CPG Scheduler
All posts

SMED Changeover Reduction: Convert Internal Setup to External and Break the 10-Minute Barrier

smedchangeoverscpg operationsoeecapacity

SMED Changeover Reduction: Convert Internal Setup to External and Break the 10-Minute Barrier

Every flavor swap, every label change, every fill-size switch eats a hole in your run window. And here's the trap: because changeovers are expensive, schedulers compensate by running bigger batches — which means more inventory, more shelf-life risk, and slower response to what customers actually ordered this week.

There are two ways out of that trap. The first is to change in a smarter order — sequencing SKUs so you pay the smallest total changeover penalty, which we cover in sequence-dependent changeover scheduling. The second — the one most plants never systematically attack — is to make each individual changeover physically shorter. That's SMED. This post is the operator playbook for doing it.

First, measure the changeover correctly

Before you can shrink a changeover, you have to agree on what it is. The clean definition: changeover time is the elapsed time between the last good part of the old run (at full speed) and the first good part of the new run (at full speed).

That second clause — at full speed — is where most plants fool themselves. The clock does not stop when the machine starts moving again. It stops when the line is producing sellable product at rate. All the slow trial runs, the dial-it-in adjustments, the first few crushed caps or off-registration labels — that's all still changeover. If your logged changeover times are suspiciously tidy, odds are someone is stopping the clock at "machine running" instead of "first good part at full speed."

Get this right first, because every improvement claim below is measured against this honest baseline.

What SMED actually promises

SMED stands for Single-Minute Exchange of Die — a method for changing equipment from one part number to another in as little time as possible. The "single minute" phrasing trips up operators constantly, so correct it up front: it does not mean every changeover should take one minute. It means getting changeover time down to a single digit of minutes — under ten minutes.

The method was developed by Shigeo Shingo, and his documented work produced changeover reductions averaging around 94% across a wide range of companies — think 90 minutes down to under 5. Treat that number as aspirational evidence that the ceiling is far higher than most teams assume, not as a promise for your line next Tuesday.

The one idea that makes it work: internal vs. external setup

SMED breaks a changeover into elements — discrete tasks. Every element is one of two types:

  • Internal elements can only be done while the equipment is stopped. Example: swapping fill nozzles, changing a forming die, pulling and replacing a sealing jaw.
  • External elements can be done while the equipment is still running the previous product. Example: staging the next label roll, retrieving and pre-kitting the change parts, pre-heating a tool, stacking the right cartons at the station.

Shingo's central insight was simple and radical: most of what plants treat as "stopped-machine work" doesn't actually require the machine to be stopped. The game is to (1) separate internal from external, then (2) convert as much internal work to external as you can, then (3) streamline whatever internal work remains.

And here's where the time hides, which tells you where to aim: in a typical setup, trial runs and adjustments account for roughly 50% of the time, and preparation/checking of materials and tooling about 30%. That's 80% of your changeover living in two categories — dialing-in and fetching — and both are prime targets. The actual mechanical swap of parts is often the small part.

The SMED method, step by step

Step 0 — Observe and document

Video a real changeover, start to finish, from last good part to first good part at full speed. Then list every element — you'll usually find 30 to 50 of them. Capture who does each one, how long it takes, and whether it currently happens with the machine stopped or running. You cannot improve what you haven't itemized.

Step 1 — Separate external elements

Go down the list and tag each element internal or external as it exists today. Then move every genuinely-external element out of the stopped window: the operator fetches, stages, and checks everything while the previous product is still running. This step requires no engineering and no capital — just discipline and a staging plan. Simply separating out the external elements often cuts changeover time nearly in half on its own. This is the cheapest win in the entire method; do it before anything clever.

Step 2 — Convert internal to external

Now the engineering starts. Take elements that currently require a stopped machine and redesign them so they don't:

  • Pre-stage and pre-kit all change parts on a cart, sequenced in order of use.
  • Duplicate tooling and jigs so the next job's parts are pre-assembled and pre-set offline while the current job runs.
  • Pre-heat, pre-set, or pre-adjust anything that currently gets set up cold on the stopped machine.
  • Convert cleaning offline. A classic illustration: a team replaced an in-place component that needed cleaning with a roll-in/roll-out standalone unit, so the dirty component could be cleaned off the line while a clean one ran. That single conversion took a setup from about one hour to six minutes. For food-and-beverage lines where wash-down is a huge chunk of changeover, this pattern is gold.

