More than 18 million employees play fantasy football, according to a recent Forbes article.
Nearly 70 percent of companies deploy big data technology, like predictive analytics tools, according to Gartner.
What’s the correlation?
Although both are prominent in organizations, the former is largely considered relatable, fun and understood, while the latter is typically regarded as intimidating, tedious and mystifying.
The good news is, fun fantasy football can help us better understand daunting predictive analytics—because it spotlights and leverages many predictive analytics principles and take aways.
For instance, when we examine the NFL’s top quarterbacks’ fantasy points per game, per year, from 2009-2013, we see why it’s important to create analytical models that not only consider various data points specific to the individual player, but are custom, robust and update in real time.
Let me explain this as it relates to Medalogix’s healthcare-specific modeling philosophy.
Custom. Not all analytics tools are created equal. Numerous companies offer analytics based on state or nationwide generic data. If you consider that in terms of fantasy football, that’s like offering predictions about a quarterback based on his college division’s trends. You may assume that if a quarterback doesn’t play division 1 in college, he’s not a good fantasy pick. Based on that prediction, you wouldn’t pick Joe Flacco and that would cost you some major fantasy points. You don’t have to be a mathematician to see that that broad analytics model is incomplete. The prediction fails to account for numerous factors starting with the most obvious: Joe Flacco’s specific abilities and historical performance.
Not only do generic models fail to account for an individual’s performance and predictive factors, they don’t consider data around teams, teammates or coaches. Teams, teammates and coaches largely affect a quarterback’s success just like venues, staff and physicians affect a patient’s outcomes. For instance:
Teammates/Staff. When star wide receiver Santonio Holmes hurt his leg in 2012, quarterback Mark Sanchez’s productivity dropped 35 percent. Without Santonio there to catch the ball, Sanchez was less effective.
An important component of a patient’s success is dependent on those who are assisting or caring for him. A healthcare predictive model must consider the clinical team’s strengths and weaknesses to accurately predict a patient’s outcome.
Coach/Clinician. St. Louis recruited a new head coach in 2012. The same year, Sam Bradford’s productivity increased by 30 percent. A coach’s instruction can take a player to the next level.
A physician’s decisions directly influence a patient’s outcome. A clinician’s strengths at treating specific diagnosis must be considered when predicting a patient’s success.
Robust. Our fantasy football example only considers one data point—points per game per year. We’re missing the affect that opponents, game location and weather have on a player’s performance. Although points per game per year is enough to illustrate our point for this article, it’s not enough to generate accurate predictive insights. Robust models consider thousands of data points to generate accurate insights one can confidently act upon.
Real Time. Predictive analytics tools must be able to account for real time occurrences. Tom Brady may be a good pick today, but if he injures his arm, predictive tools must be able to adjust and take his injury into account. Much like new lab results, vitals and medications must be reflected in the patient’s real time predictive risk analysis so that a clinician may react accordingly.
That’s just scratching the surface when it comes to predictive analytics. There are many different factors and considerations when developing a predictive modeling solution—especially one that takes your business growth goals from fantasy to reality.