MIT Sloan Sports Analytics Conference, Boston
Here are nine rapid-fire observations about the leadership and management of data analytics drawn from the just-completed MIT Sloan Conference:
1. Resist Insularity: Learn outside your specific domain. Specialization is essential, of course, but hyper-specialization can produce insularity. Widen your lens by looking at your subject matter from from different angles. A sports analytics conference is a superb way to challenge your assumptions about the use of data in, say, higher education or business.
2. Calibrate Readiness: We heard about owners, GMs, and coaches who still resist the application of data analytics to their organization's performance. Sound familiar? Ask yourself just how receptive your leadership (or that of your client) is to receiving the valuable insights analytics deliver and what you can do to help build a culture that improves those odds.
3. Speak English ... or whatever the appropriate language. Know your audience. Translate research results into accessible language. Provide leaders with meaningful, actionable insights for making decisions and not with jargon. Don't overemphasize the methodology or statistical tools unless the leaders want it. As the Celtics' Mike Zarren told the Conference, "The communication of the data is more important than the data itself." He added, "You need to integrate the findings so they don't seem alien to the organization." And NYU Law Professor Anne Milgram said, "The technical part is easy. Getting people to understand how to use it is tough."
4. Define the Problem: Ensure that leaders and researchers alike agree on what problem is being solved as well as the definitions of key terms. It's stunning how far down the road market research projects can travel without congruence among key players on precise problem specification and, with it, agreement on the right research questions, definitions, and problem-solving context. Without problem definition, Ticketmaster President Amy Howe told us the research can be little more than a "scavenger hunt." FiveThirtyEight's Nate Silver added in a separate session that, "We put too much emphasis on technical skills and not enough defining the problem, asking the right questions, and using our intuition."
5. Find Wisdom: Wisdom is elusive everywhere these days. Few fields need it more than market research and analytics, however, where there remain giant gaps in understanding between the producers and consumers of data. So, yes, be influenced by the right data applied to he right problem in the right context, but never be cowed by it. Remember that data can mislead or lie anytime humans are involved in its collection, interpretation, communication, and application. As Nate Silver reminded us at the Conference, "personal agency still matters most." Legendary McGill management guru Henry Mintzberg has long talked about the primacy of judgment over measurement. And I've always maintained that the evidence provided by market research must be leavened by expertise and experience.
6. Match the Methodology: It's easy to fall in love with a particular methodology or statistical tool, but these love affairs sometimes end poorly. Choose methods and tools that best help solve each specific problem or at least advance understanding of it. Decide on approaches that best answer the research questions and deliver data-driven guidance for decision-making. And by all means, please don't get lost in the methodology or statistical tools. Researchers need to see the big picture if they are to help their leaders and clients do so.
7. Beware Overfitting: Less can be more in data analytics, especially in building useful predictive models. Remove unnecessary complexity, where possible. More factors in a model will generally make for a better fit with the data you already have. A better fit for the available data does not necessarily make for better prediction, however, especially for time periods outside the range of the data. I worry about the zeal for more data exhibited at the Conference. We need better, cleaner data, not more data, per se. As poker superstar Dr. Maria Konnikova told us, "People are always over calculating in poker. Generally, you only need a few of the right data points to get the prediction right most of the time."
8. Use Visualization: And speaking of the big picture, present your data in interesting, visually stimulating ways. There are plenty of good data-visualization tools out there. FC Barcelona's Javier Fernandez and his team deployed terrific visualization techniques in the Best Research Paper competition yesterday and they won. Their victory owed to many factors, of course, but their use of visually interesting "surfaces" advanced our understanding of their work and its meaning.
9. Stand and Deliver: Researchers at the Sloan Conference did a superb job presenting their findings, balancing confidence with humility. They were presentable and accessible. They considered the judges' tough questions with grace, but did not become defensive. They stood by their findings. The best research in the world will crater if it is delivered poorly or the presenter lacks sufficient confidence.
I hope this helps. Happy data hunting.
