OEE tells you how close your output came to the ideal. Process reliability, in its best-known form from Paul Barringer, tells you how often daily output holds the plant's own demonstrated level. The term has no single standard definition. A plant can have a higher OEE and a lower process reliability than its neighbor, so the two answer different questions.
First, which process reliability?
Unlike OEE, this term means different things to different people. Before you compare it with anything, find out which one is on the table.
| Who | What it means there | Data needed |
|---|---|---|
| Barringer Process Reliability, as set out by Accendo Reliability, a training provider | The probability that the plant is in its demonstrated capability mode | Daily production only |
| PTC's software guide, a vendor | The highest reliability point at which the data shows the process is under control | Production data |
| Everyday plant speech | A process that runs without trouble | None, so ask what number is meant |
It also differs from classic machine reliability, which is about one item surviving between failures. That is the world of reliability-centered maintenance and the MTBF calculator.
The difference in one line each
OEE: how much good output did we get, compared with what the ideal rate allows in the scheduled time?
Process reliability: how much of the time did the plant behave like its usual self, and how badly did it behave the rest of the time?
OEE has an outside yardstick, the ideal rate. Process reliability measures the plant against its own record, so it can read high in a plant running far below design.
Two plants, one design rate
Example: two plants each have a design rate of 1,200 tonnes per day, and both are scheduled for 20 days. Assume all output is good.
| Plant A | Plant B | |
|---|---|---|
| Total output | 840 x 20 = 16,800 t | (1,140 x 16) + (300 x 4) = 19,440 t |
| OEE, against 24,000 t | 70.0% | 81.0% |
| Days at its own usual level | 20 of 20 | 16 of 20 |
| Process reliability, roughly | about 100% | about 80% |
| Lost on bad days | 0 t | (1,140 - 300) x 4 = 3,360 t |
| Gap from its usual level to design | (1,200 - 840) x 20 = 7,200 t | (1,200 - 1,140) x 20 = 1,200 t |
OEE says plant B is better, and it did make 2,640 more tonnes. Process reliability says plant B is the one in trouble, and that is also true.
These plants need opposite work. Plant B should find out why it crashes, because fixing that alone would lift it to 95.0%, while plant A has a steady process stuck at a low level.
How the Barringer method works
The real method fits a Weibull line to the daily data, so our 16 of 20 is only a rough stand-in. Accendo's introduction says it needs only production data, which is a large part of its appeal.
Good production per day, bad days included. Accendo's example uses a full year.
Sort from low to high, give each day a median rank, and plot it on Weibull axes.
Steady days form a straight line, and crash days bend away from it below a clear break.
The break gives process reliability. The gap to the line on crash days gives the reliability loss.
The median rank for a day in position i of n days is (i - 0.3) ÷ (n + 0.4), which is a standard plotting position. On plant B's 20 days, the 5th lowest day has a rank of 4.7 ÷ 20.4 = 23.0%.
The slope of the line matters too, since a steep line means tight control. Accendo's own example has a process reliability of 92%, yet shows about 1,005 tonnes a day against a design of 1,500, and it calls that plant a poor performer.
What each one can't see
- OEE can't see variation. A month of steady 81% and a month of wild swings averaging 81% give the same score. A plan or a customer feels the difference.
- OEE needs a sound ideal rate. On a process line it can be argued over, as OEE for continuous processes discusses.
- Process reliability can't see a low standard. Plant A reads close to perfect while it leaves 7,200 tonnes on the table.
- Process reliability can't name a cause. It works at plant level, so it tells you that bad days happen, not why they do.
Which to use where
| Your situation | Start with |
|---|---|
| You run a discrete line and work shift by shift | OEE and its loss breakdown |
| You run a continuous plant and have years of daily output | Process reliability for the strategic picture, then OEE on the key units |
| You need to size a project before funding it | Both: crash-day losses from one, loss minutes from the other |
| You need a daily number the crew can act on | OEE, since a Weibull plot is not a shift tool |
The two work well in sequence. Process reliability picks out the bad days, and then the OEE loss tree and a root cause analysis explain them.
A simple first test needs no software. Sort a year of daily output in a sheet, and look for the point where the numbers fall away.
Process reliability flags the bad days, and someone then has to explain them. Fabrico shows OEE per line and shift, with stops and speed from the machine and the cause tagged by operators. See how Fabrico shows OEE per shift.
What people ask next
What is process reliability in manufacturing?
It has no single definition. The best-known form, Barringer Process Reliability, is the probability that a plant is in its demonstrated capability mode, read from a Weibull plot of daily output. Elsewhere it may just mean a process that runs without trouble.
Is process reliability the same as OEE availability?
No. Availability is the share of scheduled time a machine ran. Process reliability looks at the spread of output from one period to the next, so a plant with many short stops each day could still score high. For the first, see the availability calculator.
Can a plant have high OEE and low process reliability?
Yes. Plant B above reads 81% OEE and about 80% process reliability, because it is close to design on good days and well below it on bad ones. A steady plant can show the reverse.
How do I calculate process reliability in Excel?
Sort daily output from low to high, give each day a median rank of (i - 0.3) ÷ (n + 0.4), and plot ln(output) against ln(-ln(1 - rank)). The break between the steady line and the low days marks process reliability. For serious use, check the fit in reliability software.
What data do I need for each one?
Process reliability, in the Barringer form, needs only daily output. OEE needs scheduled time, stops, total units, good units and the ideal rate. The OEE data collection guide sets out how to gather those.
Fabrico is a light MES for OEE. It takes stops, speed and counts from PLCs or IoT sensors, with add-on AI cameras counting output where a machine has no usable signal, and operators tag the cause at the line.