Why we alert on change, not on thresholds
Fixed thresholds go stale. Comparing this period with the last one keeps alerts useful as you grow.
The first alert most teams create is "tell me if visitors drop below 500 a day". It works for a month. Then traffic doubles and the alert never fires again, even when something is clearly wrong.
What goes wrong with fixed numbers
A threshold encodes what was normal on the day you created it. As your product grows, normal moves, and the number quietly becomes meaningless.
Compare with yourself
Orbit alerts compare the latest period with the one before it. A rule such as "signups drop by 30% over 7 days" keeps working at 50 signups a week and at 5,000.
- Pick a metric that maps to a decision: signups or revenue are better than raw visits
- Choose a period long enough to smooth out weekends
- Set the drop at a level you would act on, not a level that would merely be interesting
Let the alert explain itself
An alert that says "visitors fell" starts a hunt. An alert that says "organic search traffic went from 420 to 250 visitors" starts a fix. Every Orbit alert includes the source or page that moved the most.