Expand my Community achievements bar.

Webinar: Adobe Customer Journey Analytics Product Innovations: A Quarterly Overview. Come learn for the Adobe Analytics Product team who will be covering AJO reporting, Graph-based Stitching, guided analysis for CJA, and more!
SOLVED

Anomaly detection on a weekly level using 1 extra day?

Avatar

Level 1

Hey there, I'm looking for some assistance with Anomaly detection in AA.

 

It appears that if you set the alert cadence to weekly, it will include 1 extra day in the calculation.

 

If this day is used in anomaly detection as a full week, it will surely trigger thresholds as it's 1/7th of the week's data representing a full week.

 

Can someone help clarify whether this is being used in calculations, or if we can exclude this single day?

Screenshot 2024-01-18 at 3.00.01 PM.png

 

Thank you!

Topics

Topics help categorize Community content and increase your ability to discover relevant content.

1 Accepted Solution

Avatar

Correct answer by
Community Advisor

Hi, may I ask what metric you are using in this alert? I don't actually use the anomaly detection in my alerts, so I can't see this directly, but in theory I should be able to build this table in Workspace to replicate.

 

I know that the Bounce Rate and Entries metrics always seems to have this oddity (where there is one extra day returned, and this happens no matter what granularity).. so I am wondering if this could be related to one of those metrics... Bounce Rate is "Bounces / Entries" - so it seems that the Entries is the real offender here, and I have never been able to figure out why this happens... It's likely a bug that no one reported?

 

If you are dealing with Alerts, it's unlikely that this will cause a trigger.. since this one day out of the week is always on the older weeks, not the new... your Alerts will fire for the week that just passed.. the report when you look at it, shows the 12 week trend so you can understand the anomaly, but the alert itself (if it were to be sent this week - probably on Sunday), would be for the week of Jan 7th (to 13th)... the oldest week of Oct 15 (which only contains Oct 21) wouldn't trigger a 'current' alert.

 

Not sure if this helps... 

View solution in original post

2 Replies

Avatar

Correct answer by
Community Advisor

Hi, may I ask what metric you are using in this alert? I don't actually use the anomaly detection in my alerts, so I can't see this directly, but in theory I should be able to build this table in Workspace to replicate.

 

I know that the Bounce Rate and Entries metrics always seems to have this oddity (where there is one extra day returned, and this happens no matter what granularity).. so I am wondering if this could be related to one of those metrics... Bounce Rate is "Bounces / Entries" - so it seems that the Entries is the real offender here, and I have never been able to figure out why this happens... It's likely a bug that no one reported?

 

If you are dealing with Alerts, it's unlikely that this will cause a trigger.. since this one day out of the week is always on the older weeks, not the new... your Alerts will fire for the week that just passed.. the report when you look at it, shows the 12 week trend so you can understand the anomaly, but the alert itself (if it were to be sent this week - probably on Sunday), would be for the week of Jan 7th (to 13th)... the oldest week of Oct 15 (which only contains Oct 21) wouldn't trigger a 'current' alert.

 

Not sure if this helps... 

Avatar

Community Advisor

I should add.. when building reports using Bounce Rate or Entries... I've never been able to figure out a proper way to exclude that extra row or data... If you select rows 2-13, right click and select "display only selected rows" you will break the "rolling" dimensions... the table will perpetually show those dates, but the ones out of the scope of the range will actually show no data.... basically, it will break the report....