Expand my Community achievements bar.

Check out the November edition of the Analytics Community Lens newsletter to see what's been trending in the last two months!

Adobe Alert - How do they calculated "expected" value


Level 2


I have an alert that the daily visits were 35% above expected - how does adobe calculate the "expected" value (and thus are able to calculate the % above/below)? 


3 Replies


Community Advisor

The "Expected Value" is essentially a value that is derived from your historical trends.I don't know all the math that they put into it specifically, but it will take into account the granularity of the data, probably the day of the week and time of the day... then based on the history of your data collection, and throwing out some of the outliers in the historical trends so as to not be heavily swayed by other anomalies (probably using some standard deviation rules), it determines what it believes to be a accurate representation of your data and compares the current data against it....


Whether or not you feel that the data is truly outside of expectations is another story entirely


If you are using the Anomaly Detection in Workspace, this shows you where it believes the data should be, and flags items that fall outside of it's range... personally (at least in my reports) I have never felt these to be terribly accurate or useful (but I work in News... so it's a pretty volatile market, depending on what is happening... big things happening in politics? "Feel Good" community stories that go viral? Product recall notices? There are so many factors that impact what and when people are reading, that Adobe's anomalies don't really help us out much)


Here is some more reading on Anomaly Detection: https://experienceleague.adobe.com/docs/analytics-platform/using/cja-workspace/virtual-analyst/anoma... which is what is also used to create the type of alert you are seeing....


Level 4

One of the sessions at Adobe Connect 2023 suggested trying to use ratios (orders per visitor for example), more than raw numbers (orders) so as to receive less false alarms.


I believe it was in the Rockstar session if you're interested.  

Lol as I looked for the session to give the link, I realized it was Jennifer who posted above!  She's got some great ideas that she gives!  Starts around 28 minutes in:  https://business.adobe.com/summit/2023/sessions/2023-adobe-analytics-rockstars-top-tips-and-tricks-s...


Community Advisor

Lol, yup.. that was me  


I focused on custom rates (where we make the math) as opposed to the Adobe provided thresholds... but play around with both!