The 6 big losses: what your manual log doesn’t see

A solid framework, but only half filled in

The six-big-losses model is one of the most robust in continuous improvement. Born with TPM, it structures OEE analysis by breaking down each source of yield loss. Its logic is unassailable: to improve a performance, you first have to know where it is lost. The framework remains perfectly valid today, and that is precisely why it deserves a close look.

The problem is not the framework; it is the data you put into it. When you fill the six categories from manual logs, you actually populate only the ones that are visible: breakdowns and long stops. The others stay empty or approximate, not because they do not exist, but because nobody has been able to measure them. The framework is solid, but you only fill it in halfway. Note that OEE is the English term; the French equivalent, TRS, refers to exactly the same indicator.

The six losses, one by one

The first two families are the availability losses. Breakdowns cover the long stops requiring an intervention. Setups and changeovers cover the downtime to switch from one production to another. These two losses are the most visible, so the best tracked, but they tell only part of the story.

Then come the performance losses: micro-stops, those short stops cleared without maintenance, and slowdowns, meaning production below the nominal pace. Finally the quality losses: scrap and rework on one hand, start-up losses on the other, when the first parts after a launch are not conforming. Six families, three components of OEE: availability, performance, quality (OEE = Availability x Performance x Quality).

Where the manual log breaks down

The manual log excels at long, striking losses. An hour-long breakdown, a forty-minute changeover: these events are long enough to be noted, attributed and recorded. Conventional tracking captures them correctly, and it is on them that most action plans concentrate.

But it breaks down on short, frequent losses. The few-second micro-stops, the diffuse under-pace after a restart, the small unrecorded hiccups: all of this escapes human entry, by simple physical impossibility. Yet these are often the losses that represent the largest recoverable share of OEE. So improvement gets steered on the best-seen losses, while the costliest ones are ignored, simply because no human entry could ever capture a loss that lasts a few seconds and repeats relentlessly.

The paradox of invisible losses

There is a troubling paradox in the everyday practice of OEE: effort concentrates where you can see, not where it counts. Because breakdowns are visible, you launch reliability projects. Because micro-stops are invisible, you ignore them, even though they often weigh more. Visibility steers the action, and invisibility protects the biggest losses.

This bias is not irrational: you cannot act on what you do not measure. As long as the short losses stay in the blind spot of the manual log, they will keep escaping improvement plans, whatever the quality of the analysis framework. It is the data, not the model, that has to change, and that is exactly what automatic measurement does without touching the proven analysis framework.

Completing the framework with automatic measurement

Automatic measurement does not replace the six-big-losses model: it finally fills it in completely. A sensor fitted on the machine records every event to the second and automatically classifies the losses by category, frequency and duration. The six families stop being half-empty boxes and become a complete, quantified map of performance.

You then get the view that was missing: not just the visible tip of the iceberg, but the submerged mass too. You see the real share of each family in the OEE loss, and you often discover that micro-stops and slowdowns weigh more than breakdowns. This completeness radically changes how actions are prioritised, because the biggest recoverable share finally appears next to the losses everyone already knew about.

From descriptive analysis to prioritised action

A framework filled with real data finally lets you move from description to action. Instead of spreading effort uniformly or following intuition, you apply the Pareto principle: you identify the two or three dominant loss families and concentrate the effort where the gain is highest. Data turns an analysis model into a battle plan.

This quantified prioritisation avoids the classic trap of improvement initiatives: dispersing energy on secondary losses because they are easier to see. With continuous measurement, you act on the losses that really count, verify the gain in real time, and adjust. Hutchinson improved its OEE from 42% to 75% with the same headcount and machines, sensor installed in under an hour.

This approach also changes the rhythm of continuous improvement. Where a manual diagnosis demanded weeks of approximate logging before hazarding a priority, automatic measurement provides the complete map in a few days. You then chain improvement cycles faster, run several in parallel on different lines, and each action rests on fresh data rather than a memory. The six-big-losses framework, long confined to after-the-fact analysis, becomes a day-to-day steering tool.

A common language for the whole plant

The six-big-losses model has another, often underestimated merit: it provides a common language. When every line, every team and every site classify their losses by the same categories, fed by the same automatic measurement, comparison becomes legitimate. You finally speak the same language from one workshop to the next.

This standardisation is a major lever for multi-site groups. It reveals where the real gaps are, lets you identify best practices and spread them. Without a common measurement base, comparing the OEE of two plants stays an exercise in interpretation; with it, it becomes a decision grounded in facts, and a best practice found on one site can be transferred to the others with confidence.

Putting it in place without a heavy project

Completing the six-big-losses framework with automatic measurement does not require an MES project or a heavy investment. The measurement layer is fitted in under an hour, with no production stop, on old machines as well as new ones. The first data, classified by loss family, is usable within 48 hours.

The most effective approach is to start with a representative pilot line, let the measurement fill in the framework for a few weeks, then compare the real map to what the manual log was saying. The gap is usually telling, and it is often enough to decide to extend the approach. A free 60-day pilot lets you do this with no risk.

This gradual approach is also what reassures teams. You do not impose a big system to learn overnight: you instrument one line, observe together the map of the six losses as it really appears, and draw concrete actions from it. Trust is built on the first measured results, not on a theoretical promise, and the extension to other lines follows naturally, line by line, as each one proves the same recoverable gap between the declared map and the measured one.

Key points to remember

The six-big-losses framework still holds, but the manual log only fills it in halfway: it captures breakdowns and changeovers, and misses micro-stops and slowdowns, often the costliest. Automatic measurement completes the framework by classifying each loss by category, frequency and duration. You then move from a partial analysis to a quantified prioritisation, with a common language for the whole plant. More than 450 plants across 30+ countries already monitor their OEE to the second with TeepTrak.

FAQ

What are the 6 big OEE losses?
Breakdowns, setups and changeovers, micro-stops, slowdowns (under-pace), scrap and rework, and start-up losses. They split across the three components of OEE: availability, performance and quality.

Why is the manual log insufficient?
It captures long, visible losses (breakdowns, changeovers) well but misses the short, frequent ones (micro-stops, under-pace), which are often the costliest and the most recoverable.

How do you get a full view of the six losses?
With automatic measurement that records every event to the second and classifies losses by category, frequency and duration. The framework is then fully filled in, not just its visible part.

Which loss should you start acting on?
On the two or three dominant families revealed by the measurement, following the Pareto principle. Very often, micro-stops and slowdowns offer the biggest gain, with no capex.

Do you need an MES project for this?
No. The measurement layer is fitted in under an hour, with no production stop, and provides data classified by loss family within 48 hours, on old machines as well as new ones.