operations-management · git:20260728.2d6d1c9 · 2026-07-28 · sha256 9730797b9e502837
operations-management git:20260728.2d6d1c9A
Immutable. This exact content is served forever at /api/v1/blob/9730797b9e502837.
--- name: operations-management description: Run repeatable delivery reliably by finding the constraint, reducing variability, and measuring flow rather than utilisation. Use when delivery is inconsistent or slower than the work itself requires. --- # Operations management Operations is the discipline of making a process reliably produce an outcome. The counterintuitive core is that keeping everyone busy makes throughput worse, because full utilisation guarantees queues. ## Method 1. **Find the constraint.** One step limits the throughput of the whole system, and improvement anywhere else changes nothing (see agent-workload-balancing). 2. **Protect and exploit the constraint.** It should never be idle or doing work that something else could do, which is where capacity comes from without spending. 3. **Measure flow time, not utilisation.** How long work takes end to end is what customers experience; high utilisation with long queues is a system performing badly. 4. **Reduce variability before adding capacity.** Inconsistent arrival or processing time creates queues even below full capacity, and smoothing is usually cheaper than expanding. 5. **Limit work in progress deliberately.** Less started simultaneously means faster completion, which contradicts intuition and is consistently true. 6. **Standardise the routine, escalate the exception.** Repeatable work follows a defined process; genuinely novel work is routed to judgement rather than forced through it. 7. **Make the process visible.** A queue nobody can see is a queue nobody manages, and visibility alone improves most processes. ## Boundaries Operational efficiency serves delivery; optimising a process nobody values is waste at higher speed. Standardisation reduces variability and adaptability together, which is the wrong trade for creative work. People are not machines, and utilisation targets applied to humans damage both quality and retention.