3 ms·

I recently reread The New Economics. Deming talks about a factory worker screwing up on the production line, ruining a large quantity of product, and getting pu

I recently reread The New Economics. Deming talks about a factory worker screwing up on the production line, ruining a large quantity of product, and getting punished for it. He makes the argument that the design of the line, company policy, etc, have a far greater impact on workers’ productivity than the intrinsic qualities of those workers. I told this to a colleague and she said “well yeah, I can see how that would be true in a factory but it doesn’t apply to software”. I pointed out that I haven’t dereferenced a null pointer in years, and haven’t been the cause of a buffer overrun ever, and that's all because of tooling, not because I’m such a badass engineer. Having been convinced by Deming’s arguments, I can’t think of rank-and-yank as anything other than a deeply flawed, costly, and counterproductive optimization strategy, all based on a false assumption.

vorpalhex 5y ago
It depends entirely on what you're doing. I've worked in shops were I helped setup a golden way of building software.. and it worked great. We could take in junior engineers with no clue how to code and have them writing really good stuff in like two weeks, basically impossible to screw up. The tradeoff was that all of our tasks were.. basically the same kind of task. It was rare to do "revolutionary" work. I work now in a situation where every single case that comes across my desk is actually unique. I still have a toolbox, but I need a really big variety of tools and picking the wrong tool is very easy. Suddenly instead of a golden way I'm actually focusing on testing designs quickly and in isolation to make sure we're picking the right tools for some given design.