There’s a great scene in the movie Moneyball where the analyst, Peter Brand, is explaining his analysis to the general manager of the Oakland Athletics baseball team, Billy Beane. He gets all the key findings onto a big whiteboard, and says,
“I was trying to understand which baseball statistics are the biggest predictors of wins. I ran a multiple logistic regression with wins as the outcome variable, and these fourteen variables as predictors. The R-squared value of the best model....”
...just kidding. As you might have guessed, Peter Brand did not say that. What he did say tells us something about how we might communicate the results of our people analytics research:
“So using this equation in the upper left right here, I project that we need to win at least 99 games in order to make it to the postseason. We need to score at least 814 runs in order to win those games, and allow no more than 645 runs. ... People who manage ballclubs think in terms of buying players. Your goal shouldn’t be to buy players. It should be to buy wins. And in order to buy wins you need to buy runs.“
People analytics is compared imprecisely to Moneyball far too often, but Peter took a complex and novel analysis and boiled it down to what Billy cared about: wins, and how to get more of them. We can learn a lot about how to talk about our analyses by taking a closer look at the sports world and particularly sports reporters.
Sports reporters don’t assume their readers have stats degrees, but they use a lot of data in their articles. They make that data accessible by (1) giving numbers the context they need to make sense, (2) writing to help readers understand the key points, and (3) using qualitative research, or quotes, whenever appropriate.
Good communication is valuable
I’m not going to spend too long in this note persuading you that it’s worth the time and effort to make the results of your analyses comprehensible and interesting to business leaders. There are plenty of other articles that do that, including one from McKinsey that talks about the “analytics translator” role. My own role at Facebook, the analytics business partner, is a similar type of role meant to help bridge the complexity between subject matter experts on the business/HR side and the people analytics side.
People analytics professionals often talk about reporting and research as separate, but both require effective communication to be useful. I spend more than half of my time doing reporting -- not creating dashboards or reports, but pulling metrics and research results from a variety of sources together into a single report that a leader can read and understand easily. That’s another reason I find sports reporting to be such a good example of how to do my job effectively.
With that out of the way, here are some examples from the sports world.
#1: Give numbers the context they need to make sense
With apologies to any readers outside the US, my first example is Nov. 4, 2020, ESPN article by Kevin Seifert that ranks the best quarterbacks in the National Football League (NFL) up to that point.
Candidate No. 1: Russell Wilson, Seattle Seahawks
Wilson has thrown an NFL-high 26 touchdown passes, more than any player in NFL history through seven games except for Tom Brady, who connected on 27 in 2007 on the way to 50 for the season.
This sentence should be clear even to somebody who doesn’t know much about the NFL. You do need to know that a touchdown is a good thing, but beyond that, our columnist gives us the context we need to help us know that 26 touchdown passes are a lot, even by historical measures.
Next, our columnist preempts the question: “why has Wilson thrown so many touchdowns”? He uses a lot of metrics here but continues to give us the context we need so we don’t have to be the expert in all of them (I’ve bolded the context for emphasis).
By now, it's safe to say that Wilson has mastered the art of a chaotic pocket. He has been pressured on 35.9% of his dropbacks, third most in the NFL, in part because he has held the ball for an average of 2.94 seconds before throwing -- the fifth-highest rate in the NFL. But 12 of his touchdown passes have come while under duress, and his next such scoring throw will be the most in a season for a quarterback since ESPN started tracking them in 2009.
Every single number is wrapped in context.
#2: Write to help people understand the key points
In Keith McNulty’s excellent article, he persuasively argues that “the primary responsibility of the data professional is to pursue a sound approach based on their expertise and to generate reliable, trustworthy results. Only then should the communication of those results be considered.”
With that counsel firmly in mind, I’d like to share another recent example, this time from the National Basketball Association (NBA). On Nov. 22, 2020, Adrian Wojnarowski (“Woj”) reported the relatively simple fact that “Utah Jazz All-Star guard Donovan Mitchell has agreed to a five-year designated rookie extension worth up to $195 million.” However, he then took the time to help me, the reader, understand why that might be interesting. Here is a sampling from the article:
“Mitchell, 24, rapidly developed into one of the most explosive and popular young players in the NBA, an instantaneous franchise cornerstone for the Jazz organization.
“Mitchell has averaged 27.3 points in 23 career playoff games, including 36.3 points in last season's seven-game, first-round loss to the Denver Nuggets, when he joined Michael Jordan and Allen Iverson as the only players in NBA history to have a pair of 50-plus-point games in a playoff series.
“During his Rookie of the Year runner-up campaign, Mitchell became the first rookie since Carmelo Anthony to be the leading scorer for a playoff team with a winning record, averaging 20.5 points per game. His scoring average has increased each season.
“Utah expects to have one of the best teams in the Western Conference this season, a roster that'll include the return of forward Derrick Favors in free agency and the re-signing of guard Jordan Clarkson.
