SMED Results in OEE: How to Measure a Changeover Reduction

Key takeaways

• Measure a changeover from the last good unit to the first good unit at rated speed, not from machine stop to machine start.
• Set a baseline on the same format pairs before the workshop, and count the same number of changeovers after it.
• Start-up losses belong to the changeover: slow running goes to Performance and start-up rejects to Quality.
• Keep changeovers in planned production time. Booked as planned stops, they leave the OEE sum and the SMED gain shrinks.
• In the worked example, a yogurt cup line cuts changeovers from 58 to 32 minutes, and OEE rises from 60.2% to 72.2%.

To measure a SMED result in OEE, time each changeover from the last good unit to the first good unit at rated speed. Do it before and after, on the same format pairs. Count the start-up rejects and slow running too, and keep changeovers inside planned production time.

This guide is from Fabrico, the OEE (overall equipment effectiveness) and light MES platform. It shows how to set the baseline, where each part of a changeover lands in OEE and how one dairy line proves its result. For the method itself, see our SMED guide.

Where does a changeover start and end for SMED results?

Shigeo Shingo's SMED aims at a changeover in single-digit minutes, under 10. The method moves internal setup, done with the machine stopped, to external setup, done while it runs. Our guide to internal and external setup time explains the split.

The result has to include what comes after the stop. A changeover has three parts:

• Run-down: the last good unit comes off the line. The clock starts here.
• The stop: parts and settings change, checks are done. This is Availability.
• The run-up: the line restarts slow and makes rejects until it runs good product at rated speed. The clock stops here.

Many plants time only the stop, from machine stopped to machine started. That misses the run-up, and the run-up is where better settings show their gain.

How do you set the SMED baseline?

• Same format pairs. A flavour change and a cup size change are different jobs. Compare like with like.
• Enough changeovers. Take the last 20 or more of each pair, across all shifts, not one changeover on a good day.
• The median and the range, not the best one. The range tells you later whether a change is real.
• Frozen definitions. Write down rated speed, the rule for when start-up ends and how start-up rejects are counted. Keep them for the after.
• The same count after. If the baseline has 24 changeovers, the check has 24, on the same pairs.

Where do start-up losses land in OEE?

The six big losses put one changeover in three places:

• The stop is a setup and adjustment loss, in Availability.
• Slow running in the run-up is reduced speed, in Performance.
• Rejects in the run-up are reduced yield, in Quality.

So a SMED result shows up in all three factors. A team that reports only stop minutes leaves part of its result out. Our guide to start-up loss versus reduced yield shows how to keep start-up rejects apart from running rejects.

Why does booking changeovers as planned time hide the SMED gain?

OEE divides by planned production time. If changeovers are booked as planned stops, their minutes leave that time. A line can lose an hour and a half a shift to changeovers and OEE never sees it.

Then a SMED workshop that cuts those minutes barely moves Availability. The saved minutes come back as run time and as planned time at once, so the ratio hardly changes.

Christoph Roser lists this as one of three common OEE fudges: shrinking the time basis by turning losses into planned stops. Keep changeovers inside planned production time, as their own bar. Our guide to planned and unplanned downtime definitions shows where to draw the line.

What does a worked SMED result look like in OEE?

Sample line (illustrative numbers, not a customer): a yogurt cup line whose filler sealer is the slowest machine, rated at 120 cups a minute. Each shift has 450 minutes of planned production time and two changeovers. The team measured 24 changeovers on the same three format pairs before the SMED workshop, and 24 after.

One changeover, median of 24:

The range moved too: 49 to 71 minutes before, 28 to 38 after. The ranges do not overlap, so the drop is not noise. Timing only the stop would show a cut of 20 minutes. The real cut in lost output is 27 minutes a changeover, and the run-up gave 7 of them.

OEE per shift, with 40 minutes of other stops, 25 of other speed loss and 480 other rejects:

• OEE check: 32,520 good cups / (450 x 120) = 60.2% before, and 39,000 / 54,000 = 72.2% after.
• The gain is 6,480 good cups a shift, or 54 filler minutes: two changeovers of 27.
• At a contribution of $0.05 a cup, that is $324 a shift, or €324 in a euro plant.
• Over 750 shifts a year: $243,000, or €243,000, if the cups sell.

The trap. With changeovers booked as planned stops, planned time is 358 minutes before and 398 after. OEE reads 75.7% and then 81.7%: a higher number and half the gain. The 40 stopped minutes won a shift barely show, because Availability moves only from 88.8% to 89.9%.

How do you know the SMED result will last?

• Keep counting. Track the median and range of each format pair after the workshop, not just the first few changeovers.
• Watch all three factors. A faster stop that adds rejects in the run-up has moved the loss, not removed it.
• Check the neighbours. If the filler restarts faster but the case packer now starves, price the whole line.
• Write the new method and its set points into standard work, for each format.
• Rank the losses again. Once changeovers drop, another loss may lead.

Our SMED data playbook covers how to check a changeover gain after the workshop.

What mistakes make SMED results look bigger or smaller than they are?

• Timing machine stop to machine start, so the run-up is left out.
• Comparing the best changeover after the workshop with the average before it.
• Mixing format pairs, so a run of easy flavour changes looks like a result.
• Booking changeovers as planned time, which hides the gain and inflates OEE.
• Cutting a validated clean or the first-off quality check. Those minutes are not the target.

Where Fabrico fits

Fabrico is a computer vision enabled Manufacturing Performance Platform. OEE and light MES come first: stops are logged automatically, and OEE is live by line and shift.

• Stops logged automatically, so each changeover has measured minutes before and after the workshop.
• The operator confirms the reason in one tap, so changeovers get their own bar, apart from breakdowns.
• OEE live by line and shift, so slow running and rejects after the restart show in Performance and Quality.
• Losses priced in money and ranked, so the team knows which format pair to attack first.
• A camera clip as evidence for a stop, where a camera watches the machine.

Fabrico measures the changeovers. Your team runs the SMED. You start on one line, with a baseline and a decision date agreed in advance. There is no free plan and there are no tiers: each plant gets a quote. See how stops are captured on our computer vision OEE page.

Frequently asked questions

How do you measure changeover reduction?

Time changeovers from the last good unit to the first good unit at rated speed, on the same format pairs, before and after. Use the median and range of a fixed count, such as 24 against 24. Then compare Availability, Performance, Quality and good units over the same shifts.

What results can SMED give?

Shingo's aim was a changeover in single-digit minutes, under 10. How close a line gets depends on the machine and the formats. Measure your own baseline first. On the sample line, the median changeover fell from 58 to 32 minutes.

Should changeovers count as planned downtime in OEE?

Not if you want to see SMED results. Booked as planned stops, changeovers leave planned production time, so cutting them barely moves OEE. Keep them inside planned time as a setup loss in Availability, with their own reason code. That bar is what the SMED team works on.

When does a changeover end?

When the line makes good product at rated speed, not when the machine restarts. Write the rule down, such as the first good unit at rated speed. Then each shift stops the clock at the same point. Without it, each shift reports a different number.

Where do start-up rejects go in OEE?

In Quality, as reduced yield. Slow running in the run-up goes to Performance. Tie both to the changeover that caused them, so the SMED result includes them. Count them from the restart until the line runs good product at rated speed, with one written rule on each shift.

See your changeovers in a demo

Bring one line and its main format pairs. We will show its changeover stops, the losses after each restart and what they cost.

Request a demo · Contact us