Alerting can be used to send notifications based on anomalies in your data. Alerting can be accessed by "Manage"-Dropdown menu.
Users can set the following parameters:
Name
Give a descriptive name for your alert. Numbers and special are supported.
Date Range
Users can use this option to set the time range, that will be taken into account, when Condition Type is set to "moving mean".
Dimension Filter
The considered data set - needs to contain at least 1 dimension. Users can use our extensive dimension filter engine.
Read more about our filtering logic here.
Metric
The metric to be considered. Currently only calculated KPIs are supported.
Granularity
The time granularity on which the alarm is to be based. If, for example, it is set to "weekly", the system will send alarms if a certain weekly threshold is exceeded or not reached.
Condition Type
Choose between 2 different types of conditions that determine under which condition the alert gets triggered:1. Fixed Boundaries:
Manually enter lower and upper boundaries values2. Outliers By Moving Mean:
The upper and lower boundaries are defined by a moving mean, based on calculations of your data set, with the following additional options:
- Outlier percentage: Percentage of data points to be considered outliers
- Use bounds: Choose between using upper, lower or both bounds
- Window size: Defines the window size of the rolling average. The higher the value the smoother the band of non-outliers becomes. Use a window size double the length of the array of datapoints for threshold based outlier detection.
- Mean method: Choose between average
or median
(median is less prone to singular outliers)
- Cold start: Calculate the rolling average for the first and last data points (as opposed to only calculating on full window sizes)
Frequency
Choose the frequency of how often you want to send out Emails in case an alert gets triggered by a condition.
- Fire for every value: Every time the condition is met, a notification email will be sent.
- Fire at most once: The recipient will only be notified the first time the condition is met, but not for any other occurrences within the selected time granularity (e.g. weekly, daily, etc)
- Fire once for all consecutive outliers: The recipient will be notified for every new occurrence of the condition being met, but not every single occurrence within the selected time granularity.
Send as Email - Enter one or multiple Email addresses you want the alerts being sent to (you can also leave the address input field empty and just mark the checkbox Send to my email
or use both options)
The email notification will be sent at 8am CEST.
EXAMPLE: FIXED BOUNDARIES
EXAMPLE: OUTLIERS BY MOVING MEAN
Comments
0 comments
Article is closed for comments.