Here are nine rapid-fire observations about the leadership and management of data analytics drawn from the just-completed MIT Sloan Conference:
1. Resist Insularity: Learn outside your specific domain. Specialization is essential, of course, but hyper-specialization can produce insularity. Widen your lens by looking at your subject matter from from different angles. A sports analytics conference is a superb way to challenge your assumptions about the use of data in, say, higher education or business.
2. Calibrate Readiness: We heard about owners, GMs, and coaches who still resist the application of data analytics to their organization's performance. Sound familiar? Ask yourself just how receptive your leadership (or that of your client) is to receiving the valuable insights analytics deliver and what you can do to help build a culture that improves those odds.
3. Speak English ... or whatever the appropriate language. Know your audience. Translate research results into accessible language. Provide leaders with meaningful, actionable insights for making decisions and not with jargon. Don't overemphasize the methodology or statistical tools unless the leaders want it. As the Celtics' Mike Zarren told the Conference, "The communication of the data is more important than the data itself." He added, "You need to integrate the findings so they don't seem alien to the organization." And NYU Law Professor Anne Milgram said, "The technical part is easy. Getting people to understand how to use it is tough."
4. Define the Problem: Ensure that leaders and researchers alike agree on what problem is being solved as well as the definitions of key terms. It's stunning how far down the road market research projects can travel without congruence among key players on precise problem specification and, with it, agreement on the right research questions, definitions, and problem-solving context. Without problem definition, Ticketmaster President Amy Howe told us the research can be little more than a "scavenger hunt." FiveThirtyEight's Nate Silver added in a separate session that, "We put too much emphasis on technical skills and not enough defining the problem, asking the right questions, and using our intuition."
5. Find Wisdom: Wisdom is elusive everywhere these days. Few fields need it more than market research and analytics, however, where there remain giant gaps in understanding between the producers and consumers of data. So, yes, be influenced by the right data applied to he right problem in the right context, but never be cowed by it. Remember that data can mislead or lie anytime humans are involved in its collection, interpretation, communication, and application. As Nate Silver reminded us at the Conference, "personal agency still matters most." Legendary McGill management guru Henry Mintzberg has long talked about the primacy of judgment over measurement. And I've always maintained that the evidence provided by market research must be leavened by expertise and experience.
6. Match the Methodology: It's easy to fall in love with a particular methodology or statistical tool, but these love affairs sometimes end poorly. Choose methods and tools that best help solve each specific problem or at least advance understanding of it. Decide on approaches that best answer the research questions and deliver data-driven guidance for decision-making. And by all means, please don't get lost in the methodology or statistical tools. Researchers need to see the big picture if they are to help their leaders and clients do so.
7. Beware Overfitting: Less can be more in data analytics, especially in building useful predictive models. Remove unnecessary complexity, where possible. More factors in a model will generally make for a better fit with the data you already have. A better fit for the available data does not necessarily make for better prediction, however, especially for time periods outside the range of the data. I worry about the zeal for more data exhibited at the Conference. We need better, cleaner data, not more data, per se. As poker superstar Dr. Maria Konnikova told us, "People are always over calculating in poker. Generally, you only need a few of the right data points to get the prediction right most of the time."
8. Use Visualization: And speaking of the big picture, present your data in interesting, visually stimulating ways. There are plenty of good data-visualization tools out there. FC Barcelona's Javier Fernandez and his team deployed terrific visualization techniques in the Best Research Paper competition yesterday and they won. Their victory owed to many factors, of course, but their use of visually interesting "surfaces" advanced our understanding of their work and its meaning.
9. Stand and Deliver: Researchers at the Sloan Conference did a superb job presenting their findings, balancing confidence with humility. They were presentable and accessible. They considered the judges' tough questions with grace, but did not become defensive. They stood by their findings. The best research in the world will crater if it is delivered poorly or the presenter lacks sufficient confidence.
I hope this helps. Happy data hunting.
Image courtesy of Wikipedia.