1 - It's imperative that the criteria for ranking employees is well-aligned with company goals. At Microsoft for example, on most teams you are ranked mostly on what features you ship. That sounds great but it can lead to extreme feature bloat in products. It's hard to take a stand for simplicity and elegant design in a meeting when the other guy's job depends on shipping the feature, and your job depends on getting the other guy to agree to include your feature.
2 - AMZN and MSFT both merge each manager's ranked list of direct reports with the lists created by other managers who report to the same Director or Senior Manager. This process can often be much more political than meritocratic, since other managers usually don't know much about an individual engineer's performance. If a team is flat-out better than other team and all of their engineers deserved to be ranked higher, that can be very difficult to defend in a meeting in front of a VP.
So you can't really blame a policy when it gets screwed up by bad management.
+ Executive Staff gives an assignment to a certain manager.
+ Manager divides that assignment up and gives it to his team members
+ Team members complete the assignment
+ Manager submits the completed assignment back to the Executive Staff
Given that scenario there's no way anyone other than the manager can know who accomplished what meaning no matter what system you implement it's always going to come down to the judgment of managers.
(which is why hiring good managers that don't try to game the system is so very important)
+ with ~1000 employees, chances are you as a programmer are most likely doing mundane crap. The "extremely smart and focused" people don't take well to having to do bullshit projects.
Have you seen how in-depth the translations on Facebook are? I used Facebook Pirate for three or four months because I was so delighted at how many variable phrases they introduced regarding every small item on the page. It was as thorough a translation as I've ever seen online. (As an unrelated aside, Facebook supports the Konami code, which makes it the most mainstream geek site on the planet, and their Konami code triggers something wholly unexpected and hilarious.)
I don't know Facebook's employee breakdown, but remember that Facebook offers multiple products, each of which is a fairly complex offering. I can see their Photo app being worthy of a large interesting team; ditto Events, Notes, Video, their link sharing, their search (and doing search for Facebook sounds much more interesting than doing search for Google, because you get to play with so many concrete variables), their public pages offering, their iPhone application, and who knows what else they're cooking up. They've also got a unique graphic style that I'm sure a handful of people work on, and there's the usual stuff of marketing and advertising that's fun no matter where it is.
It doesn't seem completely insane to me that working for Facebook might be really fun, even as a grunt. Unlike Google, which always sounds like working there is 90% backend, Facebook has a lot of stuff that's immediately noticed by its users. It's one of the few big companies that I'd actively like to work for after graduation.
http://www.google.com/intl/xx-pirate/
http://www.google.com/intl/la/
We have Klingon too -
With 20,000 engineers now, Google is as "cool" as HP or IBM.
[Edit for obscure reference: In film Annie Hall, someone criticises Marshall McLuhan in a line at the cinema, and he pops out from nowhere and says 'You know nothing of my work'. It's quite a famous scene.]
http://www.techcrunch.com/2009/08/22/facebook-hires-tipjoy-c...
Basically I thought you might have gotten a better than average contract there as a devel since you had a background doing cool stuff and was clarifying that Tipjoy wasn't actually acquired.
Let's say at a large company the HR department worked out an "objective" means of ranking employees, but when they ran it it turned out that the bottom 10% were overwhelmingly women/an ethnic minority/whatever. They're going to tweak it until it delivers the result they want, in which case how confident can anyone by that it was "objective" to start with?
It's simply an alternative employment strategy to Google's attempts at early optimization (making it hard as hell to get in) which they even concede produces a lot of false negatives.
The reality is a company the size of IBM is always going to have a turnover of 10% or more so the "lower 10% rule" is just a way of getting rid of those people who can't do the job. The 70% might feel a little demoralized but if they're good employees they'll use that to make themselves better. The upper 20% are the best of the best and deserve to be treated as such.
So to me the 70-20-10 principle just seems like IBM being honest with employees about where they stand and giving them a chance to recognize it and do better if they so choose (some people are happy with average).
I have been in a startup, where the company had matured some, and the options given to new people were almost irrevelant, while the people that had higher positions, had been there longer, and would push people to work hard, b/c they had the incentive to do so.
People realized this, that they were working startup hours, but not getting options, or a big corp pay either, and getting the shaft. A lot of employees and managers just left. I went through 6 different managers in one year. It got ridiculous, and damaging to the morale.
Meritocracy, relative performance, individual incentives...these are all Microsoft innovations from decades ago. Kudos to Zuckerberg for taking note of a successful formula but credit still should go to Gates an co.
I think the real value of this article is the info on Facebook's internal policies, though. Ranking people based on relative performance? Relentlessly replacing lack-luster employees? A focus on providing individual incentives? Sounds like a really good (though scary) environment for ensuring that people do their best work.
It's well-known that big companies are much less efficient than their smaller counterparts, but this seems like a step in the right direction for closing that gap.