Availability, Performance, Quality: Where Losses Actually Hide
Two lines can post the same 60% OEE for completely different reasons — and need completely different fixes. Reading the three factors separately is where the diagnosis starts.
A single OEE number is like a single temperature reading: useful for spotting that something's wrong, useless for saying what. Two lines can both sit at 60% — one because it breaks down constantly, the other because it never stops but crawls. Same score, opposite problems, opposite fixes.
The diagnosis lives in the three factors. Here's how to read each one, and — just as important — how losses sneak between them.
Availability: the honest-looking one
Availability is downtime, and downtime feels like the most objective of the three: the machine is either running or it isn't. But Availability has two classic blind spots.
The first is the threshold. Most plants only log stops longer than some cutoff — two minutes, five minutes. Everything shorter simply vanishes from Availability and resurfaces as a mysterious Performance loss. A line with "great Availability and terrible Performance" often just has a stop-logging threshold that's too high, and a hundred 40-second jams a day slipping under it.
The second is the planned-time boundary. Every minute you exclude from planned production time — breaks, cleaning, changeovers if you've made that choice, "scheduled" anything — is a minute Availability can never flag. The boundary decisions get made once, in a meeting, and then quietly shape the number forever. Worth re-examining yours once a year, because plants accumulate exclusions the way garages accumulate boxes.
When Availability is your weak factor, look at the reason breakdown before assuming it's a maintenance problem. In our experience the biggest single bucket is often changeovers or material starvation — organizational losses, not mechanical ones. The fix lives in scheduling and logistics, not in the maintenance backlog.
Performance: where everything hides
Performance compares what you made against what the rated speed says you should have made. It's the least glamorous factor and, on most lines, the biggest and least understood loss.
Performance losses come in two flavors. Reduced speed — the line deliberately or gradually running below rated, because of a difficult product, worn tooling, or an operator who's learned that flat-out means jams. And micro-stops — the endless little halts too short to log, which we've covered in detail in the piece on micro-stops.
The thing to internalize: Performance is a residual. It's calculated, not observed. Nobody logs a Performance loss; it's just the gap left over after counts and rated speed do their arithmetic. Which means everything that goes unrecorded anywhere else lands here. Untracked stops, an optimistic rated speed, a miscounted batch — all of it shows up as "the line ran slow," even when it didn't. A Performance number that swings wildly shift to shift usually isn't the line changing. It's the recording changing.
That's also why the rated speed deserves more care than it gets. Set it from the machine's actual demonstrated capability, per product, and leave it alone. A rated speed that drifts down to match reality is Performance's version of grade inflation.
Quality: small number, expensive minutes
Quality is usually the highest of the three factors — 97, 98, 99% — and it's easy to read those numbers as "not our problem." Two reasons to look closer.
First, quality losses are the most expensive per unit. A bottle rejected at the end of the line carries the cost of the material, the fill, and the machine time that produced it — value already added, then thrown away. A point of Quality typically costs more than a point of Availability.
Second, Quality losses cluster where you'd predict: startups and changeovers. The first minutes after any stop produce the worst product — fills drifting into spec, seals stabilizing, temperatures settling. If your rejects spike after every changeover, your Quality problem is actually a changeover problem, and fixing the setup procedure will do more than tightening inspection ever will.
Watch the rework convention too. Product that gets re-run should count as a loss the first time through. Plants that count eventual yield instead of first-pass yield report lovely Quality numbers while paying twice for the same bottle.
Read them together
The factors leak into each other constantly — raise the stop-logging threshold and Availability improves while Performance decays; lower the rated speed and Performance improves while nothing changes at all. The classification matters less than the consistency. Pick your conventions, write them down, apply them every shift.
Then stop staring at the OEE number and start reading the three factors like a story: this is where the time went. On any given line, one of the three is the villain. Find it, name the top two losses inside it, and you've turned a percentage into a to-do list.
