Trend shift alerts let you cut through noisy data and focus only on the signal -- the normal trend of the data. With trend shift alerts, we are looking to determine a significant change in data automatically. If there is an abrupt change from the normal performance, a trend shift alert will detect that and trigger it. It will not trigger on tiny spikes that do not cause a significant shift in the data trend.
Trend shift alerts grow and adapt with the test data. Gradual changes over time will not impact the accuracy of the alert. This is better than using a static threshold that would have to be updated whenever changes are made to the endpoint - such as optimizations or features added to the page or application that would affect performance. With trend shift alerts, you won't need to change your alert thresholds every time you roll out new features or bug fixes.
To create an advanced alert with the trend shift trigger, please take a look at the Creating Advanced Alerts article and the below for more instructions:
The trend shift alert option is available under the following advanced alert types:
- Timing
- Insight
- Ping
- Requests
- Zones
Once you have switched to Advanced Alerts, you will find the Trend Shift option under the above alert types in the Trigger section.
If Node Thresholds for runs and nodes are specified, the alert will check back the specified time from the Time Threshold and look for the specified number of runs/nodes that have experienced a trend shift. If the alert threshold is exceeded, the trend shift alert will trigger.
The Consecutive setting under the Node Threshold for runs and nodes will apply after the initial trigger to determine consecutive triggers. In order for a consecutive trend shift alert to trigger, it will require multiple trend shifts to exceed the alert threshold consecutively.
