topic Re: Percentile Formula in Incorta Analyzer in Dashboards & Analytics Discussions

Percentile Formula in Incorta Analyzer

mmastropaolo — Fri, 17 May 2024 19:20:46 GMT

Is there any work around to achieve performing an aggregation by [X]th percentile at the dashboard level?
I am attempting to provide a user with the ability to filter a dashboard as desired and show the [X]th percentile of a specific measure (for example 75th percentile). To my knowledge, incorta analyzer currently only allows us to aggregate via the 50th percentile (median).

Re: Percentile Formula in Incorta Analyzer

RADSr — Mon, 20 May 2024 18:38:08 GMT

There's always a way 😉

But to better understand your question are you looking for a single number e.g. the 75th percentile equates to n dollars of revenue? Or are you looking to create a list of customers who are in the 75th percentile?

The user choice is straightforward - use a presentation variable which can be used to filter an insight.

The percentile bit may require some work but I think you're going to either 1) be using a result set to author you insight or 2) be going back to an MV where you can use pyspark SQL to do the work.

This sounds pretty interesting. If you'd like to share your specific use case and a sample set of data it'd be fun to take a run at it...

Re: Percentile Formula in Incorta Analyzer

RADSr — Wed, 29 May 2024 22:35:28 GMT

Here's a starting point ( attachments ) I started fiddling around w/ this and then decided to use copilot to create an MV to test my results ( spoiler: they are different ).

From here you can tailor your definition of "percentile" - oddly the MV using the percentile function returns the top salary as the "100"th percentile. The definition I learned and am comfortable with says there *is no* 100th percentile since percentile measures the percent of values lower than the target number.

The source file, schema export, and dashboard export are here. This was pretty off-the-cuff so forgive the naming and lack of formatting in the insights.

HTH,