Being an Asian who suffers from what I call “red face syndrome” (I also call it “beet red” or “lobster red”) I was quite interested when the following article was forwarded to me.
People whose faces turn red when they drink alcohol may be facing more than embarrassment. The flushing may indicate an increased risk for a deadly throat cancer, researchers report. ((Nicholas Bakalar, March 20, 2009. “Drinker’s Red Face May Signal Cancer Risk.” Retrieved from NYTimes.com on April 18, 2009.))
What?!? So just because I apparently lack an enzyme that helps to process alcohol, I might be at increased risk for cancer? Read on.
But those with only one copy can develop a tolerance to acetaldehyde and become heavy drinkers… Reducing drinking can significantly reduce the incidence of this cancer among Asian adults. The researchers calculate that if moderate- or heavy-drinking ALDH2-deficient Japanese men reduced their consumption to under 16 drinks a week, 53 percent of esophageal squamous cell cancers in that group could be prevented.
OK – so wait just a minute – I can reduce my risk of cancer by limiting my drinking. ((Not that it matters – but I don’t drink)) So let me get the logic straight here:
- I am an Asian with “Red Face Syndrome”
- People with Red Face Syndrome sometimes become heavy drinkers
- Heavy Drinkers have a higher risk for throat cancer
- Therefore people with Red Face Syndrome have a higher risk for throat cancer.
Call me stupid, but this is the most illogical logic I have ever seen published in the New York Times. Did we stop to think that perhaps all heavy drinkers have a higher risk for throat cancer? And did we measure the percentage of heavy drinkers among Asians compared to other populations? In reading the this from the NY times, I certainly believe that numbers 3 and 4 above are correlated, but I’m not at all sure that numbers 1 and 4 above are correlated in any way. Let’s try this logic on or size:
- Employees are hired from employee referrals into the call center
- Employees from employee referrals have a high risk of leaving the call center within 1 year
- Therefore, employee referrals are a poor source of hire for the call center.
I’m hoping you can see my trouble with this logic. In the employee referral case, we didn’t measure the performance against other groupings of applicant sourcing, nor did we validate that turnover within 1 year was actually abnormal. Here’s all I’m saying. Depending on hour you design your analytics, they could seem like they are telling you a certain story. Hopefully you have someone in your HRIT or analytics organization who has taken a statistics course sometime in her past. Misinterpreting data can be pretty silly.
4 responses to “Analytics and Dubious Correlations”
Relevant, End-2-End Metrics have been elusive and needed for years! I have to hand it to Dubs for his “red face” and “call center” example. The call center, “bad referral” scenario is particularly noteworthy. Please allow me to explain.
While the example illustrates the need for better metrics, I would like to point out today’s metrics simply focus on “metrics in a vacuum.” More simply stated, most metrics are based on the silo in which they live. In the call center case, this is a “silo view” where the data is extracted from the pre-employment or applicant tracking system. This metric sadly excludes other pertinent data including, but not limited to: performance data (were our referrals bottom, middle or top performers?); true cost data (were our referrals simply unhealthy people that exploited the medical or Rx plan (and then simply left after their surgery?))? Were these referrals voluntary or involuntary (were they gaming the system for unemployment?)?
The good news is the data is available to address the real root cause. The less than good news is, end-2-end metrics are elusive since the world generally operates in silos. Given the cost pressures on just about everything, if HR really wants a seat at the C-suite table, HR needs to start thinking and acting in business terms. Relevant, end-2-end metrics are the future and will become the next best practice frontier in the not too distant future.
Relevant, End-2-End Metrics have been elusive and needed for years! I have to hand it to Dubs for his “red face” and “call center” example. The call center, “bad referral” scenario is particularly noteworthy. Please allow me to explain.
While the example illustrates the need for better metrics, I would like to point out today’s metrics simply focus on “metrics in a vacuum.” More simply stated, most metrics are based on the silo in which they live. In the call center case, this is a “silo view” where the data is extracted from the pre-employment or applicant tracking system. This metric sadly excludes other pertinent data including, but not limited to: performance data (were our referral bottom, middle or top performers?); true cost data (were our referrals simply unhealthy people that exploited the medical or Rx plan (and then simply left after their surgery?))? Were these referrals voluntary or involuntary (were they gaming the system for unemployment?)?
The good news is the data is available to address the real root cause. The less than good news is, end-2-end metrics are elusive since the world generally operates in silos. Given the cost pressures on just about everything, if HR really wants a seat at the C-suite table, HR needs to start thinking and acting in business terms. Relevant, end-2-end metrics are the future and will become the next best practice frontier in the not too distant future.
One of the reasons for the silo metrics situation described in this example and Howard’s comments is that HR data is so siloed. With Staffing data in one application, performance info in an Access database, separation data in the HRMS, etc etc there is no place for the wholistic analysis that can provide HR folks with the insight they need into the real correlations that exist. Luckily workforce analytics tools now exist that can easily integrate critical information independently of the technology platform and provide the data necessary for reliable and valid, strategic and operational analytics. Without accurate and actionable information HR will never have credibility with business leaders.
[…] Analytics and Dubious Correlations SYSTEMATIC HR | WEDNESDAY, MAY 20, 2009 […]