Hacker News Reader

You are absolutely right, and that is what Jack Welch intended when he first implemented stack ranking at GE in the early 90s. At the time GE was overly large with an ossified bureaucracy and thousands, if not tens of thousands, of deadwood employes who were just coasting along occupying a chair. It was killing the company. Stack ranking was the answer to forcing managers to make the hard decisions about who was really helping and who was just collecting a paycheck. It worked very well.

This was only supposed to last as long as required to turn GE around. Like all good fads in the business world, though, it was taken too far by the MBA crowd. They now see stack ranking as a great permanent solution to deadwood. They fail to recognize that: 1) all organizations have a certain minimum amount of deadwood, as a function of their size; 2) once an organization approaches its natural level of deadwood, stack ranking hurts more than it helps.

You state your two final assertions as if they're universal laws. I'm not saying you're incorrect -- they certainly sound valid to me -- but can you point to any research to back up your claims?

It wasn't supposed to be only until GE turned around. It was meant to be a permanent fixture in their HR policy.

When Jobs rejoined Apple there were huge layoffs and a cleaning up of projects. But it was never permanent, he struck hard and it was clean. After that the company could focus on innovating and not fear politics of stack ranking. Can you imagine if he kept that system in play, would Apple have been as successful?

Keep in mind that the whole office still had the fear of Steve Ranking, i.e. the fact that Steve Jobs was a dictatorial asshole who just so happened to also have good design sense. More importantly it's very dangerous to start generalizing from Apple to the whole technology industry, considering that you'll probably convince a number of otherwise underperforming companies that their best bet is to have an asshole CEO.

how do you apply stack ranking effectively if most of the dead weight is in the branches (middle management) but scoring heavily weights the leaves (workers) of an organization?

My feeling is that generally a group knows who sucks or is disruptive, and this is probably the most reliable data source. That's your scoring criteria. Other modifications I might make:

1. flatten the org chart for the purposes of ranking. 2. employees rank best and worst fellow employees, i.e. up to 10 best, up to 10 worst. 3. ranking is two-dimensional to distinguish between "this person is terrible and should be fired" from "this person isn't a good fit/being utilized well and should be moved". 4. Do the above and see if the results can be used to determine who should be fired. 5. If you want to repeat this process every year don't set forced quotas.

Agree. I am sure Yahoo has some cruft at the lower levels, but I would imagine that the real problems are in middle management, where many of the worst are the most adept at playing the game with the higher ups after years of practice. I really do not know of any great ways of solving this, but I have seen the problem many times in companies of similar size.

From what I remember, Microsoft has the employees rank their peers. 360 feedback. That's part of what then poisons the atmosphere, especially since of 10, the sucker in last place often/always has to leave. People implicitly then make pacts based often on politics... But at least it's annual rather than quarterly.

yeah, no system is perfect, but the modifications I suggested I think would address these issues to large extent. For example, I suggested that employees would rank their peers on three scales (good, bad, misused) instead of with "scores". An actual score would hopefully come from aggregating the whole company and doing some analysis. Hopefully by having some smart people try to put together an algorithm that seems "fair".

Also, if I remember correctly, MS employees might have had some say in their peer's rank, but then groups scores would be merged with other groups by having managers "calibrate" with each other, which more than likely gives the manager a lot of control in the final outcome.

This is not a condemnation but people often like to invent perfect employee performance systems, as you have done here, and it is akin to people writing their own cryptography. There is evidence-based research out there - look for papers by Elaine D. Pulakos for example. My understanding is that no system works in and of itself and training managers to get the most out of their people is probably the best thing any firm can do.

how do you apply stack ranking effectively if most of the dead weight is in the branches (middle management) but scoring heavily weights the leaves (workers) of an organization?

I'd make the system ignore the scores of any bottom ranked management.

1) all organizations have a certain minimum amount of deadwood, as a function of their size; 2) once an organization approaches its natural level of deadwood, stack ranking hurts more than it helps.

Like fingernails. You never cut the fingernail right up to the quick, you always leave a little bit behind...

As Yahoo is not on the verge of bankruptcy, I'd say start with a headcount freeze rather than immediate layoffs. That immediately changes the tone from "oops, you fell below an arbitrary line that makes no sense for this team", while providing an incentive to lay off true under-performers because you can replace them with better people.

Then start making individualized assessments of projects/teams. Start with the larger ones, where you can hopefully make meaningful cuts with the least disruption. Make individual judgements as to whether the team is pulling its weight or needs to be shrunk (or eliminated entirely).

Where layoffs from a team are forced, offer transfers to other teams, which can evaluate whether they would be better served by replacing their lowest performers with these otherwise-doomed employees.

Do you need an excuse to lay them off? Oracle laid off thousands of employees after their Peoplesoft merger (many of them in Oracle Apps). The decision on whom to cut (as I understand from friends who were in management at both companies) was a long and drawn out effort and involved many levels of management meetings.

But... do it once and leave it be (maybe revisiting in a few years) Doing it every quarter (as implied by the QPR term) or even every year is an organizational forced-march.

Maybe you could get away with that at a hedge fund or other highly profitable venture (though in those places you can usually do this by giving/withholding bonuses, etc)

Reminds me of a quote from The Prince:

"Hence we may learn the lesson that on seizing a state, the usurper should make haste to inflict what injuries he must, at a stroke, that he may not have to renew them daily, but be enabled by their discontinuance to reassure men’s minds, and afterwards win them over by benefits. Whosoever, either through timidity or from following bad counsels, adopts a contrary course, must keep the sword always drawn, and can put no trust in his subjects, who suffering from continued and constantly renewed severities, will never yield him their confidence. Injuries, therefore, should be inflicted all at once, that their ill savour being less lasting may the less offend; whereas, benefits should be conferred little by little, that so they may be more fully relished."

http://www.bartleby.com/36/1/8.html

This is precisely the passage that popped to mind when I read the article.

Most of the time, in companies like this the dead-weight is actually among the middle managers. Pruning the leaves is just bleeding talent. I think a better solution would be to do a one-time merciless chop-off of managers/people who don't own/maintain a sufficiently large chunk of a product or are part of an org losing money. This type of system just creates a toxic environment within the company and brings down employee morale. Not to mention, new talent would also start shying away from the company

>Managers, generally speaking, are never going to rate their team member down if that means their team will be effected by the layoff.

Where I work they forced all the managers to give up an employee for the Y2K team. Of course they all recommended their least productive team member. Then the entire team was laid off in mid-2000.

> Some solution I'm not seeing?

If the goal is to lay off 10%, then tell every rating manager that he has to lay off 10% of his direct reports, modulo a limited amount of horse trading for true stars.