What are the six big losses in OEE?

Equipment failure (breakdowns), setup and adjustment, idling and minor stops, reduced speed, startup rejects (defects during ramp-up), and production rejects (defects during stable running). The framework comes from Seiichi Nakajima's Total Productive Maintenance work in the 1980s and remains the standard way to decompose OEE losses.

How do the six big losses map to OEE factors?

Two losses per factor. Breakdowns and setup/adjustment reduce Availability. Idling/minor stops and reduced speed reduce Performance. Startup rejects and production rejects reduce Quality. Because OEE multiplies the three factors, closing any single loss raises the overall score by arithmetic.

What is the difference between a breakdown and a minor stop?

Duration and who resolves it. A breakdown is an unplanned stop long enough to log and usually needs maintenance. A minor stop (micro-stop) lasts seconds to a couple of minutes and is cleared by the operator, a blocked sensor, a misfeed, a starved infeed. Minor stops rarely reach paper logs, which is why they hide inside the Performance factor.

Which of the six losses is usually the biggest?

It varies by line, and the ranking almost never matches intuition. Breakdowns feel biggest because they are dramatic, but once losses are converted to minutes, minor stops and setup/adjustment frequently out-cost them, especially on high-speed or changeover-heavy lines. The only reliable answer is your own Pareto in minutes.

Do changeovers count as one of the six big losses?

Yes, setup and adjustment is the second loss, and it counts against Availability even though the changeover was planned. The countermeasure is SMED: separate internal from external setup work, convert internal steps to external, and standardize settings so adjustment time disappears.

How do you reduce the six big losses?

Match the countermeasure to the loss: planned and condition-based maintenance for breakdowns, SMED for setups, stop detection plus operator-led equipment care for minor stops, restoring true cycle time for speed losses, standardized startup procedures for startup rejects, and source quality controls like SPC and poka-yoke for production rejects. Attack them in Pareto order, one owner per countermeasure.

The Six Big Losses: Mapping Downtime to OEE

Breakdowns, setups, minor stops, reduced speed, startup rejects, production rejects, what each one costs, which OEE factor it hits, and the countermeasure that works.

The six big losses are the classic categories of equipment-based production loss: breakdowns, setups and adjustments, idling and minor stops, reduced speed, startup rejects, and production rejects. Each maps to one OEE factor, the first two hit Availability, the next two hit Performance, the last two hit Quality.

OEE tells you how much you lost. The six big losses tell you where. The framework comes from Seiichi Nakajima's Total Productive Maintenance work at the Japan Institute of Plant Maintenance in the 1980s, and it has survived four decades because it does one thing well: it turns a percentage into a to-do list. This post maps each loss to its OEE factor, gives the standard countermeasures, and shows how to rank the losses so you attack them in order.

What are the six big losses?

The six losses, each with its OEE factor and its usual disguise on the floor:

Equipment failure (breakdowns), Availability. Unplanned stops long enough to log: seized bearings, drive faults, jams that need maintenance. The loss everyone already sees.
Setup and adjustment, Availability. Changeovers, tooling swaps, and the dial-in time after them. Planned in the sense that you chose the changeover; a loss in the sense that the line isn't producing.
Idling and minor stops, Performance. Micro-stops under a couple of minutes: a sensor blocked, a misfeed cleared by the operator, a starved infeed. Individually trivial, collectively enormous, and invisible on paper logs.
Reduced speed, Performance. Running below true ideal cycle time: worn tooling, tentative rates after a quality scare, "we always run it at 80% because it jams at full speed."
Startup rejects, Quality. Scrap and rework produced between startup and stable running: first pieces after changeover, warm-up scrap, purge material.
Production rejects, Quality. Defects made during steady-state running, the ones that show up in first pass yield.

The classic mapping: two losses per OEE factor. If a loss on your floor doesn't fit one of the six, your categories, not the framework, need adjusting.

What is the countermeasure for each loss?

