Statisticians and Data Scientists Have Entered Their Leadership Era

By Adam Barone, ASA Senior Editor (Contract)

Charismatic confidence has value in leadership, and for most of history, leaders were rewarded for projecting certainty that reinforced their intuition, authority, or subject matter expertise. But leadership that depends solely on these qualities is quickly becoming outmoded by an increasing need for scientific rigor to manage the explosion of data required for decision-making—a development that’s opened the door for many statisticians and data scientists to step beyond the numbers and into a new kind of role: strategic leader.

A new study from the National Academies, Frontiers of Statistics in Science and Engineering: 2035 and Beyond, notes, “With the influx of computing capacity, proliferation and interoperability of data, and availability of increasingly complex modeling, the statistical landscape is shifting in ways that will transform its demands and broaden its contributions.”

Indeed, the world is trending away from leaders with all the answers and toward those who can answer a single question: “What do we do with all this data?”

If anything, more complex data environments generate new questions that often reshape the landscape of possibilities—a journey that requires statistical skillsets to navigate. In this respect, one of the biggest challenges statistical leaders now face is transforming data into trustworthy insights that make sense to everyone in the room. And this often requires posing uncomfortable questions that statisticians are uniquely qualified to ask: “What does the data say? What are we missing? How confident should we be? What happens if we’re wrong?”

The real tests of leadership come when uncertainty is high, the stakes are enormous, and there’s no playbook to follow. In those moments, statistical leadership reveals its greatest value, and few examples illustrate this better than ASA Fellow and Brigham Young University President Shane Reese’s experience in navigating the COVID-19 pandemic.

Inventing Datasets to Combat a New Threat

In 2020, as academic vice president, Reese faced what would become perhaps the biggest challenge of his career as a statistical leader—sorting through a flood of conflicting, rapidly changing data to discern how the university would respond to the COVID-19 pandemic.

As a private institution, BYU retained authority to decide its own response to COVID. And instead of complying with the conventional wisdom of the moment, Reese reverted to his statistical instincts as he and his team scrutinized models that were predicting catastrophic numbers of student and employee deaths.

“I was literally blown away that one group would say to do this, and yet as a super heavy consumer of data, I just didn’t see it in the data,” he said.

And when the data they wanted wasn’t available, Reese and colleagues devised novel data streams, such as monitoring levels of SARS-CoV-2 in wastewater, which provided aggregate, anonymous early-warning signals on infection trends.

“It was an awful time, but also a remarkable time to be a statistician trying to navigate something that had huge uncertainty. I couldn’t believe how much high-throughput data I was consuming to try to make every decision,” he said. “We didn’t just do what other people were saying to do or rely on conventional wisdom. The truth is we didn’t have conventional wisdom. No one had any experience to draw on for a situation like this. We were without an anchor, but we felt like we could trust data.”

In the end, the BYU leadership team used this novel data to find a suitable balance between caution and continuity of in-person learning. For Reese, the experience reinforced the power of statistical thinking for making decisions when uncertainty is high and the stakes are higher.

“Sometimes data is so clear that it doesn’t require any inference. And sometimes data itself has enough uncertainty that it requires inference and scrutiny and all the things that statisticians are trained to do,” Reese said. “When we didn’t know something, we designed a way of collecting data that would inform that question. It wasn’t just the data streams coming in. It was the formulation of questions that could be answered by collecting new data. And we were literally inventing new streams of data daily during the height of the pandemic.”

The types of data challenges Reese encountered are hardly unique. Across academia and industry, leaders increasingly find themselves confronting an overabundance of information with no clear way to deal with it. As data volumes grow and analytical tools become more powerful, organizations equipped to act with confidence that comes from statistical rigor have an advantage. Expertise still matters, but it must be paired with the ability to interrogate evidence, challenge assumptions, and make sound decisions when answers are incomplete or conflicting—like ASA Fellow Yannis Jemiai has done in the pharmaceutical industry.

