To eliminate OEE losses, treat each loss on its own: breakdowns need faster response and repeat-failure reviews, changeovers need a shorter setup routine, short stops need their causes recorded and removed one by one, slow running needs a correct ideal cycle time, and scrap needs to be traced to the moment it starts. Fix the largest loss on the bottleneck first, then measure for two weeks before you move on.
Which loss should you fix first?
OEE splits lost time into availability, performance and quality. The six big losses split it further: breakdowns (equipment failure) and setups lower availability, short stops (idling and minor stops) and reduced speed lower performance, and process defects and startup rejects (reduced yield) lower quality [1].
Before you fix anything, you need to know which of the six costs the most time on the machine that limits output. If you are not sure yet, start with why OEE is low and how to find the causes, which explains how to split OEE and confirm the main loss in about two weeks.
Two rules help you choose:
- Work on the bottleneck. An hour gained on a machine that waits for parts adds nothing to output.
- Work on one loss at a time. If you change three things at once, you will not know which one helped.
How do you cut breakdown time?
Breakdowns are unplanned stops caused by equipment failure [1]. Their cost has two parts: the repair itself and the time before anyone starts it.
Shorten the wait first. Make sure each stop reaches one named person, set a time after which the alert escalates, and measure the time from stop to response.
Then stop the same failure from coming back. Once a week, take the three failures that cost the most hours, find the root cause with the people who repair them and agree one action for each. Track mean time between failures and mean time to repair on those machines.
The DBR77 IIoT knowledge base covers the alert side in why machine alerts fail on the shop floor. For a full plan against unplanned stops, read how to reduce unplanned machine downtime on our blog.
How do you shorten changeovers?
Setup and adjustment time counts as a planned stop, and changeovers are usually its largest part. The standard method for reducing it is SMED (Single-Minute Exchange of Die) [1].
List every task in a changeover and mark which ones can only happen while the machine is stopped and which ones could be done while it is still running. Then move the second group outside the stop:
- Prepare tools, dies and material for the next product before the current batch ends.
- Pre-set what can be pre-set, for example fixtures or programs.
- Replace bolts and manual alignment with quick clamps and fixed stops where possible.
- Write the new routine down and train every shift on the same steps.
Measure the changeover from the last good part of one product to the first good part of the next.
See where your machines lose time, stop by stop. Open the DBR77 IIoT demo: a demo environment with sample data, machine statuses, alerts and operator views.
How do you remove short stops and slow running?
Short stops last a minute or two: a jam, a misfeed, a sensor that trips [1]. Most plants do not record them accurately [1]. Illustrative example: sixty one-minute stops in a shift add up to an hour.
Removing them takes three steps. Record them from the machine’s run and stop signal where it exists, because operators will not write down a 40-second stop. On older machines without a PLC, a simple status device the operator taps at each stop is enough to start. Give each one a reason from a short list. Then work on the most frequent reason, not the longest single stop.
Reduced speed is harder to see because the machine keeps running. It shows up only if you compare the actual cycle time with the ideal one. Typical causes are worn or dirty equipment, poor lubrication and material that does not meet specification [1]. Two checks help:
- Set the ideal cycle time to the best rate for that product. A value borrowed from the slowest product makes slow running look normal.
- Compare cycle time by shift and by operator. A large gap points to settings or habits you can standardize.
| Loss | OEE factor | Typical fix | How to check it worked |
|---|---|---|---|
| Breakdowns | Availability | Faster response, weekly repeat-failure review | Lost hours and mean time between failures |
| Setups and adjustments | Availability | SMED: move tasks outside the stop | Time from last good part to first good part |
| Short stops | Performance | Automatic recording, remove the most frequent reason | Number of stops per shift |
| Reduced speed | Performance | Correct ideal cycle time, standard settings | Actual against ideal cycle time |
| Startup rejects | Quality | Fixed startup routine, settings stored per product | Rejects in the first parts after a start |
| Process defects | Quality | Trace defects by code, shift and product | Reject rate during stable running |
The DBR77 IIoT knowledge base explains the measurement side in live OEE compared with OEE from a spreadsheet.
How do you reduce quality losses?
Quality losses come in two forms. Startup rejects appear after a start or changeover, before the process is stable. Process defects appear during normal running [1].
For startup rejects, store the right settings for each product and use the same startup routine on every shift. Count rejects by time since start; if most appear in the first minutes, the startup routine is the problem.
For process defects, record every reject with a defect code, the shift and the product. Look for patterns: one product, one shift, one machine after a long run. Fix the cause, then confirm with the next weeks of data. Count good parts after final inspection, otherwise rework coming back later will hide the loss.
How do you keep the gains?
OEE often improves for a few weeks and then drifts back. Three habits help keep the gain:
- A short daily review of the bottleneck’s losses with the operators who run it.
- Written rules for what counts as planned time and which ideal cycle time is used, so the number means the same thing every week.
- One owner per line who decides what to work on next.
When one loss is solved, the next largest becomes visible. A common reference point is 85% OEE, but most manufacturers score closer to 60% [2], so steady progress on the bottleneck is a better target than a single number.
If the losses are spread across many machines with no single cause, the problem may be the layout or the flow between stations. The DBR77 Digital Twin knowledge base shows how to test that in how to find production bottlenecks before they happen.
FAQ
What are the six big losses in OEE?
They are breakdowns, setups and adjustments, short stops, reduced speed, process defects and startup rejects. The first two lower availability, the next two lower performance and the last two lower quality [1].
What is SMED?
SMED (Single-Minute Exchange of Die) is a method for shortening changeovers. You separate tasks that need the machine stopped from tasks that can be done while it runs, move as many as possible outside the stop, and simplify the rest.
How do I reduce short stops if nobody records them?
Record them from the machine’s run and stop signal where it exists. On older machines without a PLC, a simple status device the operator taps at each stop is enough to start. Then ask operators to pick a reason from a short list for each stop and work on the most frequent reason first.
How long does it take to see OEE improve?
A single fix on the bottleneck, such as a shorter changeover, can show up within one or two weeks. Keep measuring for several weeks to make sure the gain holds.
Ready to see your losses in real numbers?
Start with the machine that limits your output. Try the DBR77 IIoT demo to see machine statuses, alerts and operator views on sample data, then talk to us about measuring losses on one of your lines.
