https://www.growthbook.io/ logo
#ask-questions
Title
# ask-questions
m

microscopic-table-65564

03/26/2024, 5:40 PM
Hello, I got this warning
*Multiple Exposures Warning*. 172 users (2.56%) saw multiple variations and were automatically removed from results. Check for bugs in your implementation, event tracking, or data pipeline.
This is the first time I’ve seen this in an experiment. It was after I increased the traffic percentage and used the Same Phase, apply changes to new traffic only (with sticky bucketing). I didn’t change anything in my implementation. Do you have any idea what could have happened? Should I interpret this as a signal to restart the experiment from the beginning?
Any advice?
f

fresh-football-47124

04/05/2024, 4:38 AM
did you change the split percentage?
m

microscopic-table-65564

04/05/2024, 3:48 PM
Yes, we did and we used sticky bucketing
@fresh-football-47124 Sorry, no, we did not change the percentage, just the percentage of traffic, but not in the variants’ weight. We had the experiment set for just 20% of the traffic and we increased it to 30%
@fresh-football-47124 This happened again with a new experiment, which means that the percentages or weight variations have never changed. The only change in recent weeks was that we enabled the sticky bucketing in our account and updated the SDK accordingly. Do you think that’s related?
4 Views