What Is OEE, and Why Changeover Variance Quietly Wrecks It

Overall Equipment Effectiveness (OEE) is one of those metrics everyone in manufacturing has heard of and fewer people trust. Part of that is because the number often moves for reasons nobody can explain. Nine times out of ten, when an OEE number swings week to week with no obvious cause, changeover is the reason, specifically, how much it varies and how little of that variance actually gets tracked.

OEE in plain terms

OEE multiplies three ratios together:

• Availability: actual running time divided by scheduled time. Downtime, changeovers, and unplanned stops all eat into this.
• Performance: actual speed divided by the line's ideal speed. Running slower than rated, even while "up," drags this down.
• Quality: good units divided by total units produced. Scrap and rework drag this down.

Multiply the three together and you get OEE. A commonly cited "world class" benchmark is around 85%. Most plants, especially smaller ones that have never formally measured it, run well below that, often somewhere in the 40-60% range once someone actually adds it up.

Where changeover lives in that formula

Changeover time almost always gets counted against Availability, as planned downtime. That part is straightforward. What's not straightforward is how that number gets estimated in the first place, and this is where most OEE tracking quietly falls apart.

Two common failure modes:

1. Changeover isn't tracked at all, and gets folded invisibly into run time. A line that took an hour to reconfigure between two runs just shows as "running slower than usual" for that stretch, which drags down Performance instead of Availability, and nobody can tell the difference from the report.
2. Changeover is tracked as one flat number applied to every transition, regardless of what's actually changing. A same-color, same-diameter repeat run gets charged the same setup time as a cross-family die change with a full purge. Neither number is right, and the error goes in different directions depending on which transition actually happened that day.

Either way, the OEE number you get out the other end is an average that hides its own biggest driver.

The part that actually explains the swings

Here's the concrete version. Take two weeks with identical scheduled hours and the same total order volume. In week one, the schedule happens to sequence orders so similar colors and diameters run back to back, changeovers are short and predictable. In week two, the same volume gets sequenced with several dark-to-light color reversals and cross-family die changes scattered through it. Same demand, same hours, meaningfully different Availability, and therefore a meaningfully different OEE.

If nobody is tracking changeover time per transition, that week-two dip shows up in a report as "productivity dropped," with no attached explanation. The team ends up debating operator performance, equipment issues, or material problems, none of which caused it. The actual cause, sequencing, is invisible because the data needed to see it (changeover time attached to the specific transition that produced it) was never captured.

What to fix first

You don't need a full OEE dashboard before this is worth doing. The single highest-leverage practice is tracking real changeover time per transition, not a plant-wide average, and paying attention to which transitions are cheap and which are expensive. That data alone explains more week-to-week OEE variance than almost anything else you could measure, because it's usually the actual cause, just uncounted.

Once you have that, two things get easier: you can sequence orders deliberately to avoid the expensive transitions when the schedule allows it, and when OEE does dip, you have an actual explanation to point to instead of a guess.

Where RunMark fits, and where it doesn't

RunMark does not calculate or display a formal OEE score today. If you're already computing OEE elsewhere, that calculation still lives with you.

What RunMark does do is treat changeover time as a real, attached property of the specific transition between two scheduled runs, not a flat constant applied to every changeover on a line. When a run gets placed after another one, the setup window reflects what's actually changing, resin, color, diameter, or tooling, using the constraints configured for that line. That gives you exactly the input an honest Availability calculation needs: real changeover time, attached to the transition that produced it, instead of a single plant-wide guess averaged over everything.

If your OEE number keeps moving in ways nobody can explain, changeover variance is the first place to look, and having that number attached to the actual sequence of the week (not folded into run time, not flattened into an average) is usually what turns "productivity dropped" into "we ran four expensive color changes back to back, and here's why."