Alerts tell you, when something interesting happens in your account. It works by analyzing a time series, e.g. the number of visits per day and whenever there is a significant development you will get an email and can see the notification in the notification feed.
There are two types of alerts:
For smart alerts you select all the time series’ you want to monitor by selecting. The thresholds are calculated via machine learning, looking at how the time series behaved in the past, which means you never have to adjust the thresholds, e.g. to keep up with the growth of the site. With one Smart Alert you can monitor up to ten time series by selecting the metric you want to monitor, e.g. visits and a dimension, e.g. campaign channels. Now you select either certain values for dimension, e.g. email, display and affiliate or you just choose the top X channels and we start monitoring the time series for each selected channel.
For AI-based alerts you can select the anomaly level you want to be informed about, where “exceptional” means that you only want to be informed about really big spikes or drops and “minor” is already triggered when the value is slightly above or below expectation. Usually somewhere in the middle is the best level to use. After you have configured an alert you can trigger it to see the anomaly level visualized for the time series.
For rule-based alerts you define the exact conditions when you want to be informed, e.g. when a certain threshold is passed like 100.000 Visits, or the number of conversions dropped by 10%.
You can choose between:
- Change by absolute
- Change by %
- Greater than
- Smaller than
Change by absolute checks for a change of the absolute number. E.g. change by 10 means, if you had yesterday 100 visits and today either 110 or 90 visits you will be notified. If you use a prefix, e.g. +10 or -10 then you will only be notified if that is the case, e.g. for -10 only if the value dropped by 10 or for +10, if it increased by 10.
Greater than and Smaller than take absolute numbers and will be checked against the provided value, once the selected timeframe has passed, e.g. during the night for the previous day. Important notice: For hours we check after two hours have passed since many metrics will only be written once a session is completed, e.g a visit might start at 1pm and last until 2:30pm and we will add this visit to the hour between 1 and 2 pm and all the data that was connected to this visit, e.g. device class, location, session duration. Due to this data from the last hour is always prone to change a lot as session comes to an end and if checked before nearly always leads to an alert.
Frequency determines when the time series is checked, either daily or hourly. Daily means once the day is over the value for the previous day is added to the time series and the new value is evaluated against the configured (rule-based) or calculated (AI-based) threshold. If the value is outside of the expected range you get an email.
Name: Provide a clear name for the alert. It will be used in the notifications you get when this alert is triggered.
Description: Provide a good description to give an idea what this alert is about. The description will be displayed in the alert overview.
Spike Message: When the alert is triggered you can provide a message that is added to the alert. The spike message is delivered when there is an unexpected spike in the time series, e.g. if the bounce rate goes up, or the conversion rate goes up.
Drop Message: Same as above, but this message is delivered, when there is a drop in the time series, e.g. if the bounce rate drops or the conversion rate drops.
Recipient’s Email: Add all Emails to which the alert should be sent to.