What is this alert for?
To detect a variation in the number of times a brand has been mentioned in social networks with or without sentiment assigned in the platform.
How does it work?
From a brand (selected when the alert is configured) the system takes into consideration the average of times the brand was mentioned during the last 7 days. Based on that average, the system sets the normal value of mentions to the brand.
However, overcoming the average does not necessarily indicate an abnormal situation (e.g there are brands with more mentions at noon that at midnight). It is necessary to calculate an additional metric to understand variation between peaks and valleys. For this, it is calculated the standard deviation.
Variation. What determines?
The type of variation establish the sensitivity threshold the system takes to trigger the alerts.
You can choose between:
Low = Superior 1.5 standard deviations above the historic average calculated.
Medium = Superior 2 standard deviations above the historic average calculated.
High = Superior 3 standard deviations above the historic average calculated.
How is this calculated?
It triggers an alert when the current average (last 24 hours) exceeds the average of the last 7 days, plus the selected threshold (low / medium / high).
Calculation: [average last 24 hours] > [average last 7 days] + [standard deviation] * [selected threshold] |
Selected threshold could be 1,5, 2 or 3.
It is advisable to start with a Low variation and increase it when the alert is released very often.
In this example, the alert will be sent if you have selected a low or medium threshold.
What is the verification interval?
The system searches for a variation in mentions to the brand, once per hour.
If it detects that the current average (last 24 hours) continues to exceed the threshold defined in the alert (low / medium / high) , the system will send the corresponding notifications.