc support can be given in the next few hours it would be very verry helpful!
f
fresh-football-47124
08/23/2022, 9:03 AM
any of these times work? https://calendly.com/jeremy-588/30min
b
brief-hair-56764
08/23/2022, 10:57 AMHave scheduled for 8:30 pm today
brief-hair-56764
08/23/2022, 10:57 AMIST
67 Views