Analytic Systems are No Good Unless They Are CRAP

Has anyone ever described your analytic system as 'crap'? You may not have been aware of it at the time, but that person actually paid you a compliment. Analytic systems of the highest quality should be CRAP. There are four reasons why this is so:

First, analytic systems should be Creative. Not in the sense that the functional requirements were not met, but rather innovative in development approach. By either using new tools or old tools in new ways, creativity keeps work interesting and advances the analytics discipline. As long as the customer's welfare is not put at risk in terms of cost, schedule or scope, then thinking outside of the box benefits managers, analysts, and end users. The goal is to push products to their limitations and then seek ways of integrating with other products. For example, here is a SUGI paper from 2005, where I integrated SAS with earned value techniques and email services.

Second, analytic systems should be Responsive. Users ought to be able to feed inputs and receive outputs in a timely and user-friendly fashion. Ideally, analytics must accommodate a wide spectrum of end user profiles – local/remote, new/experienced, American Disabilities Act eligible/non-ADA eligible -- whatever is required. Even in the area of language, it is not unusual to see commercial applications (more so ATMs and social media) offer the user the choice of language. For example, here are two examples of how to quickly notify end users when jobs have catastrophic outcomes.

Third, analytic systems should be Adaptable. High quality analytic systems should be scalable, platform-flexible, and easy to maintain. Hence, they should be designed with growth in mind. One of the reasons why SAS has been so successful is the ability to function across platforms and to integrate other products like Microsoft Add-Ins.

Fourth, analytic systems should facilitate Productivity. Analytic systems that do not increase the amount of work accomplished within a given period are not worth the time it took to develop them. If end users pass through the learning curve and still are not productive, then valuable time was spent learning to use a tool that did not yield an increase for work performed. Even if data scientists are great coders, their time would be best spent analyzing and reporting and not coding. SAS Enterprise Guide is an example of how to leap-frog coding to interpreting results.

Therefore, in that light, I suggest that all analytic systems should be CRAP -- Creative, Responsive, Adaptable, and Productive. If you embrace this vision, then please join me in singing the CRAP Anthem of Analytics Quality:

CRAP (Sung to the tune of Bad by Michael Jackson)

Baseline Introduction

Boom-boom-boom-ba-boom boom ... Boom-boom-boom-ba-boom boom

Verse

Creative is the golden rule,

Responsive and Adaptable

To what suits you.

Productivity - is Job One!

You will have CRAP when I am done.

Refrain

I'll always keep my word, and

I'll always promise that,

When I build analytics,

I'll always give you CRAP!

Chorus

I'll give you CRAP, it's CRAP, it's really, really CRAP

I said it's CRAP, it's CRAP, it's really, really CRAP

You know it's CRAP, it's CRAP, it's really, really CRAP

It's Creative, Responsive, Adaptable

And Productive too - It's CRAP.