Each loss has a known playbook. None of them start with buying a new machine:

Breakdowns → planned maintenance. Move failure modes from surprise to schedule: a real PM program basic condition monitoring on the worst actors, and root-cause work (5 whys) on repeat failures so fixes stop being repairs. Track MTBF to confirm the failures are actually spacing out. The U.S. Department of Energy's Federal Energy Management Program O&M guidance estimates plants heavy on reactive maintenance can find savings opportunities exceeding 30–40% by shifting to planned approaches, this loss is where that money lives.
Setup and adjustment → SMED. Separate internal from external setup work, convert internal to external, and standardize the sequence. Halving changeover time is a routine SMED outcome, and adjustment time disappears when settings are documented instead of dialed in from memory.
Idling and minor stops → make them visible, then fix the top three. Micro-stops don't reach paper logs, so the countermeasure starts with automatic stop detection, then operator-led care of the sensors, guides, and feed mechanisms where micro-stops breed.
Reduced speed → restore, then challenge. First restore the equipment to run its true ideal cycle time (worn parts, buildup, alignment); then challenge informal derates. Every "we always run it slower" hides either a fixable defect or a wrong standard.
Startup rejects → standardize the ramp. Documented settings per product, first-piece verification, and startup procedures that reach stable state in a defined number of cycles, not "when it looks right."
Production rejects → quality at the source: statistical process control on the variables that drive the defect, poka-yoke where human error is the mode, and fast feedback so the 200th defect isn't the first one anyone sees.

How do you rank the losses on your line?

Convert every loss to time, then Pareto it. The six categories only pay off when they carry minutes: each downtime event coded to a loss, micro-stops and speed losses computed from counts against ideal cycle time (method in the OEE calculation guide), rejects converted at ideal cycle time per unit. Time is the common currency that lets a changeover problem be compared to a scrap problem on one chart.

A hypothetical loss Pareto in minutes per week. Micro-stops topping the chart is common once measurement gets honest, they were always there, just never logged.

Two patterns repeat across plants. First, the ranking almost never matches the plant's intuition, breakdowns feel biggest because they are dramatic, but minor stops and setups frequently out-cost them. Second, the top two losses usually carry half or more of the total, so the improvement plan is short: one countermeasure per top loss, one owner per countermeasure, reviewed weekly. This is the same discipline as a downtime program extended to losses that never show up as stops.

What does a weekly loss review look like?

Twenty minutes, one page, same agenda every week. The Pareto does the talking:

Minutes 1–5: the trend. This week's total loss minutes against the last eight weeks, split by the six categories. No narrative yet, just whether the number moved.
Minutes 5–15: the top two losses. For each: what were the biggest individual events or patterns, what countermeasure is open, did it advance, what does the owner need. Repeat offenders get root-cause treatment, not another line in the log.
Minutes 15–20: decisions. Close countermeasures that worked, kill ones that didn't, and pick at most one new item. A loss review that opens five actions a week closes none.

Two rules keep the meeting honest. Operators who logged the stops should see how their codes got used, nothing kills data quality faster than logging into a void. And the categories stay fixed: when a loss seems to fit nowhere, the answer is a better reason code within the six families, not a seventh family. The framework's power is that every plant, shift, and line can be read in the same six columns.

How does this connect to your OEE number?

Directly: close a loss and the corresponding OEE factor rises by arithmetic, not hope. On the hypothetical Pareto above, the 340 minutes of weekly minor stops on a line with 2,250 planned minutes is roughly 15 points of Performance; halving setups returns about 6.4 points of Availability. Pricing improvements this way, through the OEE calculator or by hand, is how you choose between an SMED workshop and a PM overhaul with numbers instead of enthusiasm.

The loss framework is also the honest test of your data collection. If you cannot fill in all six categories with real minutes, the gaps mark what your current logging cannot see, usually minor stops and speed. Plants that instrument those gaps, feeding stop detection and counts straight off the machines into live dashboards the way Harmony's downtime intelligence does (see the platform), typically discover their true loss ranking within weeks. The losses were never hiding. The clipboard just couldn't see them.