Hi, Harshita. By design, Proctor is meant to make allocations in a way that is statistically similar to random. You can hack Proctor to do this by having multiple allocations at 100% that assign different buckets. Each allocation would have a rule that is true for some parameter that is passed in. We have done this in cases like activating features for some countries but not others, like having a rule ${country == 'CA'} that assigns test1, ${country == 'JP'} that assigns test2, etc. However, that's in the context of incremental roll-out for each of those countries; we don't start at 100%. If you're just segmenting specific users based on criteria determined in your application code, and you're not doing an incremental roll-out or unbiased sampling, Proctor seems like it may be an unnecessary moving part.