After Henry Ford revolutionized the auto industry with his conveyer belts and assembly lines, the rush for increasing effectiveness began, and the chase for higher profits stimulated the birth of management science. In the mid-1950s, few leading manufacturers did not have an assembly line. Business owners started hunting for better managers who could drive higher productivity in their workforce.

Nowadays, most business battles are won in offices, and business processes have become so automated that companies are searching for new ways to compete. Needless to say, analytics is perceived as one of the cogs in the competitive advantage engine. However, managing data scientists has its own peculiarities that are not taught in business schools – yet.

What is different about managing data scientists? What is missing in all the books on people management?

Below is a distillation of my experience in this subject, which I call the ‘5 Rules for Managing Data Scientists’:

1. Give Them The Opportunity to Publish a Paper

First of all, data scientists are scientists who research and invent. Creative minds can generate dozens of ideas a day. Preventing these ideas from spreading is like shaking a hot bottle of champagne without opening it. Eventually the wine will force the cork out and spray the precious contents everywhere. You, as a manager, must be prepared and control this process. Most data scientists are introverts, but don’t be fooled by their seemingly calm behavior. Some of them have brains with the power of a nuclear reactor – one you don’t want to explode.

Second, to a scientist, publishing a good paper feels as rewarding as going for a two-week vacation. That’s why managers should actually stimulate data scientists to write papers. Help them to formulate their ideas so that no proprietary or sensitive information will be disclosed. Scientists who have an opportunity to publish are more satisfied and more likely to stay with your organization for a longer period of time. Another good thing: for the company, this is a free reward.

2. Quiet Office Space

As a data scientist, I know how much quiet office matters. Years ago, I turned down a job offer with better pay because of one reason: the company’s noisy open space. Consider: we spend at least a third of our lives in the office. For a scientist, whose primary job function is to think, office noise leads not just to lower effectiveness, but shortens productive lifespan. Did you know that brain tumors can grow faster when you overclock your mental activity? It doesn’t have to be that bad, however, managers should always remember that a tired scientist is just a body. When a thought process experiences constant destruction, the time that a scientist spends in the office is a waste. Companies that place data scientists in noisy open spaces should expect low returns on their investments in the workforce and higher employee turnover.

3. Form Pairs

This one took me the most time to get. Have you seen project teams where the members communicate with sparkling eyes, ignite and encourage each other, openly share ideas and are willing to work extra hours? This is one of the most wonderful and pleasant sights, especially for a manager. Books on leadership teach us that a leader can try to achieve this kind of work environment by leading by example, setting the right goals for the organization, and motivating and rewarding employees. The sad thing is that this does not necessarily work with data scientists. As I said, most of them are introverts and will rarely share their ideas with strangers. Most of them, if not all, are smarter than you (yes, be prepared for that) and can read your managerial efforts as manipulations and brainwashing.

If that’s the case, a manager can try the opposite way of building teams. Rather than picking available employees for the project, form a team around existing relationships and adjust it to the project goals. Rather than taking the employees out for an evening to break the ice, invite employees who have already formed strong social ties to be on the team.

An average data scientist typically has one friend within your organization. All you need is to recognize these relationships and cultivate them. A pair of friends will work much more effectively on a single project than they would if they are on different teams. They will exchange ideas, since they already trust each other. They will work extra hours, because they will discuss their work during out-of-office hangouts. A team of two friends is more than just a team of two – and you, as a manager, have to capitalize on that.

The pair need not be the same age or gender. Senior scientists often are willing to work with a younger person, who can be seen as their successor.

4. Don’t Skimp on Software

To me, this is ridiculous. Managers who try to save their organizational budgets by limiting software access risk losing more than they gain. Software features are rarely developed or supported for free. Developers expend a huge amount of efforts to make their software more productive. If you are not willing to pay for the software features, be ready for your data scientists to put their effort into upgrading the functionality of their tools. The budget that you will save by not buying the software will be eaten up by covering the data scientists’ extra time. Without a doubt, the saying ‘time is money’ is applicable here.

There is one more consequence of not paying for the software that data scientists need. You risk being thought of as a greedy manager who tries to please his boss at his employees’ expense. You can lose your credibility, undermine motivation, degrade team effectiveness, and so on. In the worst case scenario, you will find yourself explaining to your boss why you tried to save on software by sacrificing productivity.

5. Do Not Control Office Hours

This rule is the most obvious, and the rarest. Scientists think of their work all the time. Mendeleev invented his periodic table of elements while he was sleeping. Data scientists are no exception. They can be thinking of their work while they’re eating breakfast. If a data scientist comes late to the office, it does not automatically mean that he or she was skipping off work. Measure a data scientist’s work by results, not by hours spent in the office. Never force them to be in the office unless it is an urgent case. Create a comfortable, quite environment, equip them with the software and hardware they need, form project teams around relationships, and your subordinates will rush to the workplace.

 

My actual list of recommendations for data science managers is much longer. My message here is that data science managers should not merely tend herds by following the rules in classic business books, but by starting to write new ones.