Fail … but not TOO fast.

Share this post:

A critical theme of digital transformation is the drive for innovation, and rightly so.   

Many years ago this was my “thing”. I led work on creative innovation for a client with a household name … and I loved the work.
But like any other fashion, innovation easily falls into disrepute. It’s really easy to misappropriate. One popular way to drag its name through the dirt is by experimenting badly.  

A real experiment has rigour, because it proves something. That doesn’t necessarily mean statistical significance, but it at least means recognising and acknowledging when a test is not statistically significant.

There are really two options, as explained in David Rogers’ excellent “Digital Transformation Playbook”: They are Convergence and Divergence. Anything else is hokum.  

Convergence withstands scientific scrutiny: It’s important here to randomise, maintain a control group, understand the variables at play and gather quantitative feedback (just ask any scientist).

On the other hand, a Divergent experiment is informal, but can be at least as powerful: Here the emphasis is on qualitative feedback, which is channelled to drive improvements, take decisions and spawn left-field ideas.

In the “naughties” (2000-2009) our masters told us not to fail.

In the 2010s they asked us to “fail fast”—a worthy sentiment which embedded fearlessness, and the idea of learning from everything we did.

For the 2020s I’d like us to take a deep breath, and make sure we don’t fail TOO fast. If we rush at experimentation we also run the risk of experimenting badly, which is worse than not experimenting at all. Science is “not rocket science”—we can all learn the basic tenets—and if we do, we’ll deliver much better intelligence to inform executive decisions.

Executive bravado is not the same as scientific rigour, and that’s now widely understood. No longer does anyone believe that their seniors are necessarily operating with the benefit of some divine insight. If you’re still brazenly representing an indefensible level of analytical intelligence then your days are numbered: There’s a better way.

So I want to hear from anyone out there who’s volunteering to lead an evening class in elementary data science (ideally in West London). It’s today’s critical skill, but woefully under-emphasised! We now live in an era where almost every board-level decision can be helpfully informed by data, but we still fail to exploit that. I want to see that change in the 2020s.