zlacker

[parent] [thread] 5 comments
1. thearn+(OP)[view] [source] 2015-08-06 14:43:54
"average" with respect to the company, or the population of programmers as a whole? If normally distributed and sufficiently large, you'd be getting rid of half your company under the former. The latter seems like a tough thing to measure to begin with. I can't imagine either really being the case.

Though it does sound somewhat similar to the concept of "stack ranking", which a few companies (such as MS) are notable for having used as part of their annual review process in the past.

Stack ranking has always sounded to me like an absolutely poisonous thing to implement in an otherwise healthy office. But if the organization knows that it needs to implement a reduction in force regardless, then I guess it might make sense if management does not have a feel for who their best engineers are.

replies(4): >>brudge+U >>kisna7+A2 >>Bartwe+86 >>sixtyp+7i
2. brudge+U[view] [source] 2015-08-06 14:52:35
>>thearn+(OP)
The advantage of stack ranking is that it's a formal process, and one in which managers' behavior is more measurable and reviews are definitively scheduled. It also aligns employee churn with performance to some degree, moving weaker hires out and encouraging better hires to stay, again via a formal process.

What it reduces is the tendency of less formal processes to give raises to those who ask and to stiff those who don't and to throw up the "your salary is confidential". All those things it avoids are known to create a high potential for bad company culture and moral.

replies(1): >>TheOth+a7
3. kisna7+A2[view] [source] 2015-08-06 15:05:55
>>thearn+(OP)
I think you make a good point about how to define average. I am not sure what the npr people meant when they said average. I was assuming industry average. However, the questions then is - if the managers are used to seeing only high performance engineers, how do they gauge whether the engineer is really average performer. It seems like managers subjective views and their biases would matter a lot.
4. Bartwe+86[view] [source] 2015-08-06 15:33:34
>>thearn+(OP)
This hints at a pretty interesting analysis of whether this benefits companies.

If you fire anyone below "market average", you can reasonably expect to sustain an above-average workforce as long as you compensate well enough to retain decent employees.

If you fire anyone below "corporate average", the results are more interesting. You'll bring up the average quality of current employees, but have to backfill the missing employees.

If your hiring is just a normal distribution on average skill, the result is that you quickly reach diminishing returns. You fire the vast majority of your new hires every year, while slowly turning over a few of the old ones in favor of new superstars. If your hiring is better than average, though, you can reach far high equilibrium points. In effect, it makes sense to fire all the employees up to your hiring midpoint, trusting that you can efficiently raise your average quality by doing so. After that, you'll have to cycle through too many employees per position to see speedy improvement.

Of course, all of this is crippled by training delays and the problems you'll face when "we fired half the employees last year" gets around. Interesting theoretical model, though.

◧◩
5. TheOth+a7[view] [source] [discussion] 2015-08-06 15:42:28
>>brudge+U
Considering the quality of software and innovation produced by MS while stack ranking was in place, and the drop in software quality coming out of Google after its own version of stack ranking was introduced, it seems unlikely to be effective.

As I understand it individuals were stack ranked within teams, so even if a team outperformed the company or industry average, individuals were still tagged as underperformers.

And generally it became a matter of politics, popularity, and self-promotion, not objective competence - which is hard to measure anyway.

To me, it seems like a fast road to madness.

As for Netflix - I can't imagine any company needs to be staffed entirely by ninja rockstar code demi-gods. Many development projects are mundane and by the numbers, and basic competence is fine. If you want to be disruptive, hire a core of creative innovators who can code. They probably won't be ninjas, but for product development, talented customer-oriented innovators are a really good thing in any business that sells stuff to real customers.

6. sixtyp+7i[view] [source] 2015-08-06 17:15:47
>>thearn+(OP)
Stack ranking is a solution to a larger problem, which is the issue of growth in overall "size of the tribe" and political networks that shade / obscure / fabricate performance views.

By the time you're stack ranking, you're deep into large company territory... dealing with the above issues.

Medical equivalent: People taking cancer drugs have a higher than average likelihood of dying. However, that's probably a function of having cancer vs. the performance of the drug.

[go to top]