Top contributors in anomaly detection

Q&A
, ,
1

Can the top contributors details within an anomaly detection insight be displayed in the visual on the dashboard so it can be accessed by users of an embedded website, or are they only visible for those with console access?

2

Hello @lm155 , welcome to the QuickSight community!

Can you clarify what you mean by the “details” of the anomaly detection? Do you mean that you want to display the fields that feed the data or that you want to show something else?

3

hey, I am referring to analyzing top contributors the “explore anomalies”. Is there a way to include those in the narrative if that makes sense?

4

Hey @lm155

Okay, I think you are referring to this: Exploring contributors (key drivers) - Amazon QuickSight

To do what you are looking for you can use the narrative to show all the contributors. You will just need to pull out the correct portion of the computation and find the index. For example, it would look something like this:
Top3CustomersSales.items[index].metricValue.formattedValue
This is based on this computation:

image
image523×576 25.1 KB

So, if your contributors are in the computations you should be able to pull them out using the index. Here’s a breakdown:

Is this what you are looking for?

5

That is what i am asking, however i dont see any narrative items here that fit top contributors?

image
image1915×645 74.6 KB

6

Hey @lm155

You will have to recreate it. This won’t be perfect but I think it will be the closest we can get. If you know the fields that you are using for the anomaly detection you can use the other narrative computations (like top ranked etc…) to recreate the top contributors.

7

Ok, would there be custom narratives anywhere that link directly to the top contributors of an anomaly, or would i have to be manually finding how to make those variables appear. Mostly checking if the anomaly detection gets re-run, would the manually added contributor insight you suggested also change?

Thanks

8

Hey @lm155

I don’t think there is a computation that specifically links to the anomaly detection so it would mean some trouble shooting. To your point, you will probably have to use the fields that make up your anomaly detection in computations to make this work so it could be a decent amount of dev time.

As far as the insight loading every time the anomaly detection runs, my assumption would be that if you take a snapshot of the dashboard at the same time it would load the insight narrative against your data and should be the same.

(Sorry for the late reply, I was OOO and catching up)