My great passion for cooking as well as data science eventually lead me to the idea of organizing a data science team like a large restaurant that has an executive chef and a sous chef.

In a restaurant or other large operation that involves a commercial-grade kitchen, the sous chef sets the executive chef up for success by handling a lot of menial work that would otherwise occupy the executive chef’s valuable time. On cooking shows, you typically see the sous chef chopping vegetables and pre-mixing sauces while the primary chef plans and executes the meal.

Consider a similar operation on your data science team. You can obtain great efficiencies with your data science capability and improve job satisfaction for all involved by introducing a helper role for data scientists called a sous scientist.

SEE: 5 effective leadership styles for managing data scientists

What exactly does a sous scientist do?

A sous scientist is someone who helps the data scientists by taking over the necessary tasks that all data scientists must do.

For instance, it would be great if you could just dive in and start analyzing data, but that’s not how it works. Data must be prepared before it’s analyzed. Data coming from multiple data sources must be normalized, cleansed, and merged.

In a perfect world you would have ETL scripts set up to handle this, but in the real world data preparation often becomes a tedious, time-consuming job involving Perl, Excel, and/or anything else you can pull from to get the data in an analyzable state. This is very annoying part of any data scientist’s job; however, it’s a welcomed task for an apprentice who’s trying to break into data science.

How sous scientists benefit your organization

Introducing the role of sous scientists serves two purposes that benefit everyone. First, it allows your data scientists to spend their time being more strategic and analytical. This is great for the data scientists because they’d much rather study classification patterns than remove semantic duplicates from a 10 GB source data set. This is great for you and the rest of the organization because more time spent in analysis means more strategic advantage.

The role of the sous scientist also opens up the capability of entry-level data science work. This is wonderful for the sous scientist, as it gives them an opportunity to break into the data science world without carrying the weight of being a mathematical genius. Of course, the entry-level position also comes with an entry-level salary, so the organization doesn’t have to go broke trying to staff a data science team.

5 key strategies for bringing the role to life

There are five key strategies for bringing the sous scientist role into reality, starting with buy-in from your existing data scientists. As compelling as the idea sounds, do not make a move without running it by your data scientists first. Without their support, you’ve got no hope at success.

Second, agree with your data scientists that this is a mentorship structure, not indentured servitude. At some point the sous scientist will want to be a full-fledged data scientist, and there should be a well-defined career path in place that involves the support and mentorship of your incumbent data scientists.

Third, work with your data scientists on a good job description. This will be different for each organization. I’ve given you some general guidelines for what a sous scientist should do, but ultimately it’s up to you and your data science team what that specifically looks like. It’s important to establish clear boundaries between what a sous scientist is responsible for and what a data scientist is not responsible for. That’s not to say that a sous scientist won’t — on occasion — do a data scientist’s job. But, it should be very clear based on your job description why certain activities are being performed.

Fourth, make a strategic commitment to grow your overall organizational capability for data science so your sous scientists can advance. Many leaders know the frustration of being qualified for a promotion with no room at the top for advancement. Don’t let this happen on your team. When your sous scientist is ready to be a bona fide data scientist, you should be ready to accept and use that capability.

Finally, make sure you’re constantly recruiting for new sous scientists. When your sous scientist graduates to data scientist, who’s going to do the sous scientist work? You better have these details worked out. Once your data scientists get used to having sous scientists, you can’t put that genie back in the bottle. You must be sophisticated enough to allow your sous scientists to advance when they’re ready and have backfills available to fill the vacancies.


Why waste money paying expensive data scientists to do routine data preparation? Instead, learn from the culinary world and employ the data science equivalent of a sous chef in your organization: the sous scientist. They’re cheaper; they don’t mind doing the grunt work; and they allow your data scientists to focus more on strategic analytics.

Take some time today to discuss the idea with your data scientists and come up with an organizational plan that involves mentorship, career advancement, and steady growth in data science capability. Once you get this plan in motion, you’ll be cooking on all four burners.