This table describes what got published, not what plants achieve: it is not a global benchmark, and it does not tell you what a good OEE is. Half of the rows come from Indonesia or Bangladesh, and most are single machines or lines, many chosen because they had a problem. Its 47 comparable starting values, from 38 studies, have a median of 63% and run from 14% to 87%.
Most "OEE by industry" numbers online come with no method. You cannot see who was measured, how many machines, or how the data was caught.
This page does the opposite. Each row below links to a study we read, and shows the value in the authors' own words.
Before you quote a number
- Not a global sample. 31 of the 62 rows come from Indonesia (16) or Bangladesh (15, 11 of them by the authors' address only). No row names the US, China or Japan.
- The mix moves the answer. Leave out those two countries and the median of the other 23 counted rows is 55%, not 63%.
- Big publishers are missing. MDPI, ScienceDirect, Taylor and Francis, SAGE, Wiley, Emerald and most of IOP refused our automated reader, so few of their papers are here.
- Published does not mean typical. Authors tend to publish projects that worked, on machines they chose because they had a problem.
- Not a target. This table cannot tell you what a good OEE is. For a goal, set OEE targets from your own trend.
The table
62 rows from 47 studies, each read on 3 November 2026. Click a heading to sort, or filter below.
Download the CSV for the title, source URL and full notes of each row.
62 of 62 rows shown
| Hedman et al. 2016 | average OEE is 65%; overall median OEE of all 23 companies is 70% | single reading | Mixed discrete | Sweden | Multiple companies: 23 companies, 884 machines, 6 months (Oct 2013 to Mar 2014) | automatic | conference | full text | Data came from the software firm Good Solutions AB: the paper thanks its CEO, Mikael Persson, for the dataset. 702 of 884 machines recorded 100% performance and 796 recorded 100% quality, so the figure mostly counts stops. |
| Hedman et al. 2016 | median and average OEE of 74% | single reading | Food and beverage | Sweden | Multiple companies: Food and beverage group of the same data set; group size not read | automatic | conference | full text | Subgroup of the row above. Not in the median. |
| Hedman et al. 2016 | lowest median and average OEE of 59% | single reading | Other automated discrete | Sweden | Multiple companies: Other automated discrete group of the same data set; group size not read | automatic | conference | full text | Subgroup of the first Hedman row. Not in the median. |
| Cheh 2014 | Average OEE 43, 5 % without considering Total schedule production time | single reading | Mixed discrete | Sweden | Multiple companies: Swedish companies in four groups, 867 data records, Oct 2013 to Mar 2014 by its method chapters (two results sections say Oct 2012 to Mar 2013); company total not read | automatic and manual | thesis | full text | Commissioned by Good Solutions AB, on data from its RSproduction system. Same firm, company contact and group names as Hedman 2016, so the data may overlap, though neither text says so. Average OEE here is far lower. Not in the median. |
| Cheh 2014 | Food and beverage 49,9%; Mechanical workshop 49,8%; Polymer 38,7; Other automated discrete 43,3% | several values | Mixed discrete | Sweden | Multiple companies: Same data split into four industry groups | automatic and manual | thesis | full text | Compare with Hedman 2016: food and beverage 74%, other automated discrete 59%. Not in the median. |
| Ljungberg 1998 | only around 55 per cent | single reading | Not stated | Not stated | Multiple cases: "about 20 cases" | not stated | journal (peer review claimed) | abstract | The abstract gives no sectors or country. Emerald refuses scripts, so we read the abstract in the Crossref record for this DOI, not on the publisher page. |
| Ondra 2022 | OEE <40%; approx. 13.5% of companies; OEE between 40% and 60%; approx. 9.5%; OEE 60-80%; approx. 18.0%; OEE 80-100%; approx. 5.0% | several values | Mixed industrial | Czech Republic | Multiple companies: 92 of 200 surveyed companies, 2019 to 2020 | not stated | other | full text | Bands only, no single OEE. Band wording normalised. 54.0% of the 200 companies did not monitor OEE at all, a warning for any survey average. Not in the median. |
