When making your own tool is cheaper than buying time

How to work out whether the detour is worth it.

Making your own tool is cheaper than buying time whenever the total hours it saves across every remaining repetition of a task outweigh the hours it costs to build. The first article in this set made the same comparison, and it is worth returning to because the honest version has more moving parts than a glance at the schedule reveals.

Hidden costs on both sides of the ledger

The quick version weighs build time against time saved per repetition. A fuller one adds the time spent debugging a tool that fails the first time, learning to run it well enough to realise the planned savings, and maintaining it afterward, costs that are easy to see once the tool exists and hard to imagine beforehand. The manual side has hidden costs too: fatigue, inconsistency, and hands tied up that could be doing something else. A small business weighing a delivery van against courier trips faces the same ledger, with insurance, parking and a new driver's learning curve on one side and every courier bill the van would replace on the other, and the owner who does it well also asks how many deliveries are really still ahead. Writing the comparison down, however roughly, beats trusting gut instinct in either direction, and a short note of what a tool was expected to save, checked against what it actually saved, is how that instinct improves.

If the task never recurs, because the project changes direction or the need evaporates, the tool becomes pure sunk cost. That is why the estimate of remaining repetitions deserves more scrutiny than the build itself: a wrong guess about the tool can be fixed with more effort, and a wrong guess about whether the task will repeat usually cannot.

My key error with this

There was a set of captive nuts inside the vehicle that sat somewhere a hand could barely reach, and every time an assembly came apart I lost several minutes fishing them back into position, occasionally dropping one somewhere it could not be retrieved without taking more of the vehicle apart. I believed that building something to solve this would be a detour, since the operation only took a few minutes and building a jig would take an afternoon, which is the arithmetic I did every single time and never once completed properly. The part I kept leaving out was the multiplication, because that operation was not happening once, it was happening on every rebuild, by every person working on the vehicle, for as long as the vehicle existed. When I finally spent the afternoon and made a simple jig that held the nuts in place while the fastener was started, the time came back within about a fortnight and kept coming back afterwards. What replaced the belief is that the question is never how long the job takes, it is how many times the job is going to happen.