What are the Six Big Losses?

The Six Big Losses are equipment failure (breakdowns), setup and changeovers, idling and minor stops, reduced speed, process defects (scrap and rework), and startup losses. They are the six categories of loss that reduce OEE.

How do the Six Big Losses map to OEE?

Two losses fall under each OEE factor. Breakdowns and changeovers reduce availability, minor stops and reduced speed reduce performance, and process defects and startup losses reduce quality.

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

Duration. A minor stop is a brief stoppage, often under about five minutes, that needs no maintenance, like a jam or a misfeed. A breakdown is a longer failure that takes the machine out until it is repaired.

SMED, or single-minute exchange of dies, is the standard method for reducing changeover time. It works in three steps: separate internal setup (done with the machine stopped) from external setup (done while it runs), convert internal steps to external where possible, then simplify the steps that remain internal.

What is usually the biggest of the Six Big Losses?

It varies by plant, but minor stops and changeovers are often the largest and most overlooked, because each minor stop is too small to log and changeover time is wrongly treated as fixed. Use a Pareto analysis to find your own biggest loss.

How do you prioritize the Six Big Losses?

Capture all six, then rank them with a Pareto or waterfall chart that shows how much each one subtracts from a perfect OEE. Attack the largest first, and do it on the bottleneck, since gains anywhere else do not raise output.

How do you reduce the Six Big Losses?

Measure where the loss actually sits, rank the losses by impact, and attack the biggest first with the right tool: SMED for changeovers, preventive and predictive maintenance for breakdowns, machine data for minor stops, and variation control for defects. Focus on the bottleneck.

Where did the Six Big Losses come from?

They come from Total Productive Maintenance (TPM), the framework that OEE grew out of. Splitting each OEE factor into two losses makes the causes specific enough to act on.

The Six Big Losses in OEE (and How to Cut Each One)

The Six Big Losses are the six categories that drag down OEE, two for each of its three factors. A single OEE number tells you that you are losing output, but not where it is going. Naming the loss is how you find the cause and pick the right fix. If you are new to the metric itself, start with how to calculate OEE, then use this to improve it.

What the Six Big Losses are

The framework comes from Total Productive Maintenance, where OEE originated. Each OEE factor, availability, performance, and quality, splits into two losses. The reason for the split is that the two losses under each factor usually have different causes and different fixes, so separating them is what makes the data actionable rather than just descriptive.

OEE factor	Loss	What it is
Availability	1. Equipment failure (breakdowns)	Unplanned stops from a failure
Availability	2. Setup and changeovers	Planned stops to switch products or adjust
Performance	3. Idling and minor stops	Brief stoppages and jams, no maintenance needed
Performance	4. Reduced speed	Running below the rated or ideal speed
Quality	5. Process defects (scrap and rework)	Defects produced during stable running
Quality	6. Startup losses (reduced yield)	Defects during warm-up and startup

The six losses, one by one

1. Equipment failure (breakdowns)

An availability loss, and the most visible of the six: an unplanned stop when a machine fails. It is also often not the largest, which is why it should not absorb all the attention.

Breakdowns split usefully into two kinds. Sporadic failures happen rarely but hit hard, and are usually easy to root-cause after the fact. Chronic failures are the small, repeated stops on the same asset that quietly add up and need more structured problem-solving to trace. On a line with several processes, it also matters where the breakdown sits: a failure on the bottleneck stops the line directly, while an upstream failure starves it and a downstream failure blocks it. Separating those tells you which asset to protect first.

Cut breakdowns by moving critical assets from reactive to preventive and predictive maintenance, catching developing faults early with condition monitoring, and root-causing the chronic repeat offenders so they stop recurring. The full playbook is in reducing unplanned downtime.

2. Setup and changeovers

An availability loss: the time lost switching from one product to the next. It feels fixed, but it is one of the most reducible losses in a plant, and the standard method for attacking it is SMED, or single-minute exchange of dies. SMED works in three moves.

Separate internal from external setup. Internal steps can only be done with the machine stopped; external steps can be done while it still runs. Simply staging tools, parts, and materials before the line stops cuts a surprising amount of downtime on its own.
Convert internal steps to external. Find steps that can be moved to before or after the stop, for example by pre-assembling a module that drops in quickly rather than being built in place.
Simplify the remaining internal steps. Standardize fasteners so fewer tools are needed, or replace them with quick-release mechanisms, so the stopped portion shrinks further.

Beyond SMED, sequence the schedule so similar products run back to back and fewer changeovers are needed at all.

3. Idling and minor stops

A performance loss, and usually the most overlooked: brief stoppages, jams, misfeeds, and sensor trips that each cost seconds and need no maintenance. They are overlooked because each event is too small to log, so they rarely show up in the data, yet together they are often the single largest performance drain in the plant.

