No Data Scientist is an Island: Thoughts on Effective Communication

As data scientists our job often resembles that of a consultant: we build models to answer questions for business owners whose expertise is different than our own.  Though we jump at the opportunity to learn new algorithms, tools, and languages—all integral to the role—there is one consulting-related aspect of the role in which we may find ourselves falling a little short, and in the process contributing to a not-unjustified “ivory tower” view of data scientists.  That aspect is effective communication.

After graduating from Belmont University in Nashville, I was fortunate to land a job in consulting working for economist Dr. Arthur Laffer, of “Laffer Curve” fame (or infamy, depending on one’s view).  I expected to spend most of my time researching arcane economic policy and writing pithy research papers, but I was surprised to learn that one of the most important tasks for a consultant was communicating research methods and results. This proved to be a challenge.

Early on, I struggled to bridge the gap between my preference for an academic understanding of research topics and the clients’ desire for clear, concise answers to their economic questions. I often spoke in the jargon of the typical economist, mixed with healthy doses of “but on the other hand.” My approach often fell short, and it became obvious why as I watched Dr. Laffer communicate with clients in a very different way.  Unlike most economists he is plainspoken—even folksy—when talking economics, which has made him remarkably effective at communicating not just to fellow academics, but more importantly to the other 99.9% of people who help shape the economy—those in government, clients, students, and the public.

And that is the value of effective communication for data scientists:  we need to embrace the power of communication if our work is to have influence beyond our teams and throughout our companies.  Technical expertise alone is insufficient to achieve success.  The best model in the world is useless if it is not actionable and implementable, and implementation requires consensus building and co-operation with our business owners—both very difficult to achieve without effective communication. 

Towards that end, here are a few simple (but not easy) principles I try to continually develop, all of which will help ensure a successful project:


This is especially important at the start of a new project.  The business owner is going to have unique domain knowledge that will be vital to a successful project.  They likely understand the power and limitations of existing data and know what has worked in the past and what hasn’t.  Also, consider that some crucial variables may not be captured in existing data or models.  By listening to your business owners, you may be able to improve existing data collection and modelling methods.  We often find our most successful projects come from taking business owner insight and delivering them a solution they hadn’t yet thought of.

A good data scientist tries to be keenly aware of where her knowledge and experience fall short.  Listening to business owners also serves as a good check against any potential whack-a-doo explanations offered by our models that can result from our ignorance in new subject matter areas.

Educate your business owners

Communicating our work with a non-technical audience may be the most challenging aspect of our job.  In this instance our technical skills may be a barrier to success.  Avoid getting in the weeds on new material and stick to clear, concise explanations or use analogies to explain complex topics at the outset.  Later on there will be plenty of room for “deep dives” (note that the same is true for us as we enter a new domain that is far better understood by the business owner). 

If we’re not getting our point across, the absolute worst thing we can do is blame it on our audience.  We should reflect on our word choice and our approach and try a different approach.  When in doubt, ask the audience how they prefer to take in new material; some prefer conversation and interaction, while others do better with an essay or PowerPoint and prefer to digest material on their own.   

Lastly, throughout the course of a project, the business owner may have questions about our methods and models.  We shouldn’t take offense to these inquiries, but rather see them as an invitation to build trust in our work and a stronger business relationship.  Don’t miss this opportunity.

Avoid jargon

I admit that jargon can be fun to use—it signals subject matter expertise and group identity, and makes for occasionally-somewhat-sort-of-funny inside jokes with our peers.  And data scientists are human and enjoy these things too and that’s perfectly fine when we’re in the company of other data scientists.  But when our desire for group identity or status gets in the way of communicating the purpose of our models—solving business problems—our use of jargon becomes a problem.

Heteroscedasticity… confidence intervals… multi-collinearity… these terms are meaningful to data scientists but mean little if nothing to the majority of non-data scientists.  Understanding them is helpful to data scientists in the modeling process, but throwing out multi-collinearity in an effort to communicate model results to senior leadership may keep you from getting your point across. 

The same can be said for acronyms—unless you know your audience will understand them, an initial time-saver will end up taking more time as you’ll probably need to spell out what the acronym means.  Read the room—if your audience shares your technical background, throw a little jargon and some acronyms in.  If not, avoid them—usage may even be seen as ivory tower behavior, causing tension and limiting the potential for success.

Speak/write with intent

Give consideration to how you approach communication with business owners or colleagues in general.  How many times have we misunderstood and taken offense to an email, only later to learn the author’s purpose was far different and more positive than we’d imagined? Putting a little extra thought into communication can avoid these situations.

Whether speaking or writing an email, address the audience with an appropriate level of respect and formality and be mindful of your tone.  Determine if the combination of the subject and audience requires only a brief, to-the-point message: if so, write it and fire away.  But if your email is touching on something sensitive or is intended for senior leadership, slow down and give extra thought to your word choice and tone.  Ask a trusted colleague to check for errors and give feedback.  Read it and re-read it before clicking send. 

Putting extra thought into your communication will help your project achieve success; and more importantly, build relationships and trust with your peers and leadership.