Hi, any tips on how would I go about configuring a holdback group of users that would be excluded from all experiments for a given time period? The idea is that at a later stage this holdback group will be divided to a treatment group receiving all changes at once and a control group -> we could measure first each experiment on it's own and then later measure the combined effect of multiple changes

It is possible to make it work using namespaces, which we just launched. That feature was built to enable mutually exclusive experiments, but it can also work for holdout groups. Under settings, you would create a single global namespace. For each experiment rule you create you would enable the namespace and select a range like

0 to 0.9

. That means the last 10% of users in that namespace would be excluded from the experiment. Then later for your holdout experiment, you could set the range as

0.9 to 1.0

to run a combined test on the holdout group.

Sounds good, thanks Jeremy!