| Fuad et al. 2025 | average OEE 62.5% | single reading | Carton packaging | Not stated (authors in Bangladesh) | One machine: Auto corrugation machine, 12 shifts | not stated | journal (peer review claimed) | full text | The 29% OEE gain in the abstract matches efficiency figures of 60% to 89% for the line. These compare output with expected output and are not OEE. |
| Fuad et al. 2025 | average OEE 59% | single reading | Carton packaging | Not stated (authors in Bangladesh) | One machine: Flexo printer, 12 shifts | not stated | journal (peer review claimed) | full text | |
| Fuad et al. 2025 | average OEE 62.5 | single reading | Carton packaging | Not stated (authors in Bangladesh) | One machine: Auto gluer (printed as auto glutter), 12 shifts | not stated | journal (peer review claimed) | full text | The same paper also says this machine ranged around 68%. |
| Fuad et al. 2025 | average OEE 58.5% | single reading | Carton packaging | Not stated (authors in Bangladesh) | One machine: Binding machine, 12 shifts | not stated | journal (peer review claimed) | full text | |
| Irfan et al. 2025 | average OEE increased from 65.59 % to 68.12 % | before and after | Cement | Bangladesh | Several machines: Six machines at one plant; Sep and Oct 2022, then Nov and Dec 2023 | not stated | journal (peer review claimed) | full text | The abstract says almost 65.60%. The text says five machines, the tables show six. |
| Haddad et al. 2021 | OEE = 90.50% x 95.77% x 75.54% = 65.47%; then 95.36% x 96.09% x 75% = 68.73% | before and after | Aluminium extrusion | Palestine | One line: One extrusion line, single company | manual | journal (peer review claimed) | full text | |
| Rozak et al. 2020 | increasing OEE value from 87% became 92% | before and after | Automotive | Indonesia | One line: Cylinder block machining line, one company | not stated | journal (peer review claimed) | full text | The authors say the line's 87% was already above the world class bar. The body prints 92.3% for the after value. |
| Sunadi et al. 2021 | from an average of 26,43% to 78,87% | before and after | Plastics | Indonesia | One machine: One injection molding machine; baseline covers three months | not stated | other | full text | The largest jump in the table: availability 54.97% and performance 48.54% before. |
| Darsin 2020 | Rata-rata OEE selama enam bulan adalah 85.57% (monthly low 68.27%, high 98.46%) | single reading | Sugar | Indonesia | One line: Bagasse conveyor at one sugar factory, six months of 2018 | not stated | journal (peer review claimed) | full text | A swing of 30 points inside one season. Text in Indonesian; in English, the average OEE over six months is 85.57%. |
| Ummah and Dahda 2022 | the Cutting Automatice (CNC) machine is 73.24% | single reading | Not stated (cutting) | Indonesia (from authors) | One machine: CNC cutting machine, one company | not stated | journal (peer review claimed) | full text | |
| Ummah and Dahda 2022 | the manual cutting machine is 55.32% | single reading | Not stated (cutting) | Indonesia (from authors) | One machine: Manual cutting machine, same company | not stated | journal (peer review claimed) | full text | |
| Daman and Nusraningrum 2020 | 84% for EX157 units | single reading | Coal mining | Indonesia | One machine: Excavator EX157, 2018 | not stated | journal (peer review claimed) | full text | Definition differs: quality is the bucket fill factor, and the paper compares with a mining bar of 77% (Elevli 2010). Not in the median. |
| Daman and Nusraningrum 2020 | 68% for EX158 units | single reading | Coal mining | Indonesia | One machine: Excavator EX158, 2018 | not stated | journal (peer review claimed) | full text | Definition differs: quality is the bucket fill factor. Not in the median. |
| Sumasto et al. 2024 | The calculated OEE rose from 67.42% to an impressive 77.80% | before and after | Automotive components | Indonesia | Several machines: Blowing machines at one plant; baseline Sep to Nov | not stated | journal (peer review claimed) | full text | The text also prints the baseline as 66.42%. |
| Shakil and Parvez 2018 | OEE = 45%; Proposed OEE = 48.75% | before and forecast | Garment | Bangladesh | One line: One sewing line | manual | other | full text | The second figure is calculated for a proposed layout. The abstract words it as achieved. |
| Hermanto 2016 | resulted OEE values painting division 70.8% | single reading | Not stated (painting) | Indonesia (from authors) | Several machines: Dryer, painting and oven machines, 29 Sep to 30 Oct 2015 | not stated | journal (peer review claimed) | full text | |