Step 3 — Streamline the remaining internal elements

Whatever still has to happen on a stopped machine, make it faster:

  • Replace threaded fasteners with functional clamps, quarter-turn fasteners, or cams — count the turns you're eliminating.
  • Eliminate adjustments. Adjustments are the 50% monster. Replace "dial it in by trial" with hard stops, position pins, graduated settings, and go/no-go gauges so the correct setting is achieved on the first try, not the fifth.
  • Run parallel operations — two operators working different ends of the machine simultaneously instead of one person walking the full length serially.

Step 4 — Standardize and sustain

Document the new changeover as standardized work: sequence, who does what, staged on which cart. Without this, times quietly drift back to baseline — some of your early gains are also just the Hawthorne effect of being watched. Standard work is what locks the improvement in place.

Shingo's eight techniques — use as a checklist: separate internal from external; convert internal to external; standardize function (not shape); use functional clamps or eliminate fasteners; use intermediate jigs; adopt parallel operations; eliminate adjustments; mechanize. Note the order — mechanize is last. Don't buy an automatic tool changer to speed up work you could have eliminated entirely.

A CPG worked example

Consider a documented capping-machine changeover (cork-stopper bottling line — about as CPG as it gets). The baseline changeover totaled 66:52, and when the team broke it into elements, 84% was internal and only 16% external. That lopsided ratio is the opportunity in plain sight: almost everything was being done with the machine stopped, and most of it didn't have to be.

Walk the method against that picture. Step 1 pulls the fetching, staging, and parts-checking out of the stopped window — immediately shrinking the internal share. Step 2 converts the parts swaps to pre-kitted, pre-set modular change parts staged on a cart during the prior run. Step 3 attacks whatever adjustments remain with position stops. You don't have to hit "single digit" to transform the economics — moving a changeover from an hour toward fifteen minutes changes what batch sizes make sense.

A second case sets realistic expectations on timeline: a regulated packaging line applying SMED at its bottleneck process cut major changeover time by 30% over 12 months. That's a real, sustained improvement — and it took a year, not a weekend. Treat both figures as illustrative single-case evidence, not guaranteed benchmarks for your line.

When SMED is worth it

SMED is a real project with real engineering cost. Prioritize it with a rule of thumb: if changeovers represent a significant share of lost productive time — say at least 20% — a SMED program is likely worth running. If your bigger problem is unplanned stops and unreliable equipment, a TPM (total productive maintenance) effort may deserve to come first; SMED on an unstable machine just speeds up the chaos.

Pick a good pilot: a changeover long enough to have real room to improve but not so monstrous it overwhelms the team; with large variation in current times (variation means process, not physics, is the problem); with multiple opportunities per week to test and iterate; an engaged crew; and ideally one that sits on a constraint or bottleneck, so every minute saved flows straight to throughput. If the line isn't the bottleneck, see drum-buffer-rope scheduling before you invest.

Why schedulers should care: SMED changes your math

This isn't a maintenance project you watch from the sidelines — it rewrites your planning parameters.

  • Smaller economic batches. Changeover (setup) cost is a direct input to the economic production quantity. Cut changeover time and the optimal batch size shrinks — you can run more SKUs more often without penalty.
  • A tighter production wheel. Shorter changeovers compress every-product-every-interval (EPEI), letting you cycle the full portfolio more frequently.
  • Fresher product, less write-off. Smaller, more frequent runs mean less finished goods aging in the warehouse — directly relevant to shelf-life-constrained scheduling.
  • Smoother flow. SMED is a key enabler of production leveling (heijunka) — it reduces the unevenness (mura) that big batches force on the system.
  • Higher OEE. Setup/Adjustments is one of the Six Big Losses, and it hits the Availability component of OEE directly. Faster changeovers reclaim schedulable run time.

Pitfalls and reality checks

  • "Single-digit" is a target, not a mandate. Push toward it; don't torture the line chasing the last 30 seconds when there are diminishing returns.
  • Don't mechanize first. Automation of a wasteful process just makes waste faster. Eliminate and simplify before you spend capital.
  • Guard the gain. Without standardized work and periodic re-measurement, times creep back. The first video showed people trying hard because they were watched; the standard is what makes it stick.

Operator checklist

  1. Measure right — last good part at full speed to first good part at full speed.
  2. Video it and list every element.
  3. Separate external from internal — bank the near-half "free" reduction.
  4. Convert internal to external with pre-staging, duplicate jigs, and offline cleaning.
  5. Streamline the rest — functional clamps, eliminate adjustments, parallel work.
  6. Standardize so it stays fixed.

A scheduler who knows the true, honestly-measured changeover time — and has systematically shrunk it — can plan smaller, fresher runs with confidence instead of hiding behind oversized batches. That's the whole point: SMED doesn't just save minutes on the floor, it frees you to plan better.

Sources

More from the journal