Capturing 95th Percentile from Performance Analytics

sgmj
Kilo Explorer

3 hours ago

Hi Team,

 

I am using service now as my case management system. Where I have a field named duration which stores the case duration.

 

I am working on a project that I need to identify the trend of the case durations by plotting trend line as a dashboard widget. 

Example from the widget I should be able to see the average duration for month of January, February etc.

To get a fair representation I am looking to exclude the duration of the edge cases  by removing the top 5% time consumed cases. 

 

Since I need a widget and trend lines I was planning to use the performance analytics and PA scripts. But I think that is not feasible as there is not straightforward way to get the average of the duration with filtering the top 5%. I was able to use the script includes but the issue with that is making trend lines from that is hard.

 

So is there any best practices that I can follow here? Or maybe a way to implement this?

 

If I have the ability to add a new function like Average, Sum, Count in performance analytics I think that is the most suitable approach.

 

Appreciate your inputs on this.

 

Please note that I am using the service now Washington DC version.

0 Helpfuls
  • 113 Views
1 REPLY 1

Matthew_13
Kilo Sage

an hour ago

I think your on the right track my Friend!, and PA is still the right tool for this. You just can’t use the out-of-box Average.

The best approach I’ve used is a scripted Performance Analytics Indicator that calculates a trimmed average at collection time.

At each monthly run, the script:

  • Pulls cases closed in that month

  • Collects their duration values

  • Sorts them

  • Drops the top 5% longest cases

  • Averages the remaining 95%

  • Stores that single value as the PA score

Once PA stores one value per month, trend lines and dashboard widgets work automatically. This is much cleaner than trying to calculate averages in a widget or Script Include on the fly.

You can’t really add new PA aggregation functions, but scripted indicators are the supported way to handle custom math like this in Washington DC.

One quick sanity check: if the business is open to it, median duration sometimes works just as well and avoids trimming logic but if they explicitly want exclude top 5%, the scripted indicator is the way to go.

 

@sgmj - Please mark Accepted Solution and Thumbs Up if you find Helpful!!

0 Helpfuls