Why kanban fails in most machine shops

Kanban was born on a Toyota assembly line where every station did the same thing every day. CNC job shops do the opposite — every work order is different, routing changes mid-week, and "standard work" is a moving target. So shops adopt a 12-column board, abandon it in 90 days, and conclude kanban "doesn't work here."

The version that does work is much smaller. Three columns, hard WIP limits, and a digital board operators actually look at. If you want the full primer first, see our kanban sorting techniques guide.

The three columns every CNC shop needs

That's it. Resist the urge to add more:

Ready — material staged, program loaded, fixture available
Running — at a machine right now
Done at this station — waiting for handoff to the next op or inspection

Anything else — "waiting on material," "waiting on QA," "hold" — is a reason code, not a column. Reasons live on the card; the board stays clean.

WIP limits do the real work

The point of kanban isn't visibility. Visibility is the side effect. The point is the WIP limit: a hard ceiling on how many jobs can sit in "Running" at one work center.

Set it to the number of operators on that cell, plus one. That's it. When the limit is hit, no new work moves in — even if upstream is screaming. This is what forces the bottleneck conversation that scheduling theory keeps promising.

If you've never identified the bottleneck on paper first, run our CNC bottleneck analysis playbook before setting WIP limits. Otherwise you're guessing.

Digital vs. magnet board

Magnet boards look great in trade-show photos and lose their data the moment a card falls off. For a real shop:

Digital queue every operator can read on their phone or shared display — see how it works in JobLine's queue management.
Live status that updates when an operator taps "started" or "done" — same data feeds downtime tracking and the shift handoff.
Searchable history — every card has an audit trail.

A physical board is fine as a display (we love a wall-mounted shop floor display), but the source of truth has to be digital or it won't survive shift change.

A 14-day rollout

Day 1–3 — Build the digital board for one work cell. Three columns, real WIP limit, no exceptions.
Day 4–10 — Run it. Don't change anything. Note every workaround operators invent.
Day 11–14 — Review with the cell. Adjust the WIP limit, never the columns.

If it sticks at one cell, replicate. If it doesn't, the problem is rarely the board — it's usually that material staging hasn't caught up. The downtime reason codes will tell you.

When to scale up

Once two cells are running clean, layer in work center scheduling and production scheduling so the board fills itself from the schedule instead of supervisor pushes. That's when kanban stops being a posterboard exercise and starts saving real hours.

Three columns. Hard WIP limits. Digital. Everything else is decoration.