What it does
Fully automatic -- there is nothing to configure to start receiving alerts. The same significant-change detection History uses is filtered down to real, significant, non-insufficient-data changes, persisted as alerts, and deduplicated.
Beyond the in-app list, a real email is sent to the organization's owner for each generation run that finds something significant (one email per run, not one per individual metric, even if several changed at once), with a link to the evidence. Alert emails specifically (not the in-app list) can be turned off in Settings.
Each listed alert shows an improved/declined badge, a label, the prior and current values with a delta indicator, the measurement timestamp, and a link to the underlying evidence.
How to use it
- 1Nothing to configure to start -- alerts appear automatically after the next measurement run that produces a significant change.
- 2Read the list: the badge shows direction, values show prior vs. current.
- 3Click 'See evidence' to jump to the underlying proof for any alert.
- 4To stop the emails specifically (the in-app list is unaffected), turn them off in Settings.
Known limits
- The exact statistical significance method (shared with History) wasn't independently verified at the formula level -- only its behavior as a real/noise gate was confirmed.