Hi, I felt confused when using namespace. If I want to use 10% traffic for control and 10% for treatment, should I specify the allocation? I think the allocation is redundant.
r
rhythmic-agent-34208
07/09/2024, 8:10 PM
Hello Yi, thanks for reaching out.
The Namespace allocation is not for traffic to a single experiment, it's to allow more than 1 experiment to be in the Namespace so that the experiments can be mutually exclusive.
If you need to run 3 experiments that all need to be mutually exclusive, and you want your traffic distributed evenly among the 3 experiments, you would set up your namespace allocation to be [0 to 0.33, 0.34 to 0.66, and 0.67 to 1] which would roughly be in thirds.
Does that make sense? Let me know if you have more questions.
https://docs.growthbook.io/features/rules#namespaces
f
future-printer-83389
07/10/2024, 2:46 AM
Hi @brief-honey-45610, I understand the exclusive exp usage, but I think it's better to specify the abslote traffic proportion of each varation rather than set up the namespace allocation then setup the each variation proportion. Otherwise, it's meaningless if variations use full of the namespace allocation, like the picture above.
future-printer-83389
07/10/2024, 2:49 AM
And, in Microsoft Autopilot framework, it directly use absolute proportion, it's better for user to understand. Besides, all experiment should belongs to a namespace(they called numberline).
r
rhythmic-agent-34208
07/10/2024, 4:16 PM
Thanks for the feedback and suggestions. I'll pass this along to the team.