Shifting Away from Traditional Modes of Decision-Making

In pharma, clinical development has traditionally been led by clinical professionals, who often come with “a certain authoritative self-confidence,” says Jemiai, the former chief scientific officer at Cytel, where he recently concluded 20 years of service. In the past, he says, statisticians essentially just executed the orders of clinicians.

But that dynamic has been changing. Jemiai points to the explosion in data volume and complexity, along with advances in computing and AI, as key drivers. With AI tools now automating routine programming tasks such as tables, listings, and data management, Jemiai says the need for statistical expertise and thinking has only increased, with a greater need to interrogate assumptions, quantify uncertainty, identify bias, and consider ethical implications.

“There’s this tension between expertise and application of data-driven methods. Often, the clinical experts believe they can trust their gut instincts, their intuition, and they have to be proven wrong,” he said. “Over the years, there’s been a recognition that statistics and data analysis can be powerful if properly thought out, and I think it’s been an objective of statisticians to try and resensitize that and teach that to the clinical professionals and team. I think as methods get more complex, it becomes more and more of a need to have a statistician/data scientist to wade into the data and make heads or tails of what was going on.”

Ironically, the growing sophistication of technology may be pulling statisticians back toward the center of strategic decision-making. Jemiai describes a decades-long division of labor in which programming, data management, and statistical analysis became increasingly specialized. But with AI now automating much of that work, statisticians are regaining direct access to the data itself—a shift that may further elevate their strategic role in organizations.

Jemiai argues that statisticians possess a unique ability to test assumptions before organizations commit to them. When leaders disagree with data-driven recommendations, his approach is often to model their reasoning rather than dismiss it. By simulating various scenarios—including optimistic, pessimistic, and most-likely outcomes—statisticians can help other decision-makers visualize the consequences of different strategies before implementing them.

“I think that’s a strength that statisticians have,” he said. “We can try and model somebody’s behavior and then sort of simulate what the outcome would be, so they can see it for themselves. It’s becoming easier and easier for the statistician to pull back a lot of these functions that they used to do themselves and just have an AI agent do it for them,” he said. “It’ll be interesting to see how the role of the statistician evolves over time with this new technology.”

For Jemiai, the biggest challenge is not producing an analysis, it’s distilling it. Modern organizations generate more information than any executive team can reasonably absorb. Therefore, statisticians must step in to distill data into a small number of actionable choices. He advocates presenting leaders with a handful of viable options, a recommendation, and the evidence supporting it.

“Statistics is one of those disciplines that lends itself to a lot of debate quite easily because it’s not pure mathematics, where there’s certainty that one plus one equals two. It’s much more a question of interpretation,” he added. “The problem is that in our field of application, there’s often an interest in understanding the mechanism of action or the way things are impacting the outcome. A black box is not really what people are looking for. So, if you have a very good prediction model, but you can’t explain what’s going on, it’s not very satisfying to others in the room.”

Indeed, organizations run on people. And no matter how sophisticated an analysis, its organizational value is limited if stakeholders can’t understand and trust it. This reality creates a different kind of challenge: becoming a good communicator.

Bridging the Gap: Communication Skills

Good communication is one of the most consistently cited skills that enables a statistician or data scientist to move from technical expert to strategic leader. ASA Fellow Hal Stern—provost, executive vice chancellor, and distinguished professor of statistics at the University of California at Irvine—recalled a pivotal moment that reinforced this truth early in his career.

“I’ll never forget when I became a department chair, there was a workshop where a consultant came in and told us that if you’re in charge of a group, you need to learn how those people communicate,” he said. “You can’t just send everyone an email and assume it works. If you have a colleague down the hall that doesn’t communicate well that way, you’ll probably need to walk down there and say, ‘Hey, we need to make this decision. Here’s how I’m thinking about it. Let me know if you have any different points of view.’ Even today, I find all the time that people rely too much on email. When I get in front of groups, I often say, ‘If you go back twice and you didn’t get the answer, you should talk to the person.’”

