Wednesday, 1 August 2007

How to lie with analytics

After having taken a couple of days off sick I was intrigued to come across a July post by SystematicHR on sickness metrics, How To Lie With Analytics.

As we all know, a large majority of sick days are taken on Mondays and Fridays. It's easy to think that the cause of this is employees abusing their sick days to create long weekends. A better record keeper than myself, Systematic reviewed his timesheets for the last few years to find that he also took 80% of his sick days on Mondays and Fridays.


"I began to wonder why I tend to get sick during weekends and on vacations. A friend gave me some good insight. High performing people have bodies that tend to know when there is downtime coming and can force illnesses into those times to create rest. Now I have no idea if this is true, but the proportion of time I’ve been sick during vacations versus work time is completely off kilter.

We use analytics to create truths (ok, trends) from which we extrapolate “facts.” Unfortunately these facts coming from quantitative data are about as good as anything you get with statistics – you can create a version of truth for everything. You can find a correlation for your best employees and show how they abuse the sick time system to create 5 long weekends a year. (I just took Friday and Monday – I had a 104 degree fever). There might be people who went to Las Vegas. And I guess it was true… I was laying in bed fairly drugged up."



I wonder what other errors of analysis unthinking use of HR metrics leads us to?