---
name: "ops-map-and-improve-a-process"
description: "Use when the user names a slow, broken, or expensive process, asks where the bottleneck is, or wants a before-after read on a fix that has to hold."
triggers: ["map this process", "find the bottleneck", "why is this process slow", "process improvement plan", "root cause this process", "before and after on a fix"]
version: "1"
---
# Map and improve a process
Use this when the user names a slow, broken, or expensive process, asks
where the bottleneck is, or wants a before-after read on a fix.
## What you need first
The process name, the pain (slow, error-prone, expensive), current
volume and cycle time if known, and who runs each step.
No numbers means you map from their description and mark every figure
UNKNOWN with the one read that would firm it. No owner means the fix
stays proposed until someone owns it.
## Map process-first
Work in customers-outputs-process-inputs-suppliers (COPIS) order, the
reverse of the classic suppliers-inputs-process-outputs-customers
(SIPOC) map: the process in four to five high-level steps, then
outputs, customers, inputs, suppliers — each step with its owner and
its handoff.
Show the map in chat before you diagnose anything. The map is the
define step; the measure and control loop below closes it.
## Find the bottleneck
Where work waits, where it gets reworked, where one person is the
single point. Name the step, the wait, and the evidence. One line per
candidate, then pick the one that carries the pain.
## Root-cause it
Ask why five times, or fishbone it across people, process, tools, and
inputs. Write the cause in one line. A cause with no evidence behind it
is INFERENCE, said plainly.
## Propose the fix smallest-first
Kill the step, simplify it, reorder it, then automate it — in that
order. Each fix carries effort (hours and owner), impact (time or
errors saved), and the before-after metric with its baseline.
## Pilot on a yes
A small slice, a fixed window, and the rollback named up front. Report
before-after at the end and update the SOP when the fix sticks.
## Control the gain
Name the ongoing control metric, its owner, and the check cadence that
proves the fix holds at 30, 60, and 90 days. A gain with no control
metric is a guess, said plainly.
## Rules
Never change a live process without the process-change approver's yes.
Never present a guessed cycle time as measured.
Save the map and the improvement log under processes, dated, one folder
per process.