Stern added that transparency is equally important. “Closely related, but different, is demonstrating transparency in your decision-making. I feel it’s always important. I go to the deans all the time to say, ‘We need to figure out how we’re going to do X, Y, or Z. I’m thinking about it this way. Let me know what you think.’ And people get a chance to weigh in.”

Reese highlights learning a process for writing based on rigorous revision and feedback as key to developing his communication skills. “I continue to be grateful for a technical writing teacher here at BYU and her willingness to teach me this idea of revision as art in writing—that it’s not about your raw ability—because I just didn’t have much raw ability,” said Reese. “But I learned I could grind out a paper that wasn’t too bad by writing, revising, and seeking lots of feedback and input on my writing. She taught me that in a powerful way, and it’s a skill I wish I’d learned sooner because I wouldn’t have felt so bad about my writing.”

Jemiai contends that effective communication in leadership also requires persuasion and contextual awareness. “A good statistician will understand that greater context and understand when to push for a particular statistical argument because it’s relevant,” he said, “versus, okay, this goes beyond statistics—decisions are being made in a different way according to different criteria.”

Just as important as communication skills is statisticians’ disciplined way of thinking about evidence, uncertainty, and decision-making. Statisticians are trained to question assumptions, interrogate data quality, evaluate competing explanations, and determine how much confidence should be placed in a conclusion. In an era increasingly shaped by data, predictive models, and AI-generated recommendations, those habits of mind are proving valuable far beyond the traditional boundaries of the profession.

Statistical Thinking: The Unique Contribution

Across academia and industry, the pattern is consistent. Organizations gain when someone in the room is trained to ask the following:

• What is the quality of this data?
• What are we not measuring?
• What is the uncertainty around this projection?
• What are the long-term consequences of optimizing only for what is easiest to count?

For example, according to Stern, he and his chancellor (a political scientist) “tag team” salary equity studies because they share the understanding that falling below a regression line does not automatically equal bias—natural variability and individual context matters.

Stern also described applying statistical thinking to a recurring admissions problem. Every May, projections showed a coming flood of chemistry, physics, and math majors that would require extra resources. Then in October, after students changed majors over the summer, the numbers dropped sharply. His solution was straightforward—average the historical drop and plan resources accordingly.

“Thinking hierarchically is important. In my case as provost, I have 15 academic schools that report up to me. You have to think about when the graduation rate appears to differ across the units. Is it a real thing, or is it just a random thing? To me, that’s what traditional hierarchical models would tell you to do: Recognize that a lot of the variation you see is because somebody has to be the highest on the list and someone has to be lowest. Understanding that for whatever you’re looking at is key.”

ASA Fellow and former ASA President Sastry Pantula, dean of the College of Natural Sciences at California State University-San Bernardino, faced similar realities when building a new budget model. Clean metrics such as student credit hours and number of majors could be measured, but research output in the humanities was harder to quantify. The solution blended roughly 70% metrics-driven allocation with 30% strategic cross-subsidization to protect mission-critical areas. He stressed systems thinking, separating signal from noise, and discerning between long-term patterns over short-term blips as key contributions of his statistical leadership.

“Metrics are important, but they don’t capture everything,” Pantula said. “You can measure student credit hours and number of majors pretty cleanly, but how do you measure research productivity in the humanities or the value of certain programs that may not generate a lot of credit hours but are central to the mission of the college? You have to be careful not to let the metrics drive everything. There has to be some room for strategic decisions to protect areas that are important, even if the numbers don’t fully reflect their value.”

If statistical thinking is valuable for leaders facing uncertainty, its importance only grows when those decisions are increasingly influenced by artificial intelligence. AI can produce answers with remarkable speed and sophistication, but it can also obscure assumptions, amplify biases, and project certainty where none exists. That’s why the same skepticism, curiosity, and discipline that define statistical leadership become even more valuable when a machine joins the chat.

