The San Francisco Chronicle (and several other places) have featured an article on Google using an algorithm to predict the people likely to leave.  I have been avoiding this post ever since the Chronicle article…because I have such mixed feelings.

Some things I like:

  • Google cares this much about the workforce – as we all should!
  • REAL predictive analytics finally gets a showing!  So many people are using the term predictive analytics about things which are really just metrics and reporting…it’s a wonderful thing to see real PA at least being thought about.
  • This will add value…but maybe not in the obvious way

Some things I worry about:

  • The algorithm can only use historical data, and given layoffs, Google’s slide from the #1 Best Place to Work spot, the economy and a whole lot of other current and recent change, that might miss some key items
  • Google searches are great, but they don’t get everything…and if management at Google starts to think that they do, there is a serious risk of complacency and so further loss of focus on the value of human management (“oh, Jane didn’t show up on the output of the algorithm this month, so she mustn’t be about to leave”)
  • HR data is much more limited than internet data, so there is a serious risk that the model won’t have enough data points to be relevant. Yes, great mathematicians can overcome many shortcomings…but maybe not that one
  • If we predict individual human behavior, what risks do we open up?  Lawsuits, even?  What if we get it wrong about Sally and don’t promote her because the algorithm said she’s likely to leave?  Sure, we already do that in management heads, but what’s the legal situation once it comes from an algorithm?
  • Then there are the people worrying about Big Brother:

Some lessons for us all:

  • It’s a good idea to focus at least a part of our organization’s innovative power on our workforce…after all, they ARE our innovative power!
  • We need to be more rigorous about workforce decisions
  • Maybe we need to keep our “predictions” to groups, not individuals
  • Really smart companies are focusing serious energy on doing workforce planning

There is a lot of potential in PA (Aruspex is loving it right now), but it can’t replace human decision making…or can it?  What do you think?