Alerts that do not shout
An alert only fires when the change is statistically real, and the screen explains the criterion. No alarms over sampling noise.
- The test uses Wilson intervals at 90%: if the intervals of two windows overlap, there is no alert.
- Five types: share of voice drop, presence drop, new competitor, first citation and exhausted budget.
How it works
Each rule compares two windows of the same metric with Wilson intervals at 90%. There is an alert only if the intervals do not overlap: sampling noise wakes nobody up. Five types: share of voice drop, presence drop, new competitor, first citation and exhausted budget.
What you see on screen
The active rules with their criterion and the fired events with the figure and its denominator. Notices arrive by email.
Who it is not for
There are no real-time notices: the cadence follows your plan’s sweep. And if your plan does not include alerts, the screen says so clearly instead of hiding the module.
Availability by plan
On Pro and Agency.
The most repeated question
- Why did it not warn me about a drop I can see in the chart?
- Because with your current sample that drop cannot be told apart from chance at 90% confidence. We prefer to stay silent rather than make you react to noise; with more prompts or repetitions the interval narrows and the same drop would fire the alert.
The scanner runs 10 real responses and returns your score with its breakdown in about 60 seconds, no sign-up.
Free scanner