You encounter predictive analytics daily — when you apply for life insurance, receive unique credit card offers, or view pop-up web ads. Movie recommendations from your online provider and loyalty cards that generate special coupons are other examples.
But do they actually change your behavior? When you hear that your local weather report predicts a 50% chance of rain, do you bring your umbrella?
Typically, predictive analytics are provided from an external source. But with all analytics, predictive or descriptive, you must ensure they are creditable before changing your behavior or making decisions in the workplace and in life. Let’s go back to the weather example.
What precisely is being predicted and how is it being measured? There is a difference between using state of the art Doppler radar versus the Farmer’s Almanac. The predictive model will be different if you are tracking a storm or forecasting rain. In either case, the model should be valid, reliable and updated. Ask to review studies supporting the model to see its effectiveness.
Why was the particular modeling technique selected? Was it to predict the weather representative of your geographic area? If you live near a large body of water or next to a mountain, your weather report should account for those variables. There are dozens of ways to model and predict outcomes, but they are not “all created equally” or appropriate for every circumstance.
Is their metric endorsed by a nationally recognized body? Ensuring that nationally accepted standards are used means confidence that weather reports are reliable and valid. Third-party evaluation is essential.
Who is doing the math? If your weather team does not have the knowledge or credentials to understand what they are analyzing, they will not be able to provide valid information that will provide answers to your weather questions. Ask about on-staff statisticians and qualified researchers.
The moral of the story? Analyze your analytics!
From the December 01, 2015 Issue of McKnight's Long-Term Care News
