I had to go to the doctor this week, for my annual physical. High cholesterol runs in my family, and my grandfather and father both passed away at an early age from heart related illness. So I go. I don’t like it, but I go.
So I’m at the doctor’s office with Dr. Vahameki, who is a really smart guy but a little crass even though he’s much younger than I am. This annoys me… that he is younger than I am yet he is my Doctor.
So we are doing our thing together, and out of the blue, he says with a slyly disguised snicker, “Would you like to know how long you’ll live?”
And I say, “Um, sure.”
And then he launches into this description of how medicine used to be all about diagnosing a problem, and then prescribing treatment, but that we have finally, because of data and analytics, begun moving into the Age of Predictive Medicine, which is much more useful. I agree on the usefulness piece, but remain annoyed.
So it turns out there’s study called the Framingham Heart Study, in which they took thousands of participants, and ran their LDL, HDL, age, weight and other factors through statistical analysis to create a predictive model of how likely you are to die from cardiovascular disease given ‘your numbers’.
So he takes me to this site, and plugs in my numbers from last year’s bloodwork, and it turns out I have a 2% chance of dying in the next 10 years from cardiovascular disease if my cholesterol levels remain where they are today.
And he says, “See, it’s not very likely…”
I try and hide my annoyance at him.
I reply, “But what if I were healthy and had lower “bad cholesterol” and greater “good cholesterol?” I can never remember which is which between all the Ls and Ds and Hs.
“You’d be at about 1 in out of a thousand, or 0.1%. But I have people in here all the time at 20-30% risk factor and those are the people that I’m really worried about.”
I am not thrilled at this good news. 2% seems high. A 1 in 50 chance of not making it to 2018. Now I am really annoyed.
I quickly get dressed and get out of there.
So for the last week, I’ve started eating healthy again and am back in the gym. But it got me thinking about the power of predictive modeling, and how far behind HR is in this space, generally speaking. This is why I love blogging. It makes you think differently about the world. One minute, I’m thinking about death and my annoying doctor, the next minute, HR analytics. Maybe I really do need a doctor.
I’m not a statistician, but for large sample sizes at large companies, there is a LOT of information that is just waiting to be discovered by progressive HR organizations who can pull the data and turn it into meaningful information. We had talked about doing this at Starbucks right before I left; running multivariate regression analysis against the thousands of store level staff to better predict attrition and the demographic trends that play out when you have large sample sizes of people.
But HR is rarely predictive. It tends to be more like ‘old medicine’, identifying what is wrong and then prescribing a fix. “You see, your attrition spiked so now we need to recruit more…..” Exit interviews. Employee relations. Compensation reviews. Most all of it analyzes post data.
It is admittedly difficult to be predictive, but it is also because we don’t ask enough smart questions. We ought to be significantly better as an HR function at predicting things. Because predictive HR is a lot more helpful that diagnostic HR.
For example, we can reasonably predict what range the US unemployment level is likely to be in the next 2 years, by comparing the delta in unemployment from the top of the boom in 1999/2000 to the peak in unemployment in 2003 (50 basis points, or 2% points overall) and fudge a little for the gravity of the economic issues that we face. It’s probably going to jump to about 8% (we can now wait and see if I’m right). And from that, HR should be able to extrapolate candidate flow and inform a recruiting strategy and resourcing plan. But the vast majority of groups won’t ever do this.
I expect in the next decade (provided I make it that far), that we’ll see much more predictive HR at the best companies.