Martin Fowler: Cheaper Talent Hypothesis

Corporations won't benefit from gambling on more expensive programmers. They've already created a structure that takes in programmers, anyone with even a basic proficiency, and ensures that they produce at some specified minimum. It's a system that grinds down anyone with lofty ideas about programming, but it's also a system that's relatively riskless, and that's all large corporations care about---eliminating risk. (Particularly in a department that isn't responsible for generating revenue.)

No matter how bad you are, as long as you're consistently churning out code, in whatever language, enough work is being done that the project moves ahead. It doesn't matter who does it, and at this level of quality it doesn't matter who sits down to work on it (this is huge to anyone in charge of staffing the department).

It's a model that seems to have evolved organically inside the corporation, rather than as some anti-programmer conspiracy, but certainly smart executives at Microsoft and the banks have recognized it for years.

If ThoughtWorks doesn't use this model, and it benefits them, why are they more expensive? If you're paying less overall for talent, you can increase sales by charging less for the same product. When you save on production costs you don't raise prices to celebrate. His evidence only contradicts his claim.

If you amend his blog post to say, "well...you pay more for a higher quality version of the same thing", then what he's saying isn't news to anyone, and his old thesis becomes wrong.

[Software that works 97% of the time and software that works 99.9% of the time look very, very different at the code level. The user, however, can't tell the difference. More to the point, they won't pay for it.]

Larger companies run up against the fact that this approach doesn't scale. Even if such an organization could reliably identify super-talented programmers before hiring them (which it almost certainly can't), there just aren't enough of them to meet the inexorable demand for growth. Thus, as the organization grows, so does its mediocrity. But it can coast on its earlier reputation for some time.

On projects I've observed, companies like this placed one or two of their best people on the project early, as a loss leader. These did good work, which enabled the sales people to then bring in a larger number of considerably dimmer (but highly billable) light bulbs. That's how the profits are made - land and expand. It seems to take a while for clients to figure this out.

pg made this point in one of his essays - but his conclusion was that since it's impossible to measure productivity, the more productive hackers should create startups where they will be rewarded for their superior skills.

Interesting thoughts. Visuals in the article would have been a nice touch.

What about lower pay as a cause of lower productivity?

Also, poor leadership can play a big factor in software development productivity.