How would you summarize these insights? I might say, “The Utah Jazz signed a player who has already had a historically good performance and has improved on it each year. Utah is likely to be a very strong team next year.” Woj chose a few key pieces of data to make that point clear and compelling. He also added context to many of the statistics he cited.
Now, is it possible that Donovan Mitchell is in fact much worse than this summary? Sure. He could have a history of injury, a terrible free-throw shooting percentage, or problems off the court (although he doesn’t). Woj didn’t talk about any of those things and could have cherry-picked these numbers to fit the story he wanted to write. As Dr. McNulty wrote, “when data professionals have been told that they need to show certain results to fit with a company objective, or are working to a pre-prepared narrative that they need to support with data, we have a problem.” I agree.
However, if you can do so honestly, taking the time to make the facts compelling and interesting is a worthwhile investment.
#3: Use qualitative research when appropriate
Neither of the examples so far used any quotes, but qualitative research is uniquely suited to understand why something is happening -- in this case, why Minnesota’s defense was suddenly looking better.
In this article from Nov. 11, 2020, Courtney Cronin used a combination of qualitative and quantitative data to explain what might have been going on with the Minnesota Vikings, an NFL team.
First, she gave context for why their defense might have been expected to struggle this year. Then she shared the “defensive expected points added” metric to demonstrate the recent improvement. She doesn’t explain what this metric is, probably assuming the reader will have a general idea, but still gives the context for the numbers (as a rank among the 32 NFL teams).
“The Vikings' defense has improved in two straight victories over the Green Bay Packers and Lions, but there is still a ways to go. Preseason expectations pointed to a dropoff on defense after losing three starting cornerbacks and two defensive linemen in the offseason.
“The defense has shown promise from where it started the season to now. From Weeks 1-6, the Vikings had a -0.15 in defensive EPA (expected points added) per play, which ranked 23rd, and now have a -0.07 mark, jumping to 13th, according to ESPN Stats & Information.”
Then she shared both summarized and direct quotes from Mike Zimmer, Minnesota’s head coach.
“It's apparent to Zimmer how young players thrown into the mix are maturing and getting better with proper technique for the 3-5 Vikings. Their eyes are in the right place. They're learning to anticipate better and play the proper leverage. And coaches have also excelled at modifying what they're asking of the players to make things challenging for opposing offenses.
"'We've had to change a lot of things because of our personnel, with what we've been trying to do defensively, and so part of it is narrowing some of the package down, some of it is adding some things that we feel like can help guys in coverage,' Zimmer said."
Athlete and coach interview snippets are notoriously vague, but in general, this additional information helps us as readers make some sense out of the change in a metric. Quotes can be just as helpful when sharing the results of a survey.
A final example
Sports reporters spend their career doing just that -- reporting. But they aren’t sending lists of numbers to the front page of sports pages. They bring together the important numbers, the context for why those numbers matter, and additional insight such as qualitative data to help readers take away a clear understanding of what the numbers mean.
Over time, sports reporting has demanded more from readers, and fairly advanced stats such as defensive expected points added in football, wins above replacement in baseball, and strokes gained in golf (golf!) are now fairly common. I think we can take this as yet another example of how people analytics can learn from sports reporting -- if our stakeholders are resistant to new, valuable metrics, we can be patient as they become more accustomed to them. Of course, some people will never want to hear about analytics, but their numbers are shrinking.
I’ll close with an example of how we might write up a report based on what we learned from sports reporters. The key detail is that employee relations cases are lower this quarter. I am, of course, making up these numbers, but I’ll try to introduce a clear takeaway, give context for the metrics, and use qualitative data to round it out.
- Employee relations cases are down to historic lows, and employees are feeling better about who they work with.
- We had 4 employee relations cases last quarter, the fewest we’ve had in a quarter since 2016 and far below our average of 11 per quarter over the past five years.
- “Employees seem happier in general recently, which may be contributing to the low number of cases. In our recent employee survey, 75% of employees said they “think highly of their peers”, up from 62% in the last survey, and 82% would recommend a friend to work with their manager (vs. 64% in the last survey).
- These improvements come four months after we implemented manager training, and several managers we talked to said that the processes we implemented are helping resolve conflicts earlier. “I know how to help employees who aren’t working well together now,” said one line leader. It is too soon to be sure the training is responsible for these promising trends, but we’ll continue to monitor them.
May we all have success making our research and reporting more impactful.
The post "What People Analytics can Learn from Sports Reporting" was first published by Ben Teusch here https://www.linkedin.com/pulse/what-people-analytics-can-learn-from-sports-reporting-ben-teusch/
About Ben Teusch
People Analytics at Facebook
I work with leaders to identify and answer their most important organizational questions and make the results easy to understand so they can make informed decisions. I enjoy teaching others how to do the same. You can read what I've written about people analytics at www.teuschpa.com.
View Ben Teusch's full profile