1. The role of a Data Scientist is becoming increasingly common, and continued growth seems likely, given the impact professionals in these roles are having on companies.  The presence of a data scientist in a company presents a new challenge to executives: how can one best interact with their data scientist, and what expectations should one have on their output?

    Remember data scientists are generally doing research

    Unlike a software engineer who can often estimate upfront exactly how difficult and time consuming a task might be, a data scientist is usually embarking into the unknown.  Can a company’s margins be improved?  Possibly, but there are no direct guarantees.  It is typical for a data scientist to conduct exploratory work before committing to any results.  Give your data scientist time to do so.

    When interacting with a data scientist, challenge them with a clear problem

    A good data scientist has the right toolbox of statistics, computer science, decision theory, and similar disciplines.  They should be equipped to solve challenging, non-trivial problems.  However, those problems should be expressed in terms of the metric you want to improve, not how the improvement should be achieved.  Define your problem and all it’s constraints and allow your data scientist to find a solution.

    Be realistic in your expectations

    Commoditized hardware and cheap storage have resulted in many companies saving vast amounts of data.  Yet, data is only useful if it helps inform a decision you need to make.  Data for data’s sake has no value.  Don’t send your data scientist on a “fishing expedition” to see what nuggets of insight they might uncover.  Further, keep in mind that data science, while powerful, is not magic.  There are limits to what can be accomplished.  Present a clear, measurable direction, and let the data scientist determine what’s possible and how to extract the right solution from the data.

    Ask questions and don’t be intimidated by jargon

    For data scientists that uses concepts like “multivariate logistic regression” on a regular basis, terminology and jargon can easily find their way into everyday conversation.  Sometimes this so good – an organization can grow as a whole by understanding concepts and abstractions and adopting them.  Other times this alienates and confuses listeners.  While the best data scientists need to know their audiences and communicate appropriately, when they fail to do so, don’t be afraid to ask questions.  The best data scientists are passionate about what they do and would be thrilled to have the opportunity to explain their work in greater detail.

    Be willing to consider radical solutions

    Many businesses fear change.  Many organizations also exhibit superstitious behavior about their operational processes.  An objective deep analysis might uncover surprising facts that suggest counter-intuitive strategies.  For example, in some situations, discontinuing a profitable product might provide an aggregate increase in overall revenue for a company.  Allow your data scientist to present unconventional ideas and make sure you understand the logic behind their suggestions.

    Trust, but verify

    When done correctly, data science is *science*.  Therefore it will have three important properties.  It will be consistent with available historical evidence; it can make confirmable predictions about the future; it can be falsified.  You should feel encouraged to work with your data scientist to confirm and test their efforts.  Forecast results, and revisit predictions in the future.  After all, the truth has nothing to fear from scrutiny.

Leave a Reply