, , , ,

Image Credit: <www.quantifiedimpressions.com

Image Credit: <www.quantifiedimpressions.com>

…So I really love bringing order to information – particularly where it concerns people. I’m not a Data 1. Although I’ve repeatedly played with the idea of learning R, so perhaps one day.Scientist 1, but I enjoy tinkering and learning more about why things work. As such, I’m fascinated by some of the interesting things Google’s People Analytics teams are doing to improve the company’s talent outcomes.

Of particular note, there’s a great piece on The Atlantic here reviewing Google’s data-driven approach to people decisions. The company’s People Analytics teams have uncovered some pretty interesting nuggets related to recruiting, performance, and retention over the years. Among them: The optimal number of interviews to identify a quality hire, the ideal org structures for various departments across the enterprise, and an on-boarding agenda credited with boosting new-hire productivity by as much as 15%.

This stuff (using data analytics to improve outcomes around talent) really works. I’ve had similar success firsthand: In a prior HR role, soon after joining my new department I realized the company was sitting on a treasure trove of information (buried in our ATS). We had positions that remained unfilled after 100+ days, and didn’t know why. So I started running requisition reports and pooling data from our candidate response forms. After weeks of cleaning the data in excel, I found that in postings where the median applicant (as ranked by last salary) had an asking price that was 70% or less of the position’s pay band midpoint that our times-to-fill greatly exceeded those of jobs where the p50 candidate wanted more. Further exploring causal factors, it turns out we had a job descriptions problem! By re-thinking job titles and descriptions for some of our tougher to fill positions, however, we began to see more qualified candidates apply to the roles and our fill time materially improved.

Image Credit: <video.esri.com

Image Credit: <video.esri.com>

…With all that said, cleaning the data in the example above took a lot of time and I only scratched the surface of our total requisition database. Again, I’m not a Data Scientist; there were no doubt dozens of more efficient ways to 2. My approach involved the use of a few tried and true Excel formulas/functions and good old fashioned hard work.distill the data I was working with. 2 Any way you slice things, however, big data analyses of large employee populations takes significant time and resources.

…But the evidence seems to suggest that the payoff is worth it. Deloitte’s Josh Bersin has a great article up on Deloitte University Press here highlighting the strong correlation between a robust Talent Analytics department and shareholder value. Check out Bersin’s Talent Analytics Maturity Model below:

Talent Analytics Maturity Model

Again, you can read the full article here, but Bersin’s research finds that companies at levels 3 and 4 in talent analytics were the highest performing companies as it concerns shareholder value, outperforming the S&P 500 by 30 percent over a three year period. 

Re-visiting an article I wrote last week, however, I can’t help but wonder if a company’s business strategy ultimately puts an upper limit on how much people analytics can drive performance. For companies with highly skilled work-forces the payoff is obvious; hiring and retaining the right talent is the difference between being an industry leader and becoming irrelevant (and ultimately extinct). But what about companies that have almost entirely unskilled – and largely interchangeable – talent? Can these sorts of companies accomplish the same things more tech-minded firms do via strong talent analytics teams simply by optimizing their knowledge transfer processes? 

We know turnover is expensive, but with a well-oiled knowledge transfer process I would imagine that many of the more costly expenses (like new-hire training time) traditionally associated with high turnover could be significantly reduced. 

…It’s hard to say for sure without more data, though.

Does anyone have any research (or personal experience) that they’d be willing to share on the subject?

As always, please share your thoughts in the comments section below.