"The Internet search giant recently began crunching data from employee reviews and promotion and pay histories in a mathematical formula Google says can identify which of its 20,000 employees are most likely to quit"

http://online.wsj.com/article/SB124269038041932531.html

I think Google will do good in the Auto Insurance industry. It will do good, sorting out the good from the bad drivers.

Those 'underused' employees must be the ones that only work 80 hours a week.

they should release an API for this algorithm.

This post is missing the "1984" tag

Some quantifiable performance measurements for developers:
- Number of source code commits.
- Number of lines of code.
- Number of recent vacation days, late mornings, early departures (for interviewing) .
- Number of meetings attended (can be gleaned from meeting room schedules).
- Number of incoming/outgoing workstation phone calls.
- Periodic manager feedback forms (matrix format?).

If their staff is leaving primarily because they are "under used", you could look for certain characteristics and trends:
- Steady low baseline.
- Decline. (Maybe the developer is being sidelined on their team.)

Of course, after analyzing these parameters on past cases, they will likely find meaningful graph patterns. Some of these measurements are more significant than others so they can be weighted.

Disclaimer: We all know LOCs and number of commits are not *always* objectionable measurements for contribution, though they will certainly not be high if the developer is under performing (whether voluntarily or not).

Lines of code committed are often highest for the under-performing developers. The developers who make sure their stuff is bug-free before they commit it are the ones who reason about the code before they commit it.

Metrics like these are inherently flawed. The best developers are often the outliers. If you tweak a company based on the mainstream, you might lose the underperformers but you will also lose the overperformers. The company will turn into the McDonald's of code.

Yes agreed – see disclaimer. That is also why you might want to weight some of those parameters. Remember, likely they would be using the formulas for initial profiling, not conclusive decision making.

Also, if you're looking for performance or contribution, you most certainly want to look into the number and nature of the developer's commits, even if you choose to do nothing with it. In the very least, it is not a metric to be ignored. (Zero commits *could* mean zero contribution, but that could only be determined by following up with their team leader or manager.)

Of course, they used WOLFRAM ALPHA to calculate this.

For details, send $1500 in cash to Wolfram Research.

The algorithm:

if (timeSpentOn(Twitter,Facebook,...) > workTime) {
   userIsBoredAtWorkAndCouldPossibleQuit = TRUE;
}else{
   userIsBoredAtWorkAndCouldPossibleQuit = FALSE;
}
  

Bob

Their market research team must be a false positive in that algorithm