Planned vs. Unplanned Downtime: Why the Line Between Them Matters

Sooner or later, every plant that measures OEE has the classification meeting. It starts with an innocent question — "should the changeover count against Availability?" — and forty minutes later people who agree about everything else are arguing about the metaphysics of a lunch break.

The argument feels pedantic. It isn't. Where you draw the planned/unplanned line determines what your OEE number is capable of noticing for as long as you use it. Time excluded as "planned" doesn't show up as a smaller loss — it doesn't show up at all.

The principle that settles most of the arguments

Here's the test we recommend, because it turns a philosophical debate into a practical one: could a better-run version of this plant shrink this time?

If yes, it belongs inside planned production time, where OEE can see it as a loss. If genuinely no, exclude it.

Run the usual suspects through that test:

• Lunch and legally required breaks. No plant, however excellent, eliminates these. Exclude them.
• Changeovers. Emphatically shrinkable — SMED programs cut them in half routinely. Count them, as their own reason code. A plant that excludes changeovers has hidden what is often its single largest improvement opportunity.
• Cleaning and sanitation. The honest answer is "partly." The regulatory CIP cycle has a validated minimum duration; that floor is legitimate planned time. The gap between the validated minimum and what actually happens — late starts, slow turnarounds, missing chemicals — is a loss, and it's frequently substantial. If you can't split them, count the whole thing and accept a slightly harsher number over a blind spot.
• Preventive maintenance. Same logic. A scheduled PM window is planned; the two extra hours because the parts weren't kitted is a loss wearing a PM costume.
• No orders / no demand. Exclude it — an empty schedule isn't the line's fault — but track it somewhere, because "we don't count idle time" and "we quietly park every problem in idle time" are separated by about six months of habit.

Notice the pattern: the test doesn't produce a rule per category. It produces a rule per *reason*. "Planned" describes intent, not cost.

Unplanned doesn't mean unpredictable

A note on the other side of the line, because it carries its own trap. Plants treat unplanned downtime as random misfortune — the gods of machinery, appeased with overtime. But pull three months of stop reasons and most "unplanned" downtime turns out to be spectacularly predictable: the same feeder starving on the same product, the same pump seal every six weeks, material late from the warehouse most Mondays.

Predictable-but-unplanned is the most valuable category in your data, because it's pre-announced. The distinction worth drawing inside your unplanned bucket isn't "mechanical vs. everything else" — it's "first occurrence vs. repeat offender." Repeat offenders have root causes already half-diagnosed by their own history.

What this means for your reason tree

The practical output of all this is structure. A workable downtime tree separates, at the top level: breakdowns, changeovers and setup, material and supply waits, planned interventions (PM, sanitation), and idle/no-demand. Under each, a short list of specific reasons an operator can pick in five seconds.

Two rules keep it honest over time:

1. 1.Never let "other" grow past a few percent. "Other" is where classification systems go to die. When it swells, the tree is missing a reason the floor actually experiences — ask the operators what, and add it.
2. 2.Write the conventions down. One page: what's excluded from planned time, where changeovers count, how rework is treated. Half the value of the document is the arguments it prevents; the other half is that your numbers stay comparable when the supervisor who made the original decisions moves on.

The cost of getting it wrong

A last word on why this deserves an hour of management attention. We've seen the failure in both directions. Plants that exclude too much post proud numbers in the 80s while changeovers quietly consume a fifth of every shift — the metric literally cannot see the plant's biggest problem. And plants that classify too sloppily get numbers nobody trusts, which is worse, because distrusted data doesn't get used at all.

The line between planned and unplanned isn't bookkeeping. It's the decision about what you're allowed to notice. Draw it deliberately.