Pilot the resourcing model before you scale it.

Most capacity decisions get made on theory. "We need two more developers." "We should outsource QA." "Let's bring this back in-house." The conversation happens in a meeting, the decision gets made, the budget gets approved, and six months later it turns out the assumption was wrong.

By then you've spent the money. You've signed the contracts. You've onboarded the people, or fired them, or restructured around them. Unwinding it is more expensive and more political than making the original call. So you don't unwind it — you double down, because admitting it didn't work is harder than just pretending it did.

The way out is a pilot. Same task, two or three resourcing models, run in parallel for 4–6 weeks. In-house team versus external specialists. One vendor versus another. Senior generalist versus junior plus contractor. Measure the same things on each: time to first useful output, quality, cost per unit, how much management overhead it actually consumed. Real data, real conditions, small money.

Most teams skip this because it feels slower. It isn't. A six-week pilot that costs $20k and gives you a real answer is dramatically cheaper than a $400k headcount decision made on a hunch — especially when the hunch is wrong.

Here's the test I use: before any resourcing decision over $100k, can we run it as a pilot first? If yes, do it. If no — if the decision genuinely can't be tested at small scale — at least name what assumption you're betting the money on, write it down, and revisit it in 90 days. The discipline isn't the pilot. It's refusing to pretend you knew.