WritingSystems thinkingConcepts

Sample text — layout preview, not a published essay

Slack in Systems

Spare capacity looks like waste on the day it is measured and like foresight on the day it is needed.

Draft · 2 min read

Efficiency has an obvious appeal. A process that uses every hour, every machine and every unit of stock looks, on paper, like a process with nothing left to improve. The trouble is that paper is usually measured on an ordinary day, and systems are judged on unusual ones.

Slack is the name for the capacity that is not being used on an ordinary day. It can be time in a schedule, money in reserve, people who are not fully booked, or stock that sits on a shelf. Each of these looks like a cost when nothing is going wrong. Each of them is what absorbs the shock when something does.

Buffers do quiet work

A buffer separates two parts of a system so that a disturbance in one does not immediately become a disturbance in the other. Remove the buffer and the parts become tightly coupled. Small variations travel further, and they arrive faster. A delay in one place becomes a delay everywhere, because there is nowhere for it to be absorbed.1

Spare capacity looks like waste on the day it is measured and like foresight on the day it is needed.

This is why a system can appear to run perfectly right up until the moment it does not. The slack that was removed was not doing visible work. Its work was preventive: it was the reason a range of small problems never became large ones. Its absence shows up not as a gradual decline but as a sudden change in behaviour.

How much is enough

None of this means more slack is always better. Unlimited spare capacity is its own kind of fragility; it removes the pressure that keeps a system honest about what it actually needs. The judgement is about where slack sits and what it protects. A small reserve at a critical point can matter more than a large one spread evenly across everything.2

Making it visible

Part of the difficulty is that slack is hard to defend in the terms most measurements use. It appears as idle time or unused funds, and the benefit it provides is a counterfactual: the problem that did not happen. One way through is to name it explicitly, to treat a buffer as a designed component with a purpose rather than as an inefficiency waiting to be removed. A component with a stated job is easier to keep than a gap with no explanation.

Footnotes

  1. Coupling describes how directly a change in one part propagates to another. Loosely coupled parts can fail independently; tightly coupled parts tend to fail together. ↩
  2. Where a buffer sits often matters more than its size. A reserve placed upstream of a bottleneck protects the whole flow; the same reserve elsewhere may protect very little. ↩