The first problem is simply seeing them. Asking operators to log every thirty-second stop does not work; capturing minor stops takes automated machine data. Once they are visible, a lot of the causes turn out to be chronic and well known, operators can often tell you exactly what they fight every day, and a walk down the line while it runs reveals the rest. Others are subtler, especially flow-balance problems between machines, where one process starves or blocks the next because their speeds are not tuned to each other. Make the stops visible, then engineer out the recurring causes one at a time.

4. Reduced speed

A performance loss: running below the rated or ideal cycle time. The causes vary, worn equipment that no longer holds speed, poor-quality material that forces a slower rate, operators who are not comfortable running faster, or procedures that were never written to support full speed. Often the line settles at whatever speed gives the crew the fewest problems, which means the speed loss is really a symptom of something else.

Cut it by finding what operators are quietly compensating for and removing it, then writing standards robust enough that the right speed holds regardless of who is running the machine. Sometimes simply making the speed loss visible improves it, because people run closer to rated speed when they know it is measured. As with the other losses, concentrate on the bottleneck, since it sets the pace of the whole line.

5. Process defects (scrap and rework)

A quality loss: defects produced during stable running that have to be scrapped or reworked. Defects are fundamentally a variation problem. The more a process varies, the more it drifts out of spec, so reducing variation is the core of cutting this loss. Methodologies like Six Sigma and statistical process control exist for exactly this, but you do not need a formal program to start. Catch defects at the source rather than at final inspection, identify the few failure modes that produce most of the scrap, and find the process parameters that actually drive the quality you care about so you can hold them steady.

6. Startup losses (reduced yield)

A quality loss: the defects produced during warm-up and startup, before the process stabilizes. Every changeover and every cold start carries some of this, and it compounds with the changeover loss above. Cut it by stabilizing startup, using recipes and automated settings so the machine reaches good output faster and more repeatably, and by treating startup yield as part of your changeover improvement rather than a separate problem.

How to measure and visualize the losses

You cannot prioritize losses you cannot see, and the six are not equally easy to capture. Breakdowns and changeovers are visible; minor stops and speed loss usually are not, which is why automated data collection from machine controls matters more than manual logs for the performance losses.

Once captured, the most useful way to view the losses is together and ranked. A Pareto chart, or a waterfall chart showing how much each loss subtracts from a perfect OEE, makes it obvious which one is actually costing the most. That ranking is the whole point: it tells you which loss to attack first, instead of spreading effort evenly across all six.

A method for reducing the losses

Naming and ranking the losses tells you what to work on. A simple improvement loop tells you how. The Plan-Do-Check-Act cycle, also called PDSA, is the standard.

Plan. Pick the top loss from your data and form a theory of what causes it.
Do. Make the change on a small scale.
Check. Measure whether the loss actually fell.
Act. If it worked, lock it in and move to the next loss; if not, revise the theory and try again.

Run it continuously. OEE improvement is rarely one big fix. It is many small ones on the losses that matter most, repeated.

Mistake-proofing the losses

A lot of changeover time and startup scrap traces back to small human errors: the wrong setting, a missed step, a part installed backwards. Rather than asking people to be more careful, design the process so the error cannot happen. This is mistake-proofing, or poka-yoke: interlocks that stop a machine running with a guard open, fasteners that only fit one way, settings that load automatically from a recipe so a changeover is repeatable. The principle is to fix the system, not blame the person, because even when the root cause is human, the durable fix is almost always a change to the process.

Where to start

Put it together and the order is straightforward. Capture the losses, including the minor stops and speed loss that manual tracking misses. Rank them with a Pareto so the biggest is obvious. Attack that loss with the right tool: SMED for changeovers, predictive maintenance for breakdowns, machine data for minor stops, variation control for defects. Do it on the bottleneck first, because gains anywhere else do not raise output. Then run the loop again. For a worked version in a specific setting, see improving OEE in discrete manufacturing.

Common mistakes

Chasing a single OEE number. The headline figure hides which loss is the problem. Always work from the loss breakdown.
Treating every loss equally. Effort spread evenly leaves the biggest loss under-attacked.
Ignoring minor stops. They are unglamorous and unlogged, yet usually the biggest performance loss.
Improving a non-bottleneck machine. Gains anywhere but the constraint do not increase output.
Blaming operators instead of the process. When the same error keeps happening, the fix is mistake-proofing the system, not asking for more care.

From naming the losses to cutting them

Naming the Six Big Losses tells you where OEE is going. Seeing which loss is costing the most, on which line, and what to do about it, across machine controls, MES, and quality systems, is the work that usually stays manual.

SteelTree connects to those systems and turns the loss data into decisions: which of the six is draining the most recoverable output, on which constraint, and the next action to recover it, with the reasoning attached. You keep your existing systems. SteelTree sits on top as the decision layer.

See how SteelTree turns operational data into decisions →