Startup Loss vs Reduced Yield Loss: The Two Quality Losses in OEE

Startup loss is the quality loss from defective units produced while a machine ramps up to stable running conditions, and reduced yield loss is the ongoing scrap and rework produced once the process is already running steadily. Both belong to the quality bucket that drives the third factor in Overall Equipment Effectiveness, but they have different causes, different fixes, and different owners on the shop floor. Confusing the two leads teams to chase the wrong root cause. This guide separates them cleanly, shows where each sits in the OEE calculation, and walks through the numbers so you can act on the right one.

The two quality losses inside OEE

In the classic Six Big Losses framework, two of the six sit under quality: startup rejects and production rejects. Both reduce the quality factor in Overall Equipment Effectiveness, which is calculated as Good Count divided by Total Count. The difference is when the defective units are made.

• Startup loss happens during ramp-up: the first units after a changeover, a warm-up from cold, a material change, or a restart after a stoppage. The process has not yet reached stable temperature, pressure, alignment, or speed, so early parts fall outside spec.
• Reduced yield loss happens during steady-state running: the drift, variation, and intermittent faults that produce scrap and rework even when the line is nominally warmed up and settled.

Both end up in the same quality number, but they answer to different questions. Startup loss asks "how long does it take us to make good parts," while reduced yield loss asks "how consistently do we make good parts once we are going."

What startup loss really is

Startup loss covers every reject produced between the moment production begins and the moment the process stabilizes. Typical drivers include tooling that has not reached working temperature, injection molds or ovens still equilibrating, print or coating heads clearing purge material, and operators dialing in settings by trial and error. It is closely tied to the effectiveness of your changeover and warm-up procedures.

Because startup loss is concentrated in a predictable window, it responds well to standardized work. First-article inspection, documented warm-up sequences, and a robust control plan shrink the ramp window. Where startup defects recur in a repeatable pattern, a structured 8D problem-solving effort or a DMAIC project can attack the setup conditions directly.

What reduced yield loss really is

Reduced yield loss is the steady-state cost of a process that is not fully capable. Even a warmed-up, running line produces some fraction of defects because of raw material variation, tool wear, environmental drift, and machine-to-machine differences. This is the loss you measure with your scrap rate once startup parts are excluded.

Reduced yield is a capability and variation problem, so the toolkit is different. Statistical process control catches drift before it becomes scrap, Cp and Cpk analysis tells you whether the process can even hold tolerance, and a Gauge R&R study confirms your measurements are trustworthy before you blame the machine. When defects cluster around a few causes, a Pareto analysis ranks where to spend effort first.

Worked example: separating the two losses

A packaging line runs an 8 hour shift and produces 10,000 units total. Assume 400 units are rejected in all. Here is how to split them.

1. Total rejects: 400 units.
2. Startup rejects: After the morning changeover and two mid-shift material swaps, the first parts of each run were out of spec. Tally: 240 startup rejects across the three ramp windows.
3. Reduced yield rejects: The remaining 160 rejects occurred during stable running, spread across the shift.

Now the quality factor. Good Count is 10,000 minus 400, which is 9,600. Quality equals 9,600 divided by 10,000, or 96 percent. That single number looks fine, but it hides the split: 240 of the 400 losses (60 percent) are startup, and only 160 (40 percent) are steady-state.

The lesson is that most of this line's quality loss is a ramp-up problem, not a capability problem. If the team had reacted to the 96 percent by launching an SPC and Cpk investigation, they would have spent weeks tuning a process that is already reasonably capable. The higher-return move is cutting the number of ramp windows and shortening each one, which targets the 240-unit startup slice. Had the numbers been reversed (160 startup, 240 steady-state), the priority would flip toward variation reduction.

Why the distinction changes what you do

Startup loss and reduced yield loss have almost no overlap in remedy, so mislabeling them wastes effort.

• Frequency vs magnitude: Startup loss scales with how often you ramp up (more changeovers, more restarts after breakdowns, more small batches). Reduced yield scales with how capable and stable the running process is.
• Owner: Startup loss is usually a setup, standard-work, and maintenance-reliability issue. Fewer unplanned stops mean fewer restarts, which is where a strong CMMS and proactive maintenance practice pay off. Reduced yield is usually a quality-engineering and process-control issue.
• Design defenses: Reduced yield often traces back to a design or process weakness, so preventive tools like FMEA and robust parameter design via design of experiments harden the steady state. Startup loss rarely benefits from those and instead wants better warm-up discipline.

Practically, every OEE program should log the reject reason and the production phase (ramp vs steady) at the point of scrap. Without that tag, both losses collapse into one quality percentage and you lose the ability to prioritize.

Where Fabrico fits

You cannot separate startup loss from reduced yield loss unless you capture accurate, real-time production counts with the reason and phase attached to each reject. Fabrico is the real-time data foundation that makes this possible. It delivers live OEE and production monitoring, so good and reject counts are recorded as they happen rather than reconstructed from memory at end of shift. On machines with no PLC, Fabrico can add computer vision to capture counts and states directly.

Because Fabrico is also a field-ready CMMS with work orders, asset records, preventive scheduling, and spare-parts tracking, it links the two losses to their real drivers: fewer unplanned breakdowns mean fewer restarts and less startup scrap, while cleaner asset condition supports a more stable steady state. Fabrico is EU-built with EU data residency, so the count data behind your quality metrics stays governed under EU rules.

Frequently Asked Questions

Is startup loss counted as a quality loss or an availability loss in OEE?

It is a quality loss. The machine is running and making parts during startup, so no availability time is lost; the parts are simply out of spec. Those defective units reduce Good Count and therefore lower the quality factor, not the availability factor. The time spent stopped for the changeover itself is a separate availability loss.

How do I tell startup rejects from steady-state rejects in practice?

Define a ramp window for each product or line, for example the units produced before the first passing first-article check or before a key parameter reaches its target range. Everything inside that window is startup loss; everything after is reduced yield. Tagging each reject with its phase at the moment of scrap is the only reliable way to keep the two separate over time.

Which loss should I attack first?

Measure the split before deciding. If most of your quality loss is startup, focus on reducing changeover frequency, standardizing warm-up, and cutting unplanned restarts. If most is reduced yield, focus on process capability and variation with SPC and capability studies. A quick Pareto view of rejects by phase usually makes the priority obvious.

Want to see exactly how much of your quality loss is ramp-up versus steady-state? Book a Fabrico demo to see real-time OEE with reject reasons and production phase captured automatically on your lines.