In an experiment, Impressions and Events do exist but looking at Metrics page, no data or not all unique users show in Metrics cards
- Minimum amount of unique users required to show data is not reached, see this article for more information.
- Any user that got different treatments within the same experiment version twice, meaning the treatment flipped two times using different rules will be excluded from the Metric calculation
- Any user that got different treatments within the same experiment version and rule once will be excluded from metric calculation
- A bucketing key is additionally used in getTreatment call, and its causing treatments to flip.
For more information check Split docs page
To avoid users getting excluded from Metric calculation, make sure to avoid following scenarios:
- If a percentage is used to for treatments, always increase the "On" treatment vs "Off" in one direction, increasing percentage and decreasing it afterwards will cause some users to flip treatment in the same rule.
- Be careful when whitelisting using segments, when a segment is updated during the experiment, make sure users change treatments only once.
- Bucketing key can flip treatment of users if the user has treatments with different bucketing keys.
client.getTreatment("user1234", bucket_key, "split_name")