Based on your description, if you set the Traffic Allocation to 50% and used

deviceId

as the Assignment Attribute, you would expect about 50% of devices to be included in the experiment. However, only 17% were included. Here are some possible reasons for this discrepancy, based on the knowledge sources: 1. Traffic Allocation Setting The "Percent of traffic included in this experiment" setting determines what portion of your total traffic is eligible for the experiment. If you set this to 50%, and your experiment split is 50/50, then each variation should get 25% of the total traffic, and 50% of users should be included in the experiment. If only 17% are included, it suggests something is filtering out users before assignment, or the allocation is not being applied as expected [Splitting traffic before assign users](https://linen.growthbook.io/t/2578438/hi-i-have-a-question-on-splitting-traffic-before-assign-user). 2. Attribute Mismatch or Targeting Filters If there are additional targeting conditions (e.g., only certain app versions, device types, or other attributes), these could further reduce the number of users eligible for the experiment. Double-check your experiment's targeting rules to ensure there are no unintended filters. 3. SDK Implementation Issues If the GrowthBook SDK is not being initialized correctly, or if the

deviceId

is not being set or passed properly, many users may not be evaluated for the experiment. Make sure that every device has a valid, unique

deviceId

and that this attribute is being sent to GrowthBook for every user session [Assignment attribute issues](https://linen.growthbook.io/t/16656229/hi-is-anyone-else-having-issues-regarding-the-assigned-attri). 4. Evaluation Timing If the experiment evaluation is only triggered under certain conditions (e.g., only on specific screens or actions), not all users may be considered for assignment. Ensure the experiment evaluation is happening for all relevant users. 5. Coverage and Split If you have a 50% traffic allocation and a 50/50 split, each variation should get 25% of the total traffic. However, if you are seeing only 17% included, it suggests that the coverage or targeting is more restrictive than intended, or there is a bug in the implementation. 6. SDK Version or Bug There have been reports of bugs or misconfigurations causing unexpected assignment rates. Double-check your SDK version and implementation, and consider sharing your code with the GrowthBook team for review if the issue persists [Sample ratio mismatch and assignment issues](https://linen.growthbook.io/t/26831141/hi-question-on-assignment-we-re-seeing-a-sample-ratio-mismat). Next Steps: - Review your experiment's targeting rules and ensure no extra filters are applied. - Confirm that

deviceId

is always set and unique for each device. - Check that the GrowthBook SDK is initialized and attributes are passed correctly for every user. - If possible, check the logs or analytics to see if there are patterns in which users are being excluded. If after these checks the issue remains, consider sharing your SDK initialization and attribute-passing code for further troubleshooting.