How much a food and beverage line really produces, and how much more it could
We bring together what 429 studies of real plants measured and compare it with the limit set by the physical laws of each process. Every number links to its source.
- 67produce
Typical of the plants studied: the equivalent of 67 hours of good product at full speed (an OEE of 67 %).
- 33are lost
To stops, slow running and defective product. The split is shown below.
Where do the lost hours go?
- Breakdowns27 %
The machine stops because something fails and has to be repaired.
- Reduced speed24 %
The line runs, but slower than it can.
- Minor stops22 %
Short stops the operator clears without maintenance: jams, waiting, blocked sensors.
- Changeovers and adjustments19 %
Time between one product and the next: format change, cleaning, warm-up and fine tuning.
- Defects6 %
Product that comes out wrong and is scrapped or reworked.
- Start-up losses2 %
Product lost while the line settles after starting.
Split of the lost time among the six big losses, estimated from 195 studies. The percentages add up to 100 % of the lost time.
What is OEE?
OEE stands for overall equipment effectiveness and is the standard measure in industry. An OEE of 100 % would mean making only good product, at maximum speed, without ever stopping during the planned time. It is the product of three factors: availability (how long the line ran), performance (how fast it ran) and quality (how much product came out right).
Author: Byron Cortés, Selva Systems LLC, San José, Costa Rica. Published by The Marketplace.
Choose your process
Depending on the process, physics allows between 75 and 96 out of every 100 hours. Point at a process to see its summary and open it to walk through the whole line.
What you will find for each process
How far it can go
The ceiling set by physical laws at each station of the line. Not even the best plant in the world exceeds it.
Where plants are today
What published studies measured on real lines, with its margin of uncertainty.
What stops them
The specific causes the studies report, ranked by how often they appear.
Where the numbers come from
Nothing is opinion or a rough guess. Every figure can be followed back to the sentence of the study that supports it.
- 33,254publications found by the search
- 16,954checked one by one against the criteria
- 429studies with data measured in plants
- 7,991data points extracted, each with its verbatim quote
Status of each stage
- SearchComplete33,254 records identified, 16,957 unique
- ScreeningComplete16,954 of 16,957 screened; 1,469 passed
- Full-text retrievalComplete429 studies included
- Double extractionComplete7,991 data points from 1,325 units
- Bayesian synthesisComplete460 causes estimated
- Human validationPending
- Release
- 2026.11.18.1
- Published
- November 18, 2026
- Protocol version
- 1.0.0
- Search strategy version
- 1.0.1
For specialists
Search, unit operations with their physical models, method, downloadable data and version history.
How efficient can it physically be?
Three ceilings per dimension (theoretical limit, best available technique, best documented operation), the loss cascade between them and a physics-bounded OEE, each with its full derivation and 95% uncertainty interval.
21 physical models published · 329 verified source citations
Why does it stop?
The probability of each cause of stoppage or inefficiency, classified into the six big losses, with 95% credible intervals, number of studies, evidence grade and the verbatim source of every datum.
460 causes estimated from 429 included studies
Unit operations
The physics-bounded OEE is shown where a physical model has been published and validated; the remaining models are in preparation.

