The Six Big Losses in Manufacturing, Explained

OEE tells you how much capacity you're losing. The Six Big Losses — a framework from the same Total Productive Maintenance (TPM) tradition that produced OEE — tell you where it went. Every hour of gap between your OEE and 100% lands in one of six buckets, two per OEE factor.

The point of the framework isn't taxonomy for its own sake. Each loss category has a different owner, a different fix, and a different data trail. A shop that knows its OEE is 62% knows it has a problem; a shop that knows 14 points of the gap is changeovers and 9 points is micro-stops knows what to do on Monday.

The two availability losses

Availability losses are the times the machine should have been producing and produced nothing at all. TPM splits them by cause:

1. Equipment failure (breakdowns): unplanned stops from mechanical or electrical faults — a spindle fault, a hydraulic leak, a drive trip. Usually the most visible loss and often not the biggest one. Data trail: downtime log with a reason and duration per stop.
2. Setup and adjustment (changeovers): planned in the sense that you knew it was coming, but a loss because the machine isn't producing — tool changes, fixture swaps, program loads, first-article checks, and the dial-in period after. Includes the sneaky tail: the machine is 'running' but you're still adjusting.

The classic mistake is treating changeovers as untouchable overhead. They're the most improvable availability loss in most job shops — SMED-style changeover reduction (externalize whatever can be done while the machine still runs, standardize the sequence, stage tooling in advance) routinely cuts setup time dramatically without capital spend. Breakdowns, by contrast, improve through boring consistency: a functioning preventive-maintenance calendar and operators empowered to flag early symptoms.

The two performance losses

Performance losses happen while the machine is nominally running but producing slower than its ideal rate. They're the least visible losses because nothing looks wrong:

3. Idling and minor stops (micro-stops): jams, misfeeds, a sensor that needs a wave of the hand, a part that needs reseating — stops short enough that nobody logs them, often under a minute each. Thirty of them a shift is a silent half hour.
4. Reduced speed: the machine runs, but below ideal cycle time — a conservative feed rate someone set years ago after a bad batch, worn tooling, marginal material, or an operator running cautious on a new job. Speed loss hides inside 'run time' and only surfaces when you compute Performance from actual counts.

Performance is the factor that surprises shops most on first measurement, precisely because neither loss generates an event anyone writes down. You don't find micro-stops in a downtime log; you find them by standing at the machine for an hour, or by noticing that Performance says 78% while the operator says 'it ran fine all day.' Both are telling the truth — that's the point.

Tip If Performance is your weak factor, spend one hour physically watching the machine before buying any monitoring hardware. Most shops can name their top two micro-stop causes after sixty minutes of honest observation.

The two quality losses

Quality losses are parts the machine spent time making that didn't come out right the first time:

5. Process defects: scrap and rework produced during steady-state running — an out-of-tolerance dimension, a surface defect, a missed operation. In strict OEE accounting, reworked parts count as losses even if they're later saved, because the machine time to make them right the first time was consumed.
6. Reduced yield (startup losses): scrap produced between startup and stable production — the first parts after a changeover, a cold-start warmup, or a material lot change, before the process settles into tolerance.

Separating these two matters because the fixes live in different places. Steady-state defects point at the process itself: tooling condition, machine capability, inspection timing. Startup losses point at the changeover and warmup procedure — and they scale with changeover count, which means a shop that wins on flexibility pays for it here unless first-article stabilization is deliberately engineered. If your batch sizes are shrinking, expect loss #6 to grow unless you act on it.

Using the framework without drowning in it

Map your existing shift data to the six buckets. Downtime log entries split into breakdowns vs setups (losses 1-2). The gap between run time and ideal-rate output is losses 3-4 combined. Scrap counts split into startup vs steady-state (losses 5-6).
Express each bucket in minutes per shift, not percentages. '55 minutes of changeover, 30 of speed loss, 12 of scrap time' is a to-do list; '71% OEE' is a mood.
Rank the buckets and attack the biggest one only. Six simultaneous improvement programs is zero improvement programs.
Pick the countermeasure that matches the bucket: PM discipline for breakdowns, SMED for setups, observation-and-fix for micro-stops, ideal-rate audits for speed loss, process capability work for defects, first-article procedure for startup scrap.
Re-rank monthly. The buckets shift as you improve — that's the framework working.

One caution: don't build a 40-code downtime taxonomy on day one. The six losses plus a handful of shop-specific reason codes under each is plenty. The framework earns its keep by directing attention, and attention doesn't survive a form with nine dropdowns.

Common questions

Where do planned maintenance and no-demand time fit?

Outside the six losses. The framework covers losses against planned production time. Time you deliberately excluded — scheduled PM, breaks, no scheduled demand — isn't an OEE loss at all, though it does show up in TEEP if you measure against the full calendar.

Is waiting for material a breakdown?

It's an availability loss, but give it its own reason code rather than burying it under 'breakdown' — the fix (scheduling, staging, purchasing) has nothing to do with maintenance. Many shops add reason codes under the six headline losses for exactly this.

Which loss is usually the biggest?

It varies too much by operation to generalize honestly. High-mix shops tend to bleed most in setups and startup scrap; older equipment bleeds in breakdowns and speed; highly automated lines bleed in micro-stops. That's the argument for measuring your own split rather than borrowing anyone's ranking.

Do I need the six-loss breakdown if I already track OEE?

OEE without a loss breakdown tells you the size of the problem but not its address. Even a coarse version — stop reasons on the downtime log and a startup-vs-steady-state scrap split — turns the same shift data into a prioritized work list.