By Roger W. Hoerl, Diego Kuonen and Tom Redman
The best data scientists are not defined by technical sophistication alone, but by how they work with others to improve business results
Despite major investments in data, business intelligence, analytics, and (generative) AI, most organisations struggle to translate these capabilities into enduring value. Initiatives fail for many reasons: problems are poorly defined, solutions rely on complex analyses when simpler ones are better, internal politics intrude, and users are poor decision-makers anyway.
Fortunately, a terrific partnership between business leaders and great data scientists can do much to resolve these issues. Unfortunately, most business leaders have not met a truly great one. They don’t know what one looks like, what they do differently, or what to expect. We sympathize. We have seen hundreds, perhaps thousands, of data analysts, statisticians, and data scientists in action. Most were technically competent; few were truly effective.
This article aims to help business leaders understand what they should look for in, and expect of, a truly great data scientist versus those who possess outstanding technical skills but still miss the mark. Five traits stand out: they help crystallize the problems to be solved; they see things others don’t; they follow the evidence, not the hype; they bridge technical and non-technical worlds; and they help create enduring value. In short, they are accountable for results.
What Should Senior Leaders Expect in Practice?
1. Help Clarify the Problem
Tempting though it may be, the best data scientists do not dive into data without first obtaining agreement on the problem to be solved. They see their roles as helping clarify that specific business problem in the face of competing perspectives.
To illustrate, one day a business person asked a data scientist to calculate a sample size for them. It is a simple enough question, with a Statistics 101 answer. But this data scientist probed more deeply, asking why they needed a sample size. The answer involved seeking additional budget for an experiment. Further probing revealed that the business person didn’t really know enough about the process they wished to study to design a good experiment. Developing that understanding was a necessary first step.
Note the progression:
Need a Sample size → Need Budget → Don’t know enough
This illustrates a recurrent pattern and one way the best data scientists bring value, as billions are wasted solving the wrong problem.
Senior leaders should expect any data scientist to solve the problem they are given.
They should expect the great ones to help clarify and come to agreement on the real problem, and understand the decision to be supported and its overall business context.
2. See things others don’t
Great data scientists have a nose for “things that just don’t smell right,” including quality issues, methodological errors, and results that don’t make sense. One example involves the data scientist working at an investment bank. In reading through an important report, he spotted an error in a returns’ calculation. Digging deeper, he recognized that what should have been treated as a multiplicative relationship was treated as an additive one. A simple, but potentially embarrassing mistake. What is most important in the story is that hundreds of people had read the report. But this data scientist was the only one who spotted it!
While senior leaders should expect all data scientists to select the right statistical, analytical, or AI techniques, the best bring value by seeing things that others don’t. They may call out the need for a quality program to support an AI initiative, recognize that decision makers need training, or recommend caution when something just doesn’t look right.
3. Fair Brokers
Outstanding data scientists act as fair brokers, following the evidence where it leads.
To illustrate, one was a member of a team whose direction was “fix our flagship product.” This direction came from the CEO, so the team felt considerable pressure to find some critical flaw and fix it. However, extensive data analysis revealed that:
- The positioning of the brand in the “mid-quality range” was consistent with marketing strategy.
- The key product features were also in the mid-quality range and had been relatively consistent for years.
- Nothing suggested the market had shifted.
In short, nothing needed fixing. Reluctantly, the team leader honestly reported the results. Eventually they learned that the origin of the project was that the CEO’s spouse hated the product. An off-hand comment had been translated into “fix the problem.” By standing up, the team saved the company from an expensive and possibly disastrous intervention.
Business leaders should expect all data scientists to be completely truthful and intellectually honest, about assumptions, limitations, uncertainties, and risks. They should expect the great ones to show great courage in doing so, even in the face of enormous political pressure.
Said differently, a great data scientist should occasionally make a senior leader uncomfortable. When the evidence is weak, the uncertainty large, or the favoured interpretation unsupported, saying so is part of the job.
4. Expect Them to Bridge Worlds
It is easy for data scientists to sit comfortably in their offices, making incremental improvements to their models. The best ones do their most important work outside their offices.
One did so while serving as project manager for the development of a cross-government agencies’ Data Science Strategy. An early step involved developing a shared vocabulary, and first on this list was an understanding and operational definition of “data science”. Without a shared vocabulary, people talk past one another, think they are in agreement when they are not, and fail to spot opportunities to work together. This data scientist insisted the group make the needed upfront investment. Once that shared vocabulary was established, it became the foundation for bridging organisational and disciplinary boundaries, breaking down silos, and speeding up collaboration.
Outstanding data science sits at the uncomfortable nexus of data, technology, domain expertise, organisational priorities, politics, policy, and society. All data scientists can contribute, while only the best ones can help build the bridges needed to pull everyone together.
5. Improve Decisions and Create Enduring Value
Understood properly, data science is about improving decisions, and creating enduring value.
A great example of this involves the data scientist who was asked to support senior leaders in deciding how robust their communications network needed to be. Such problems can become quite technical, involving questions such as, “what fraction of requests is it okay not to address within X seconds, during the busiest parts of the day?”
Rather than stepping the executives through the dizzying array of choices, this data scientist crystallized the problem and simplified their options as follows, “The first thing we must decide is what kind of network we want: a ‘baby bear,’ a ‘mama bear,’ or a ‘papa bear’ network.
This sort of work adds value by moving the frame from one decision makers don’t understand to one they do, where they can ask meaningful questions, and make informed choices.
Senior leaders should expect data scientists to take responsibility for monitoring and continuously improving their analytical solutions. They should expect the great ones to build trust, help leaders make better decisions, and improve decision-making throughout the organization to create value that lasts.
Leaders Should Demand More!
Other than dystopia, there is no foreseeable future in which data, data science, and AI don’t play larger roles in companies, government agencies, and non-profits. Organizations have responded, hiring increasing numbers of data scientists and upgrading their roles. Still, few have reaped the benefits they should.
Data scientists may complain they don’t have a seat at the table. But that seat cannot rest on technical expertise alone. It must be earned by building trust and real results. Business executives should demand data scientists step up to the traits described here. These stories illustrate how five data scientists have done so. No magic needed.
Roger W. Hoerl
Diego Kuonen
Tom Redman
