Sustaining SMED gains: measure every changeover after the project

This article is for production managers, manufacturing engineers, continuous improvement leads and plant managers who have run a SMED project and want the changeover time they gained to still be there six months later. The project itself is well documented. What comes after is not: how to check, machine by machine, that changeovers stay short once the project team has moved on, and what to do when they start to creep back up.

What the SMED method says about the follow-up

GH Training, a French training provider specialising in lean management and Lean Six Sigma, published La méthode SMED : Tout comprendre (in French). The article recalls the origin of the method (created by Shigeo Shingo, an engineer at Toyota) and its goal, bringing a changeover from several hours down to under 10 minutes, then describes its four steps: measure the current state, move as many internal operations as possible to external, reduce the time of internal tasks and, finally, continuous improvement. On that last step, the wording is clear: “Le but est de continuer à améliorer les temps de changement de série en pilotant un indicateur associé […]” (in other words: the aim is to keep improving changeover times by tracking an associated indicator).

Their article provides the method. This one focuses on the indicator in step 4: how to build it so that it shows, changeover by changeover, whether the project gain is holding, and when it starts to slip. For how data supports the SMED project itself, see our article SMED: How to Reduce Changeover Time and Improve OEE with Data. Here we assume the changeover definition is already set: from the last good part of the outgoing run to the first good part at nominal speed.

Why the time you gained comes back

Shop floor teams know the pattern. On the day of the project review, the timed changeover is impressive. Three months later, nobody has timed it again, and the general feeling is that “it is more or less holding”. The causes of drift are rarely dramatic:

• Not every shift follows the standard. The team that took part in the project uses the new sequence; the night shift, or the temp who arrived after the training, goes back to old habits.
• External operations become internal again. The tooling cart is no longer prepared before the stop, the next tool is not preheated, the wrench is not in its place: each slip adds minutes of machine downtime.
• Adjustment quietly gets longer. Trial runs and tweaks after restart creep up, ramp-up stretches out, and a stopwatch that only times the stop does not see it.
• Nobody looks at the indicator at the right level. Changeover time is tracked as a monthly average for the whole shop. One drifting machine disappears among the others.

In every case, the problem is not the method. It is the lack of a measurement fine enough to see the drift before it settles in.

Measure every changeover, machine by machine

To sustain a SMED gain, the indicator must be calculated for every changeover, on every machine involved, with the same rules used during the project. A monthly shop average is not enough: it smooths out exactly what you are trying to see. The table below sums up what to track after the project.

Two rules prevent pointless debates. First, keep the same definition of the start and end of a changeover as during the project: if the baseline included ramp-up, tracking must include it too. Second, compare like with like: a changeover between two similar part numbers does not take as long as a change of product family. A changeover matrix (from product, to product) keeps the comparison fair.

The 3-month and 6-month reviews

A SMED gain is judged over time, not on the day of the project review. A formal review at 3 months and again at 6 months, prepared with data from every changeover, settles the question without argument. A simple five-step method:

1. Lock the reference at the end of the project. Target duration per changeover type, start and end definition, list of machines. These rules do not change during tracking.
2. Set an alert threshold between reviews. For example, if a changeover type brought down to 25 minutes goes back above 30 minutes two weeks in a row, the gap is dealt with immediately, without waiting for the 3-month review.
3. At 3 months, compare by machine and by shift. Median duration, spread, gaps between shifts. A drifting machine is a machine where the standard is no longer applied: go to the gemba and find out why.
4. At 6 months, decide. If the gain is holding, the standard is validated and the project can be replicated on another machine. If it has eroded, launch a short targeted workshop on the operations that slipped, rather than starting over.
5. Name an owner for the follow-up. A name, not a job title: the person who presents the figures at each review and triggers the targeted workshop when needed.

Standards and routines: what keeps the gain

Measurement shows the drift; standards and routines prevent it. A few practices that make the difference on the shop floor:

• A changeover standard posted at the workstation, with the sequence, the split between external and internal operations, and the target duration. It is updated after each improvement, otherwise it describes a changeover that no longer exists.
• Training every newcomer on the standard, temps and night shifts included.
• Five minutes in the start-of-shift meeting on the previous day’s changeovers: which ones exceeded the target, and why.
• A monthly review of overruns with manufacturing engineering and maintenance, to turn recurring causes into actions.
• Collecting operators’ ideas, as the continuous improvement step described by GH Training recommends: the people who do changeovers every day see the minutes to be gained.

Without data for each changeover, these routines run on opinions. With it, every meeting starts from a fact (yesterday’s changeover on a given machine took this long) and the team focuses on the gap to target.

This is where TEEPTRAK comes in. It measures OEE, downtime and micro-stops in real time: a module wired into the electrical cabinet and a sensor at the line, about one hour of installation per machine, no PLC connection, and OEE available within 48 hours. Every stop, changeovers included, is time-stamped and its duration recorded automatically, which gives SMED tracking the same data from one week to the next, with no stopwatch and no manual entry. TEEPTRAK does not replace the SMED method or team training: it provides the measurement that shows whether the gain is holding. More than 700 plants in over 30 countries use it.

Further reading

• La méthode SMED : Tout comprendre (GH Training, in French)
• SMED: How to Reduce Changeover Time and Improve OEE with Data
• Eliminating Micro-Stops: Pareto, Top 3 Root Causes, and Targeted SMED Project

FAQ

How long should a SMED project be tracked after the review? At least six months formally, with one review at 3 months and another at 6 months. After that, changeover time joins the routine indicators, tracked machine by machine like the rest of OEE.

Is a one-off stopwatch check enough to confirm the gain is holding? It gives a snapshot of one changeover, often done by the best-trained shift on the day someone is watching. Drift shows up in the series of all changeovers, across all shifts, which a one-off check cannot capture.

What should you do if changeover time creeps back up after a few months? First find out where: which machine, which changeover type, which shift. Then go and see which operations have become internal again or have slipped. A short targeted workshop on those operations, followed by an update of the standard, is often enough.