The Most Important Key Performance Indicator for High Performing Data Science Teams

This post is the first installment of a series that shares some best practices for cultivating data science talent and creating high performing data science teams.

Don't Measure Work. Measure Power Output.

You may recall from elementary physics that power is defined as the amount of work performed per unit time. For example, if you do more work in the same amount of time, or if you do the same amount of work in less time, then your power output increases. Conversely, performing less work in the same amount of time or the same amount of work over more time produces a lower power output. Tracking work over time provides the fundamental basis for virtually all key performance indicators, but you're still left with the task of defining and quantifying a unit of work.

Power is defined as the amount of work performed per unit time.

In data science, I've found the best abstraction for quantifying a unit of work to be an experiment, and it follows that the most important key performance indicator for data science is the number of experiments performed per unit time. In a future post, I'll more rigorously define and experiment and outline some best practices for conducting effective experimentation; for now, let's just define an experiment as a repeatable process that tests a hypothesis.

The most important key performance indicator for data science is the number of experiments performed per unit time.

Inextricably linked to this key performance indicator is the notion that repeatability yields collaboration. In other words, the easier it is to repeat an experiment, the more likely others will participate in the scientific process with you, the more hypotheses you can test, the faster you can explore complex decision trees, and the better you can estimate for complex sets of variables in optimization problems.

Repeatability yields collaboration.

There's a true beauty in this kind of simplicity. Openly and transparently tracking experiments per unit time is a a powerful agent for creating high performing data science teams with a culture that's focused on producing real results that translate into business value in record time.

If you've found this post useful, please share it with a colleague, and be sure to leave a comment if there's anything you'd like for me to cover in future posts.