Analytics Zen - The Art of Mindful Analytics

A while back while doing a presentation for some executives, I came across a methodology for creating presentations called Presentation Zen. While there are a number of principles associated with it, three outcomes really resonated with me. These are:

• It’s all about telling a story, and the message from a presentation must be accessible and useful for the audience.
• The structure needs to be solid.
• Remove the non-essential - simple is better.

I realized that these are the same principles I try to apply to my analytics work. I’ve often spoken about how analytics needs to be decision-useful, and this is coming across more often now. I’m starting to see lots of analytics showing insights into the pandemic. While there are many beautiful graphs with fascinating datasets, I often find myself asking ‘So what?’

I, therefore, believe that many data scientists should try to adopt these ‘Presentation Zen’ principles into their analytics projects. I’ve tried to adapt them below, in what I call ‘Analytics Zen’.

Analytics Zen Principles

1. The first principle is simplicity. A data analytics model should not try to solve every problem for every person - it should be focused on addressing one or two key problems comprehensively.
2. The second principle is that of being decision-useful. Analytics should enable a user to take some form of action after looking at the outcomes. This means that the model outcomes need to be carefully thought out and designed. Will the outcomes be shown in a dashboard? Perhaps an alert will be emailed to certain individuals? 
3. The third principle is to have a clear theme. I find a lot of analytics encompasses multiple themes - pick what theme the analytics will address. By this I mean is the analytics for cost reduction or revenue maximization? Is the intended purpose for internal analysis, or to take to clients?
4. The fourth principle is to become technology agnostic. I won’t elaborate on this point - there are too many emerging technologies for data scientists to become too attached - that said never underestimate being an expert in one tool or language!
5. Finally, the fifth principle is data quality. Remember the importance of data quality - you can have the best model and dashboard, but show dodgy data once and neither your results nor your model will be trusted.

I hope that these principles will help to align some of the thinking amongst data scientists - and help some exceptional data scientists to transcend to the next level of analytics - Analytics Zen.