Let’s talk today about the most important soft skill you can grow, one that benefits you no matter how deeply into data science you dive, change management. Our field would be simple if people changed how they do something just because they were presented with facts and analysis. Instead, if you want your work to change your organization – that is, if you want your work to be used – there are steps you can take to make this more likely.

Change is work! And when you’re working on a data science project you’re eventually going to ask people to change something. You’re going to ask them to do work.

Think of change as a hill with the outcome you’re changing at the top. You’re leading others on a hike up this hill, but they don’t want to take the hike. It’s nice at the bottom and being at the bottom has worked well enough so far and besides who are you to make them hike. Your job is to get them there anyway. Sounds fun, right!

Sure you can cajole them, you can get behind them and push or you could call their bosses and demand that they hike. And sometimes those approaches will work but over time you will grow irritated about begging, exhausted from pushing and your colleagues will bristle if you keep going over their head. There’s an different way – make the hill smaller

How does that work? Well, of course since this is all an over-extended metaphor the hill is just the perception of your stakeholder. Climbing a hill is work; changing how something is done is also work. So you need to make that work seem smaller. And three factors determine the perception of that work: Control, Understanding and Timing.

Three Components of Successful Change Management

Control
The truth is that no one likes change a change that’s forced upon them. Making a New Year’s resolution to get in shape feels very different than your doctor telling you to start exercising. Even people who say they thrive on change really mean they thrive on change they’re able to control.

Add to this the fact that in many organizations Data Science is a newer function. This means that most data science projects try to influence processes or departments who already have a way they do their work. Which is to say, process or departments that already have owners over them. If the results of your project depend on those owners doing something differently, you want to include them.

This makes your project stronger because you’re bringing in perspectives you don’t have. In Data Science it’s easy to convince ourselves that the data will tell the story all by itself. And if we believe that then we can build our models by ourselves. It’s much easier and faster that way, at least at first. But you will eventually be slowed down when the owners of the process push back on your assumptions or methods.

So take time to understand these owners and bring them into your work. It may be obvious who to include or it may require building a process map to see all the touchpoints. In either case if you push ahead on your project and don’t consult with the groups that manage the current state until you’re done it’s unlikely you’ll get a receptive audience at the end.

Understanding
While people dislike change they don’t control they dislike it even more when that change is something they don’t understand. Imagine being told that your company has moved your office to a new building in a different part of town, one that’s farther from your home. If that’s all you’re told you would likely be annoyed. Not only will you have to pack your desk but your daily commute is longer and that costs more gas and and time and so on.

Instead, imagine your manager explains that the new building has a lower rent so that the company will be more successful and grow and your bonus will be better. Or that the old building has failed a recent safety inspection and your company doesn’t feel it’s safe enough. The exact reason is less important the effort being made to help you understand.

This issue is especially acute in the data science world. Too many data science projects are presented as a black box or a gray box where the approach is explained, but in such a technical manner that it’s not accessible to the intended audience. Too often the explanation is some akin to ‘data goes into the model and results come out’. Followed by we know the results are legitimate because we applied statistical techniques on them and so we know. What’s really being said it ‘trust us’.

I’m not implying you’ll need to deliver a course in statistics but you do have to find a way to open the black box and explain, in simple terms, how your project is doing its work. Simplicity is better. Analogies are useful too but not always applicable. The best way to help someone understand your project is to communicate frequently, which is the third way we lower the hill.

Timing
This is the easiest factor to influence as long as you plan for it. Talking about your work and how it will impact someone gets them comfortable with the upcoming change.

Keeping with the example above about moving offices your reaction to the news will be different if you’re told the office is moving in a few months as opposed to moving that afternoon. With a few months of notice you can start to look at the nearby restaurants where you can eat lunch and plan the best route to make the commute better. It gives you time to imagine the new future, which has the affect of reducing the disruption when the change happens because it’s partially already happened in your mind.

Data science is the same way. When you give your stakeholders time to think about the change what you’re actually doing is giving them time to imagine it. That leads to time to ask questions (Understanding) and the conversations that those questions create engages your stakeholders in the project (Control).

Wrapping It All Up

Learning how to manage change in your career will help ensure your work is implemented and used. In other words, it will mean your works matters because unused data science might as well be undone data science.

Your goal in change management is to reduce the perception of how much work it’s going to take to make this change. By including those affected in your decisions, making an effort to help them understand your work and doing this throughout the project you’ll have a much easier time getting your organization to adopt your results.