| Santos 2020 | press (OEE = 83.47%) | single reading | Wine | Brazil | One machine: Press at one winery, 28 days in 2018 | manual | other | full text | |
| Santos 2020 | two filling machines (69.19% and 69.38%, respectively) | single reading | Wine | Brazil | One machine: Filling machine 1 at the same winery, 28 days in 2018 | manual | other | full text | |
| Santos 2020 | two filling machines (69.19% and 69.38%, respectively) | single reading | Wine | Brazil | One machine: Filling machine 2 at the same winery, 28 days in 2018 | manual | other | full text | |
| Djunaidi et al. 2022 | OEE value on the Flask Less Horizontal Moulding Machine was 58,53%; average overall equipment effectiveness value is 46.65% | single reading | Metal foundry | Indonesia | One machine: One moulding machine; June to August | not stated | journal (peer review claimed) | full text | Two different values for the same machine: 58,53% in the abstract and 46.65% in the text. Not in the median. |
| Hermanto et al. 2018 | rata-rata nilai OEE adalah 75,69% | single reading | Plastics (injection moulding) | Indonesia (from authors) | One machine: One Krauss Maffei KMF 250 A injection machine making lids, Jan to May 2017 | not stated | journal (peer review claimed) | full text | In English, the average OEE value is 75.69%. The text also gives 76,13% for January to May 2017, up from 52.70% for July to December 2016. |
| Junaedi et al. 2023 | December 2021 (53.4%), January 2022 (60.6%), February 2022 (61%), March 2022 (63.3%) | several values | Food and beverage | Indonesia (from authors) | One machine: Filling line 3, Dec 2021 to Mar 2022 | not stated | journal (peer review claimed) | full text | Monthly series, no overall figure. We did not average it. Not in the median. |
| Nurprihatin et al. 2023 | the OEE value increased from 70% to 74% | before and after | Porcelain | Indonesia (from authors) | One machine: Jigger 01 machine, 2019 | manual | journal (peer review claimed) | full text | |
| Ardy and Widiasih 2024 | before scheduling 69.34%; after, increasing to 77.78% | before and after | Fasteners | Indonesia | Several machines: 10 bolt former machines at one company, Mar 2023 to Feb 2024 | not stated | journal (peer review claimed) | full text | The tables give 67,50% across machines, and 66,62% then 77,53% for machine B22. |
| Al-Khazraji et al. 2020 | OEE value is increased in 2017 to 75% in comparison with its value in 2016 which was 67% | before and after | Concrete products | Iraq | One line: One production line, 2016 and 2017 | not stated | journal (peer review claimed) | full text | The conclusion prints the 2016 figure as 66%. |
| Tau et al. 2023 | from 51.17% to 72.93% | before and after | Coal fly ash processing | South Africa (from authors) | One machine: One classifier, 31 days before and 31 days after | manual | other | full text | Daily values run from -6.00% to 93%. The text also gives 73.02% for the after average. |
| Smirnov and Abdilova 2021 | the overall equipment effectiveness (OEE) accounts for 57% | single reading | Meat processing | Kazakhstan (from authors) | One machine: One calculation with planned output of 800 kg | not stated | journal (peer review claimed) | full text | Reads as one illustrative calculation, not a measured period. Not in the median. |
| Veroya et al. 2021 | from 66.59% to 85.09% | before and after | Semiconductor | Philippines | One line: Lapping process, one company | not stated | conference | abstract | |
| Nurprihatin et al. 2023 | the OEE is expected to increase from 82% to 85% | before and forecast | Palm oil packaging | Indonesia (from authors) | One machine: Pouch filling machine, Apr to Jul 2021 | not stated | journal (peer review claimed) | full text | The second figure is a forecast: nothing had been implemented. |
| Martínez-Mercado et al. 2023 | the line presents a low OEE level of 65% | single reading | Not stated (cutting) | Mexico (from authors) | One line: One cutting line | not stated | conference | abstract | Two-page extended abstract. |
| Ratanawilai and Srisuk 2015 | from 55.8% to 68.3% | before and after | Wood products | Thailand (from authors) | One line: One wood plank packing process | not stated | journal (peer review claimed) | full text | The abstract gives TEEP from 35.7% to 68.3%, the same after figure as OEE, which is unusual. |
| Vannan et al. 2025 | improvement in OEE from 30% to 38.18% | before and after | Metal stamping | India (from authors) | Several machines: One press shop, Mar and Apr 2022 | manual | journal (peer review claimed) | full text | The conclusion prints 38.14%. |
