Hello , is using uuids instead of numeric integers...
# give-feedback
d
Hello , is using uuids instead of numeric integers or vice versa as the key improve anything?
f
improve?
d
I'm currently using integers and I can see a big ratio mismatch, thats why im asking
To elaborate more
f
oh, the sample ratio mismatch shouldn’t depend on the hashing attribute
d
I have a case that i should filter out people based on some conditions
So instead of adding these as attributes and then adding forced rules which could be quite complex
Im adding a condition in the app code before even checking of this user is in the experiment or not
Also the given these strict conditions, the overall exposed users is very low , so not sure if this could be related
f
would you like to share with me the targeting your using?
it could just be the small sample size
d
Yes so I'm basically we are a subscription service where customers have a schedule of appointments/bookings , and I'm targeting users who are only transitioning from their 5th to 6th appointment during the time of enrollment to the experiment, after an appointment is done I check if they meet there requirements and if yes then check for their exposure,
Note that if customers already passed this step , for example from their 6th to 7th appointment, then i exclude these customers
And also I'm currently applying an exposure percentage of 20% only
f
are you adding those rules to the targeting conditions, or manually doing it?
d
No I'm doing it manually in the application code before checking isFeatureOn
Interested to know actually if it makes a difference
f
you could add the number of schedules as an attribute, then use the UI to target it to 5