I know, you weren’t expecting a reference to Sonny & Cher.
Technical people (programmers, doctors, scientists and the like) aren’t typically considered to be good communicators to the public at large. Good communicators like Bill Nye and Neil DeGrasse Tyson stand out because they’re adept at explaining very technical subjects in a way that’s understandable to everyone. Sure, they have time to prepare, but that doesn’t guarantee content everyone else can understand.
This is one reason why we’re so frustrated with the inaccuracy of “predictions” about things like weather, fantasy football player performance, stock market behavior, hurricane tracks, asteroid paths, and COVID impacts.
How many science-y people in these roles are saying something like… “This is a model. This is how models work. A model is not a promise. It is a set of results from a bunch of calculations based on the data we have today – and the data we don’t have yet. When the data changes, the results coming from the models will change.“
The lack of this kind of communication causes modeling to be devalued by everyone else.
What you don’t know
Data changes rapidly – weekly, daily, hourly. Some of today’s data could be inaccurate. We may not know that until tomorrow’s data arrives, or a sensor fails.
Consider hurricanes. Hurricane models “predict” their path & severity. The output changes as variables are added /changed / deleted, and as varialbe importance changes. As the hurricane gets closer to shore (or as the time to make your third round draft pick nears), models become more accurate because there are fewer variables, & the possible range of still-useful variables shrinks.
What don’t we know?
When Donald Rumsfeld was Secretary of Defense, he was asked about then-recent discoveries about WMDs in Iraq. The questions were legitimate as was his answer, though he was mocked for it at the time.
“Reports that say that something hasn’t happened are always interesting to me, because as we know, there are known knowns; there are things we know we know. We also know there are known unknowns; that is to say we know there are some things we do not know. But there are also unknown unknowns—the ones we don’t know we don’t know. And if one looks throughout the history of our country and other free countries, it is the latter category that tend to be the difficult ones.”Donald Rumsfeld (2002), speaking as U.S. Secretary of Defense
Anyone who has worked with business metrics, science, or fantasy football knows that he was right.
Despite this variability & the knowledge that tomorrow could look much different, we often have to make decisions using today’s data.
Predicting people performance
If you’re trying to predict the performance of a NFL player, it’s equally difficult. We know a player has a 44″ vertical, runs a 4.2 second 40 yard dash, and is a three year All-Star (and more), yet we still can’t accurately predict his stats for next game.
We don’t know that his mother is sick, or that a tiny injury is bothering him intermittently. We might not notice tiny performance differences that affect a game’s outcome. Perhaps only the player who covers him will notice.
After the game, the coach might tell the press that they called different plays “because he’s hurting a little bit” as a ploy to distract their next opponent. It’s Rumsfeld’s “unknown unknown” to most of us.
You don’t know when you draft a great quarterback that you’ll lose him for the season in week seven because he tripped over his own feet during practice. Likewise, if you don’t know your best salesperson’s mother has terminal cancer, you won’t know that (or how) it affects their work.
Will models help you?
How’s your team? Is anyone challenged by something that impacts them like a nagging injury? How distracted would you be in that situation? What would help you? What needs do your people have that they don’t normally have? How can you help? Can they help each other?
What aspects of your clients’ performance could be predictive? What data is indicative of their performance? What *was* indicative but has changed? What don’t you know? Have you checked in with them? How can you help? Can they help each other?
Can performance modeling help you see performance changes earlier? Can models help you make better decisions earlier?
What don’t you know?