| Pius et al. 2022 | 61.27% OEE | single reading | Water utility | Nigeria | One plant: All pumping stations of one water system, six months | not stated | other | abstract | Not manufacturing. The abstract swaps the usual 95% and 99% factors it attributes to Nakajima, and its printed factors multiply to about 63.4%, not 61.27%. Not in the median. |
| Njeshu et al. 2017 | from 43.7% to 49.9% | before and after | Rubber | Cameroon | One line: One crumb rubber line, one month | manual | journal (peer review claimed) | full text | Machines on the same line ranged from 24.1% to just over 60%. |
| Mojapelo et al. 2024 | OEE of Machines = 0.74 x 0.75 x 0.95 = 52.7~ 53% | single reading | Refractories | South Africa (from authors) | Several machines: Machines at one refractory plant; count not read | not stated | conference | full text | |
| Jamal and Raia 2024 | rising from 27% to 62% | before and after | Food and beverage | Brazil | One line: Cake mix filling line, one plant | not stated | other | full text | The authors left out the underlying data for confidentiality, so the two values cannot be recomputed. |
| Khan et al. 2026 | average OEE increased from 45.15% to 54.12% | before and after | Food and beverage | Bangladesh | One plant: One beverage facility, daily and shift records | not stated | journal (peer review claimed) | full text | |
| Ahamed et al. 2025 | Blower & Fan Blade 84.90% 86.50% | before and after | Air conditioner manufacturing | Bangladesh (from authors) | One line: Blower and fan blade process | automatic and manual | conference | full text | The abstract gives a range of 63% to 75%, which this process falls outside. |
| Ahamed et al. 2025 | Compressor Jacket 65.06% 71.20% | before and after | Air conditioner manufacturing | Bangladesh (from authors) | One line: Compressor jacket process | automatic and manual | conference | full text | The printed factors multiply to about 63%, not 65.06%. |
| Ahamed et al. 2025 | Evaporator & Condenser 75.32% 77.50% | before and after | Air conditioner manufacturing | Bangladesh (from authors) | One line: Evaporator and condenser process | automatic and manual | conference | full text | |
| Ahamed et al. 2025 | Microchannel Condenser 67.30% 67.50% | before and after | Air conditioner manufacturing | Bangladesh (from authors) | One line: Microchannel condenser process | automatic and manual | conference | full text | |
| Ahamed et al. 2025 | Powder Coating 67.30% 63.76% No Change | before and after | Air conditioner manufacturing | Bangladesh (from authors) | One line: Powder coating process | automatic and manual | conference | full text | The table labels a fall of 3.54 points as no change, and an earlier table gives 63.76% as the starting value. Not in the median. |
| Mohamad et al. 2020 | it improved from 48.8% to 66.5% | before and after | Lubricants | Malaysia (from authors) | One line: Bottle line 3, one plant | not stated | conference | full text | The project goal was 65%. |
| Ramadhan et al. 2024 | OEE value from the period of June 2024 to Ugust 2024, which was 43,4 % | single reading | Tobacco | Indonesia (from authors) | One machine: One packer, June to August 2024 | not stated | journal (peer review claimed) | full text | Monthly values were 48.7%, 43.3% and 38.3%. |
| Hassan et al. 2025 | 47.16; 48; 48.93 by shift | several values | Pharmaceuticals | Bangladesh (from authors) | One machine: Hoonga HM 200R blister machine, three shifts | not stated | conference | full text | Shift values, no overall figure. We did not average them. Not in the median. |
| Hassan et al. 2025 | 28.75; 28.02; 29.28 by shift | several values | Pharmaceuticals | Bangladesh (from authors) | One machine: Horn Noack N-740 blister machine, three shifts | not stated | conference | full text | Shift values, no overall figure. We did not average them. Not in the median. |
| Laghari et al. 2025 | resulting in an overall OEE of 27.79% | single reading | Wood panels (MDF) | Pakistan | One machine: One hot press, 60 days | manual | journal (peer review claimed) | abstract | |
| Sriputtha 2023 | from 38.18% to 81.56% | before and after | Ceramics (plaster molds) | Thailand (from authors) | One line: Plaster mold process, 30 days | not stated | journal (peer review claimed) | full text | The text also prints a baseline of 37.90%. |
