Friday, November 06, 2009

Daft stats

I have stated in the past that incompetence does not confine itself to the Public Sector; it lives happily in private companies, particularly large ones; and it would appear that HR is a preferred abode of it.

Please allow me to explain.

Most organisations recruit people because of specialist skills: mine, for example, prides itself in recruiting "the best and brightest" in the field of software development and ruthlessly select on entry. 
Subsequently, employees progress in their career (or not) on the basis of their skills and abilities that, increasingly these days, means an ability to effectively deal with information and data, and typically (certainly in our company) in very large volumes.

This means that, on average, our staff is highly numerate, with a computer or scientific background and a keen grasp of statistics; and we are not talking the sort that helps you figure your chances of drawing an ace from a deck of cards, we're talking multi-variate statistics, clustering and inference.

Not so HR -- after all, they need to possess so-called "soft" skills, "people management" abilities; the sort, you understand, that cannot really be assessed by asking a question of the sort that has a "right"/"wrong" answer; and, similarly, their career progression cannot be assessed really by whether what they did crashed the data center, or achieved massive economies of scale.

The result is that you end up with, for the most part, with Humanities graduates, who have, at best, a very basic understanding of primary-school level arithmetic, usually rather pretty girls with more legs than brain (if the above, or the following, sounds sexist and is offensive to you, then I'm pleased: it is meant to).

Take for example, a recent decision to terminate a sabbatical program that entitled workers with a few years' tenure to take a couple of months' unpaid leave, without losing access to benefits and options vesting: nothing earth-shattering, a nice little perk that was meant to allow people to "recharge batteries" before coming back for a new sprint.

Now, in these days of mass redundancies, foreclosures and people taking their lives because of stress at work (granted, they're French: they find stressful working more than 35 hours and not having Bordeaux to wash their two-hour lunch) complaining about that perk being taken away sound at best insensitive, and at worst churlish.

However, I found rather hilarious the rationale that was given for the decision: they looked at the data, analysed it, and spent a great deal of time debating it; then decided to terminate the programme, because they found that "more than 50% of those who were taking the sabbatical, would subsequently leave the company."

Which, to me, is the perfect example of inverse-causation: in other words, having forgotten that you are analysing data for a biased sample, you reverse the cause with the effect.

I considered suggesting in jest that, based on my "detailed statistical analys of a random sample of leavers," I'd discovered that 100% of them were observed taking a monthly paycheck, so they ought to take that one perk away too.
But then I thought of those leggy HR colleagues, and desisted: one never knows, they may actually think it's great advice based on solid statistical analysis.

Blogged with the Flock Browser

No comments: