Notification Fatigue

Notification fatigue

Fatigue is not a user problem, it is a product signal. When engagement falls the instinct is to send more or make them louder, which accelerates the collapse. The correct response is always to send less and make what remains matter.

Method

1. Measure per type, not in aggregate. One high-volume, low-value type usually drags the whole channel down and is invisible in the average.
2. Watch the leading indicators. Declining open rate, rising time-to-open, and preference changes precede muting, which is the point of no return.
3. Treat a mute as a strong signal. Users rarely unmute, so a mute is a permanent loss of the channel rather than a temporary setting.
4. Cut the lowest-value type first. Removing one type entirely beats trimming everything slightly, because the goal is fewer interruptions, not smaller ones.
5. Consolidate rather than suppress. Several notifications becoming one digest preserves the information and removes the interruptions (see digest-design).
6. Give granular controls before users need them. Someone who can turn off one type will; someone who can only turn off everything does.
7. Re-earn the channel slowly. After cutting, resist adding new types for a period so engagement can recover.

Boundaries

Fatigue metrics show engagement, not value; a notification people ignore may still be legally required. Reducing volume can reduce short-term engagement metrics that reward interruption, which is a measurement problem rather than a reason to continue. Some users want high volume, which is what preferences are for (see unsubscribe-and-preferences).