We’Re very confused about growthbooks temporary rollouts: How do they work? Let’s say we have: • Baseline=a And in the temporary rollout we roll out

b

Will now every single user get

b

? Or will only users who were part of the test get

b

?

Yes, all experiment traffic will be sent to the winning variant (but no more tracking callbacks will be fired).

@strong-mouse-55694 What does Experiment traffic mean? If we stop an experiment, would it have any traffic?

Hey @strong-mouse-55694, I work together with Paul and it's really a bit confusing for us at the moment. I'll try to illustrate it with a "simple" example. It would be great if you could confirm that the following statements are correct or give us an explanation/correction if this is not the case. Thank you very much 🙏 Example: We have a test with

variant A = baseline

and

variant B

. Variant B shows significantly better results during the test period, whereupon we stop the test with status Won, select variant B as the winner and activate a temporary rollout. We only selected a single condition for user targeting:

appversion >= 34

. Statements 1. after the end of the test, no new users are assigned to the two variants 2. after the end of the test, all users who were assigned to one of the two variants during the test period now see

variant B

3. all users who were not assigned to either variant during the test period and now have an

appversion >= 34

will see

variant A = baseline

after the test is completed 4. all users who were not assigned to either of the two variants during the test period and now have an

appversion < 34

will see

variant A = baseline

after the end of the test 5. only if the default value of the corresponding feature flag is changed in the base code,

variant B

is also rolled out for all users who were not assigned to either variant during the test period 6. the temporary rollout does not have to be stopped manually in Growthbook Points 3-6 in particular are questionable for us. Thanks for your help 🙏

1. Yes. 2. Yes. 3. No. If a user satisfies the condition, they'll see variant B. 4. Yes. 5. If I understand, then, no. Like 3, B will applied to all users who meet the targeting condition. Users who don't will get the default value. 6. The temporary rollout, as the name suggest, is intended to be temporary. Ideally, you'd make the change to your codebase to serve the winning value and then remove the flag/experiment altogether. There may be ways to simplify if you can share a bit more about what you're trying to accomplish overall.