Google, building better bosses. I doubt.

In this New York Times article, the reporter explains how Google is using data to “build better bosses.” Their first amazing discovery: bosses are wanted not for technical skills, but for management skills. Wow. That’s an eye-opener. I’ll bet no one’s ever observed that before. Actually, last time I checked, companies often promote people to management based on technical skill, give them virtually no training, and then those managers do a piss-poor job. This has been going on for decades, and don’t think I’ve ever met an engineer over 25 who would find this surprising. Of course, I haven’t worked at Google.

Fortunately, they’ve now hired statisticians to analyze a gillion pages of interviews and measurements. What I love about statisticians is how they’re known for making keen, non-obvious observations and distinctions in human behavior, and measuring them. Excuse my incredulity, but … really? No sociologists? No psychologists? No cognitive behavior people? The human race knows an incredible amount about ways of understanding and measuring human behavior in data-driven, statistically significant ways.

Are Performance Evaluations Examples of Their Management Expertise?

By the way, Google has four performance evaluations a year. Frequent feedback, right? Maybe. There’s increasing evidence that performance evaluations serve almost no function except to stress out everyone involved. If an employee and their boss communicate well, there should be no surprises at such evaluations, rendering them unnecessary.

And if they really believe performance evaluations are valuable, do they bother to quality control them? Do they make sure that their managers are trained in an objective way of evaluating behavior? Performance evaluations are measurements, and the managers are the yardstick. Measuring something with a broken yardstick produces a meaningless measurement.

Are They Accounting for Cognitive Biases?

They go so far as to mention cognitive biases in the article but don’t seem to have considered whether or not there’s a halo effect in interviewing their employees about what makes a good manager. Perhaps if you ask someone what makes a good manager, they always give you the same answer, which would imply we have a built-in explanation that may or may not even relate to external reality. And if that’s so, turning a description like “listens well” into actual teachable behaviors is a trick in and of itself. (I’ll bet some of them don’t know how to rigorously realize that a phrase like “listens well” is too vague a concept to be teachable.)

Just look at the field of leadership research; pretty much every leadership book says the same thing about what makes a great leader. Yet with 80,000 leadership books on the market, we still suffer from a terrible lack of leaders. Maybe we’re wired to think about leadership in terms that aren’t specific-enough to be useful or are simply wrong. There’s ample evidence that people are extremely bad at predicting their own reactions. So asking someone “What kind of boss would be good?” may produce well-meaning but inaccurate answers.

I admire Google’s desire to be data-driven. And I also envy and admire their persistent desire to give their entire company the intellectual freedom and comfort of a college campus. Yay!! More companies should do this. (And more high tech companies have done it in the past, as long as they were pulling in the kind of gross margins Google pulls in. Most such companies abandon those cultural artifacts and revert to more traditional and soul-destroying modes when under economic pressure. I hope if it ever comes to that, Google has the fortitude to hold on!)

But at least as much as I love what I’ve heard about their culture, at least as reported in this Times article, their attempt to build a better boss is high on data manipulation and gathering, and very low on data quality.