The Wizard of Oz was a favorite childhood movie of mine, and one of my favorite characters was the Cowardly Lion. As a young child, I liked him because he was animated and goofy; as I got older, I was intrigued at the paradox of lacking courage to obtain courage. It’s a self-supporting dysfunction.

Many data scientists show the same absence of courage. Fortunately, there is something you can do to mitigate this expensive problem.

Aha, that’s what’s missing

Courage is one of those unexpected ingredients of data science that usually becomes an unpleasant aha moment for leaders in the middle of strategy implementation. Courage is the fuel for experimentation and, without experimentation, there is no data science.

But it’s more than experimentation in the scientific sense. To succeed, programmers need to take risks with code; mathematicians need to reinvent math; and data artists need to color outside of the lines. Intellectually, data scientists know this, but their personality doesn’t generally support taking risks, so they don’t.

One of my college professors said that MBA graduates don’t start their own business because they know better — the same theory follows with data scientists. They have the unique ability to calculate the risk of taking any route, and they’re smart enough to know when something probably won’t work. Data scientists also hate it when something they try doesn’t work (especially when there are others around to witness them get the wrong answer), so they don’t attempt to answer the question. You need to neutralize the downside.

The power of undo

Lack of courage (or fear) is the anxiety attached to an anticipated downside. People who are afraid of heights aren’t literally afraid of altitude — they’re afraid of what might happen if they suddenly lose altitude. For a data scientist, there’s a psychological fear of being wrong; however, there are some fundamental, tactical fears that must be addressed first — the strongest of which is losing work. You must install a failsafe mechanism that I’ll generally classify as the power of undo.

The power of undo comes in many forms and will be specific to your team, though here is some general advice: To mitigate the fear of losing work, you must work with your team on ways to step back when experimentation goes wrong. This means getting serious about testing infrastructure and software change control. You should always have automated methods of testing system. You should also build functionality in small increments, so the most recent working version is from yesterday rather than six months ago. The confidence of having yesterday’s build as a failsafe goes a long way in mustering the courage to experiment.

The courage to be courageous

Although the power of undo is very important, it pales in comparison to consequences that reinforce experimentation and all the other behaviors that demonstrate courage. This is where most leaders get it wrong.

Leaders spend most of their time and energy in setup (change control system, automated test harness, team building) and then very little time in reinforcement mechanisms. And yet, research shows that only 20% of behavioral influence comes from the antecedent (i.e., what prompts the behavior) and 80% comes from the consequence (i.e., what reinforces the behavior). So spend some upfront time on the power of undo, but then focus the majority of your effort on reinforcing courageous behavior.

To do that, be clear about the behaviors you expect. Remember, courage cannot be directly observed, but the behaviors that demonstrate courage can.

Is your team experimenting or playing it safe? Are they attempting bold leaps in functionality, or are they inching their way to the solution? Are they restoring from previous builds? Normally, reverting to a previous build has a negative effect, so it’s important that you flip this around by making it a positive effect.

Instead of criticizing the team for wasting time on an idea that didn’t work out, celebrate them for making a bold move. Don’t forget that once in a while those bold moves pay off and wind up saving months of development time.


Courage is unnatural for a data scientist, let alone a team of data scientists, so it must be nurtured and developed. It took a special medallion from the Wizard of Oz for the Cowardly Lion to find his courage. Your medallion of courage is the power of undo and the positive consequences that reinforce risky behaviors. Offer this to your data science team, and welcome the brave new world.