Functions in calculated metrics - aggregate vs. per-visit calculations

[RyanTerrell]

Hello,

I created a calculated metric for the standard deviation of content velocity, which is pageviews/visits. The metric is not behaving as I would initially have suspected. When I pull in the day dimension, I am given the standard deviation for the aggregate page views / aggregate visits value over each day of the timeframe. Rather than getting the standard deviation for the average content velocity each day, I am hoping to get the standard deviation for an individual visit's content velocity. Is this possible?

I was able to pull in visitor id (this is not at the visit level, but closer to my goal) as the row dimension, which I believe gives me the standard deviation of content velocity on the visitor level. However, when I filtered/removed visitor ids that were null from the table and several specific visitor-ids, I noticed that the standard deviation metric value did not change. Is pulling the visitor id in as the row dimension giving me the standard deviation of content velocity for unique visitors, or is this also giving me the standard deviation of an aggregate metric? Also, when I remove/filter out specific rows from a table, how can I make these changes reflect in my standard deviation metric (or do I need to create a segment to filter them out)?

This question is related to a previous thread here.

Thank you!

[Jennifer_Dungan]

This is an interesting question.. I don't know if the built-in functionality will give you the granularity that you need... Since there is no way to really choose the scope, it's basically determined automatically which isn't really what you need here...

This may be better in something like Report Builder, where you can get the data and then use more advanced calculations to achieve your needs..

That said, I wouldn't give up just yet... try playing with different options and thinking outside the box and see if you can make it work... 

[RyanTerrell]

Thank you for your reply and encouragement. I will keep trying and look more into report builder!

[leocwlau · Accepted solution]

Hi @RyanTerrell, referring to the documentation that the standard deviation function has the (Table) on it, 

https://experienceleague.adobe.com/docs/analytics/components/calculated-metrics/calcmetrics-referenc.... So it is giving you the standard deviation over the table rows, that is the dimension you are using. If you have two levels of dimensions, day and hour of day, the first level of dimension day has the same value of stddev calculated from the content velocity of all days, and the second level dimension of each day could have different values but same under each day, calculated from the content velocity of all hours from the same day.

Using visitor ID as the dimension gives you the closest answer but you need to be aware that it is content velocity per visitor where the same visitor may have multiple visits with different content velocity. For the value remains unchanged if you apply a filter or remove some values from the visitor ID, my guess is the change is indeed too small to be reflected in the freeform table and rounded out. You can try some drastic filter or segment, such as mobile device, to observe if the value remains the same or not.

[RyanTerrell]

Thank you for your help! I will try creating a content velocity metric based on unique visitor instead of on the visit level to see if that avoids the above concern, and will try filtering as you suggested.