What’s In A Number?

Topics: HRExaminer, John Sumser, by John Sumser

For the past 18 months, I’ve been asking a question every time the subject of data (big, little, analytics, metrics, privacy) comes up. It’s been my way of illustrating the difference between how HR currently uses data and what’s actually required. I’ve probably put the question in front of 7500 people if you count the various presentations.

The question is:

Imagine that you are recruiter and I am a hiring manager. I have to hire somewhere between 100 and 120 people this year. The hiring demand happens seasonally; there is absolutely no level flow. One month I need a lot of people, the next month I need almost none. When I have an empty chair, I have to pay overtime to get the work done. (So every week a desk is empty, I have to pay for those 40 hours at time and a half.)

“Average time to hire” is a singularly useless number for me. It turns out to be different depending on the season and the local economy. While I understand that that’s how the recruiter is evaluated, I need a better metric to predict my costs. My bonus depends on hitting financial targets and payroll is my largest single expense.

(I want to be sure that you understand that I’ve been directly asking industry thought leaders and innovators as well as regular practitioners.)

As I’ve asked the question, I get pretty much the same range of answers. There’s a fair amount of ‘beats the crap out of me’. The more enlightened folks say something like “that would involve predictive analytics and we’re not there yet.” I have yet to have a conversation that goes “Oh, that. Here’s how you measure and explain it.”

So, I’ve been working with the assumption that this was a problem that requires some sort of rocket science applied to recruiting metrics. I’ve used the example to point out the fact that most HR measures are internally focused and do not offer interesting insight to the users of HR services. In this case, Average Time to Hire is a great tool inside HR. It is of little use to a hiring manager.

The other day, I was working with some tracking statistics that I follow. (I’m a ‘Quantified Self’ hobbyist and track all sorts of stuff.) I was trying to solve a problem that most averages have: the bigger the sample size, the less impact an individual measurement has. In this case, I was tracking my blood sugar level.

Over time, and generally over the course of a month, my blood sugar stays pretty constant at 135ish. A new measurement never changes the lifetime average. It rarely changes the monthly average.

But I know that there’s a lot of variation in any given month. If I let my sweet tooth get the better of me, the measurement gets high and I have to bring it down. Using the average simply doesn’t help me.

So I started playing with different kinds of averages. It was really interesting to see that there was important formation in moving averages that covered 2,3,4,5,6 or 7 consecutive measurements. At that level, you can see the impact of performance improvement.

I realized that I had discovered my own answer to the question of how do you predict recruiting performance in a situation where hiring demand is variable. You don’t need rocket science or complex big data projects. You simply need to care about your customer and work with her to find a measurement that gives her the insight she needs.

Big data is not about how complicated you can make something. It’s about how quickly you can deliver actionable insight.

Tagged with:  
Read previous post:
Is Engagement A Near Miss Solution?

Homer Simpson, Paul Hebert and Fermat’s Last Theorem. Hmmm. What could go wrong with this Employee Engagement scenario?