AI: The Statistician’s Role

ASA Fellow Galin Jones, who leads the University of Minnesota’s AI Hub as vice provost for AI, argues that statisticians have a distinct and necessary role to play with AI. “You cannot do modern AI without statistical thinking,” he said. “It’s impossible.” He’s made responsible use and academic integrity central to the university’s AI strategy. Every incoming student now receives baseline AI literacy training, and academic units across campus are developing their own “AI blueprints” to guide thoughtful adoption.

In Jones’ view, the disciplined system for dealing with uncertainty, variation, and limitations of data are key aspects of statistics that must be applied to working with AI. In fact, the ability to distinguish knowledge from confidence may be one of the most valuable statistical leadership tools that can be applied in marshalling AI within academic or corporate environments.

“We tend to get pigeonholed as a STEM discipline, where we focus on math and computing, and that’s all we do. But I think we’re much more closely aligned with philosophy than people appreciate,” Jones said. “We really worry about: ‘When do we know something?’ And that’s epistemology. And so, I often like to think of us as applied epistemologists. Somehow, we work at the intersection of math and computer science and philosophy. And that type of thinking really leads us to think much more broadly about things than just kind of doing the algorithm, implementing the algorithm, and getting the output.”

Without a statistician’s grounding influence, the application of AI risks becoming untethered from rigorous evaluation and ethical scrutiny—a function Jones believes will become increasingly important as AI tools become more powerful. He stresses that transparency, reproducibility, and honest communication of uncertainty are essential and shouldn’t be done in isolation.

“Statistics is really a way of viewing the world. It’s not just the technical stuff that we teach people. Understanding the role of chance in the world, understanding causal inference, understanding how these things can be used to make decisions—I think that’s really the fundamental set of skills. How the data tell the story and how that story affects our decisions.” 

For Jones, understanding what can be known also means recognizing what cannot. The deeper statisticians engage with uncertainty, the more they appreciate the limits of their own knowledge. That realization, he argues, often leads to a leadership style rooted not in certainty or authority, but in curiosity, humility, and respect for the expertise of others.

“It’s about having humility. It’s about having a sense of curiosity. It’s about valuing other people’s contributions to the world the same way you would a statistician,” Jones said. “A lot of people don’t take that approach when they’re working with other groups. They think they have the answer, and they want to drive the conversation. But if you approach it with curiosity and humility, it’s easier to get people on your side. It’s easier to get them to trust you. And ultimately, you’re going to end up with a better solution because you’ve incorporated perspectives that you wouldn’t have otherwise considered.”

The Future Needs Statistician-Leaders

The question is no longer whether statisticians belong in leadership roles. Increasingly, the demands of modern leadership are requiring the very skills statisticians and data scientists spend their careers developing.

But the statistician-leaders featured here didn’t succeed because they always had the answers. They succeeded because they knew how to interrogate assumptions, evaluate evidence, communicate uncertainty, and adapt when new information emerged. They understand that leadership in complex environments is rarely about projecting certainty. More often, it’s about helping organizations make better decisions when certainty is unavailable.

These capabilities are only becoming more important. The future will need leaders who can distinguish signal from noise. Leaders who can balance evidence with judgment. Leaders who can collaborate across disciplines, communicate with clarity, and remain intellectually humble in the face of uncertainty.
In short, the future will need leaders who think statistically.

The National Academies study calls for statisticians to take on larger leadership and co-leadership roles in shaping research, education, and innovation. The experiences of Reese, Stern, Pantula, Jemiai, and Jones suggest that future has already begun.

Adam Barone

ASA Senior Editor (Contract)

Barone has been recognized seven times for excellence in journalism by the New England Press Association, New Hampshire Press Association, and Hoosier State Press Association. He’s covered a diverse range of topics, including technology, business, social issues, and sports. He also adds 20 years of award-winning experience as a copywriter in the marketing/advertising world. He holds a BA with a double major in journalism and philosophy. Deeply curious and excited about the future of statistics and data science, his goal is to tell stories that spark curiosity and intrigue at the intersection of statistics, data science, and society.