I have created an experiment, and I want two groups to be added to that specific experiment, group 1 (Internal Users) should only be shown experiment version and remaining users should have divided traffic. I can't find a way to link this group to the experiment. Can someone help me with this

You just need to create a "Forced Rule" on your feature for the internal users as the first rule and an experiment rule as the next rule.

I have created experiment without creating feature flag. And in the code, I run experiment using growthbook using SDK. Isn't it the correct way of performing experiments

There are inline experiments in the SDK but they are not as powerful as creating a feature flag that would allow you to do the Forced Rule targeting

Understood. So kindly verify the process 1. I will create a feature flag add forced rules which will pick something for default values from saved groups, if not found in forced rules, it will return a random response (I mean splitter traffic) 2. And then add the Rollout rule to just control the rollout. 3. Then attach the experiment dashboard for the analytics. Will the variations mentioned here will be taken care of while deciding to enable or disable the feature? In step 3, can I attach the existing experiment that is already created, if yes how? I tried using experiment_id and tracking_key on the dashboard, but it was showing to create new experiment.

I think you probably don't want the rollout rule, but for the experiment rule don't include everyone in the experiment. The experiment itself traffic split you probably want to be 50/50. With 0 being the "control" and 1 being the "new version". If you include only 1% of users in the experiment, then only 0.5% of users will be seeing the new version. If no bugs are shown you can increase the number of users in the experiment then, so that you get enough data to make a decision in a reasonable amount of time.

Yes, I understood the rollout rule. But can you comment on these 1. Forced rule will override everything (including A/B) experiment rule. Which means even if I have split of 90/10 or 100/0, if the id is found in the forced rule, will return ON (as I have configured) 2. If say I have same kind of experiments for different features. Is there any simpler way to use one type of experiments only for different feature flags? As I don't want to create experiment dashboard all the time. I have too many feature flags. 3. Assume, I have created experiment from feature so in this case I can modify phase or other stuffs that are supported in experiment. And the ongoing experiment will incorporate such changes directly at feature flag level, or do I have to remove the experiment and create another one or what is it like?

Hi Aditya - the rules are applied in order, and when there is a matching rule, no other rules are used.