Hi group,
Has anyone else experienced issues with...
# ask-questionsp
polite-hairdresser-26298
03/18/2025, 8:43 AMHi group,
Has anyone else experienced issues with sticky bucketing? We enabled it to help to improve multi-exposure, however, for one test that had resulted in negative impact on pricing element on cart. According to development this happened since sticky bucketing ignores targeting. Can anyone suggest anything?
We've been having a lot of issues with GB lately.
š 1
f
fresh-football-47124
03/18/2025, 9:27 AM
sticky bucketing will keep a user in the experiment based on their initial assignment. So its possible that the targeting is being skipped
fresh-football-47124
03/18/2025, 9:29 AM
I'd have to check on the specifics
p
polite-hairdresser-26298
03/24/2025, 12:00 PMOne similar question. started later on what is the guarantee that the split between first test and variant does not cancel out the introduction into the second test. I have often seen that results change dramatically as soon as another test is started which normally should not be the case I would guess.
I guess my question is if you think it is better to always use namespace to exclude ab test crossing or if perhaps this will ensure the outcome is reached sooner?
f
fresh-football-47124
03/25/2025, 6:40 AM
there can be interactions between experiments (and certainly if its the same part of the product, you should use namespaces), but meaningful interactions are rare.
š 1
p
polite-hairdresser-26298
03/26/2025, 3:09 PMIs the multi exposure notification shown based on the same phase date range or is it calculated somehow differently. Currently I have a situation in Android app with multi exposure count being twice as high as users exposed to the test. Any suggestion what can be to blame?
f
fresh-football-47124
03/26/2025, 3:10 PM
I Believe its within a phase
fresh-football-47124
03/26/2025, 3:12 PM
here are some reasons why you can get multiple exposure warnings:
ā¢ you changed the split mid experiment phase
ā¢ you changed the split in its own phase, but have sticky bucketing turned on
ā¢ the attribute you're using for assignment is different than the id you're using for reporting, and the one for assignment is not as long-lived
ā¢ the attribute you're using for assignment is conditional
ā¢ you're not using the built in trackingCallback and instead firing you're own
š 1
p
polite-hairdresser-26298
03/26/2025, 4:12 PMWe did have an issue with id, which were (not set) as expected so I will know tomorrow if this was the reason
polite-hairdresser-26298
03/27/2025, 4:51 PMWe still have issue with multi exposure. Can this be due to fallback not being set and left as "null"? I mean we see a, b and "null" as the values.
I guess what I mean by that if "null" or fallback also participates in the multi-exposure trigger count.
12 Views