In process manufacturing, measure OEE at the stage that limits the plant, using mass or volume as the unit: OEE = (good output / ideal rate) / scheduled time. Batch vessels and continuous units use the same formula with a different ideal. Do not chase OEE at stages with spare capacity, because a low number there is often just waiting.
Why process plants read OEE differently
A discrete line makes countable things one after another. A process plant turns material into more material, through stages that run at different speeds and hold stock between them.
So the hard question is not the formula. It is where to apply it, and what a low number at each stage actually means.
Three decisions to make first
1. What type of process is this stage?
Most plants have more than one type under the same roof. The type decides how you set the ideal rate.
| Stage type | Unit | Ideal comes from | Details |
|---|---|---|---|
| Batch: reactor, mixer, fermenter | kg or liters | Nominal batch size / shortest proven batch cycle | Batch guide |
| Continuous: kiln, dryer, column | tons or m3 | Best proven hourly rate per product family | Continuous guide |
| Packing: filler, bagger | Packs, or kg for mixed sizes | Best proven rate per format | Beverage line example |
2. Which stage sets the plant's output?
Find the stage that limits what the plant can ship. Bottleneck analysis shows how, and in a process plant the sign is often a tank that is always full in front of a stage and always empty after it.
That stage gets the headline OEE. The others get their own figures, read as explanation.
3. What unit lets you add products together?
Use kilograms or liters where densities are close, and hours at the ideal rate when grades run at different speeds. Good output divided by the ideal rate gives ideal hours, and hours can always be added.
Batch counts look tidy but they break as soon as batch sizes vary.
Worked example: a mixer feeding a filler
Example: a detergent plant runs one mixing stage and one filling line, both scheduled for 120 hours in a week. The mixer's ideal rate is 2,500 kg an hour, and the filler's is 2,000 kg an hour.
The mixer made 216,000 kg of on-spec product in the week. The filler packed 210,000 kg of good product.
A manager who ranks the two would attack the mixer first. That is the wrong move here, since the mixer could never run flat out with a slower filler behind it.
At the filler's 2,000 kg an hour, even a perfect week ships 240,000 kg. The mixer could reach 300,000 kg, so at least 20% of its time is spare capacity that shows up as waiting.
The useful readings are these. The filler has 15 hours of real loss to attack, and the 6,000 kg gap between stages is either stock left in tanks or a yield question for production yield.
What differs from a discrete line
| On the floor | How to treat it |
|---|---|
| Recipe hold, reaction or cure time | Part of the ideal cycle, not a loss. Only time beyond the recipe counts. |
| Product or grade transition | Stop time, plus a quality loss for material made during it. |
| Off-spec sold at a lower grade | Not good output at the grade you planned. Write the rule down. |
| Rework blended back in | Not good the first time. Do not count the same kilograms twice. |
| Downstream tank full | A blocked stop at this stage, with the cause named. See upstream and downstream stops. |
| No orders for this product | Excluded from scheduled time when planned. See OEE with no orders. |
Where to get the data
You seldom need a new sensor. Most of this already sits in a control system, a weigh scale or a batch record.
- Output: flow meter totalizer, weigh scale, or tank level change plus transfers. Reconcile against shipped or packed quantity at least weekly.
- Run state: a signal from the control system, or a current sensor on an older motor.
- Quality: lab release results. Check that the release date does not move output into the wrong week.
- Stops: a log with one cause per stop, using reason codes that name process events such as "no steam" or "waiting for lab".
What OEE leaves out of a process plant
OEE only sees time, rate and good output. Several things that decide a process plant's margin sit outside it.
- Raw material yield. You can hit the same OEE while using more feed per kilogram. Track yield next to it.
- Energy per unit. Running a dryer harder can raise OEE and raise cost per ton more. Compare with cost per unit.
- Demand-led plants. When the market sets the rate, a stage can be perfectly run and look poor. Split out the time no one needed the output, as OEE vs TEEP explains.
- Utility units. Boilers and chillers follow what the plant asks for, so OEE is a poor measure for them. Log their failures as causes at the stages they stop.
How high should it be?
Hansen sets world-class bars by process type, building on Nakajima's TPM figures: 85% or more for batch and discrete, 90% or more for dedicated discrete and non-dedicated continuous, and 95% or more for dedicated continuous. Read is 85% realistic before you use one as a target.
We found no neutral published benchmark for process plants as a group. Your constraint's own trend is the better guide.
Four errors that skew the number
- Averaging all stages. A plant OEE built from every vessel mostly measures spare capacity.
- Multiplying stages together. Stages with buffers between them do not lose output like machines in tight series.
- Setting the ideal from the nameplate alone. Use what the stage has proven, as in ideal run rate.
- Ignoring stock changes. Output into a tank is real output for that stage, even if nothing shipped.
Process stages already hold the signals you need. Fabrico takes stops, speed and counts live from PLCs or IoT sensors, with OEE per line and shift, and operators tag the cause at the line. See how Fabrico connects to machines.
What people ask next
Is OEE used in the process industry?
Yes, on batch, continuous and packing stages alike. It is most useful at the stage that limits output, as in the continuous and batch guides.
How do you calculate OEE without a part count?
Use mass or volume and an ideal rate. Divide good output by the ideal rate to get ideal hours, then divide by scheduled hours. Check your setup with the OEE calculator.
Should I calculate one OEE for the whole plant?
Report the constraint as the headline and keep the other stages as supporting numbers. If you must combine lines, weight by scheduled time, as shown in how to average OEE across machines.
Is yield part of OEE quality?
Only partly. Off-spec output counts against quality, but using too much raw material to make on-spec output does not. Track production yield as its own number.
What is the difference between OEE and TEEP in a process plant?
OEE measures against scheduled time, and TEEP measures against all calendar time. A plant that runs 24 hours a day often watches both, as explained in OEE vs TEEP.
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.