| Raval et al. 2026 | OEE improved from 86% to 94.70% | before and after | Bearings | India (from authors) | One machine: One external grinding machine | automatic | journal (peer review claimed) | full text | The authors built the monitoring system they evaluate, and the before and after data differ in method. |
| Tayal and Kalsi 2019 | The OEE has been improved from 62.70 % to 74.30% | before and after | Automotive | India (from authors) | Not stated: Machining operation; size not given in the abstract | not stated | journal (peer review claimed) | abstract | |
| Abdullah et al. 2025 | The OEE of the machine improved from 63% to 73% | before and after | Tyres | Bangladesh | One machine: One tire curing machine, one month | not stated | journal (peer review claimed) | abstract | |
| Ottosson 2009 | an increase from 49.2% to 57.2% | before and after | Automotive | Germany | One line: One pilot line at one plant; after figure is the last four weeks | manual | thesis | full text | The author warns of substantial deviations, and describes an early OEE error that gave values above 100%. |
| Kumar and Sahu 2023 | The average of OEE from all the data that has been collected is 14% | single reading | Batteries | Not stated (thesis in Sweden) | One machine: Stacking equipment at a newly built plant, selected days | manual | thesis | full text | New line in ramp up. The company target was 65%, and it put the highest achievable OEE at 29%. |
| Hartman and Moussa 2023 | the average OEE score of stamping machines is around 44.20% | single reading | Outdoor products | Sweden (one machine in the Netherlands) | Several machines: Seven stamping machines, 33 weeks | not stated | thesis | full text | Calculated from quality and a combined performance and availability score. |
| Gurruchaga-Rodríguez et al. 2025 | overall operational efficiency (OEEaq) value of 76.37% | single reading | Aquaculture | Mexico | One plant: One tilapia farm, 8 batches, 2016 to 2019 | not stated | other | full text | Definition differs: an adapted OEEaq, with batch values above 100%. Not in the median. |
"From authors" means the study does not name the country, so we took it from the authors' location. Values keep the authors' own commas and wording, except a few that are shortened or translated into English.
Median and range, with limits
We computed these from the CSV by script. They describe these studies only, and half of the rows come from two countries.
The median uses one starting or single value per row, only where a study reports one comparable figure. That leaves 47 of 62 rows: median 63%, middle half 49% to 69.36%, lowest 14%, highest 87%.
Some studies give several rows. Counting each of the 38 studies once, using its own median, gives 62.85%.
3 of those 47 rows start at 85% or higher, the world-class figure usually traced to Nakajima's TPM work. For what that bar means, see is 85% OEE realistic.
What the table can tell you
- The spread is wide. Reported values run across most of the scale, so one number for "typical OEE" hides a lot.
- Related sources can disagree. Two Swedish sources below use the same four industry groups and data from the same software firm, and land far apart.
- Most rows do not say how data was caught. Many cite company records and stop there, so we mark those as not stated.
Hedman's paper thanks Mikael Persson, CEO of Good Solutions AB, for the dataset, and Persson was Cheh's company tutor.
Cheh's method chapters give October 2013 to March 2014, but two results sections say October 2012 to March 2013. So we cannot tell how far the data overlap, and we count Hedman because it gives company and machine totals.
What it cannot tell you
- Publication bias. Failed projects seldom reach print. Of 26 before and after pairs here, 25 rise.
- Typical plants. Some authors pick the worst machine on purpose, as Mohamad 2020 did with its lowest line, so starting values may sit low.
- Sector averages. Most sectors have one or two rows. That is a list of cases, not a sector figure.
- Fair comparisons. Rows differ in period, from selected days to two years, and in how speed and quality were set.
- Effect of a method. The median rise of 8.44 points across 25 pairs has no control group, and later periods may simply differ.
Small samples matter here. A machine studied for a month, like OEE for one machine describes, says little about a plant, and averaging OEE across machines needs care of its own.
Before and after pairs
25 rows in the median also give a measured value after a project. Their median moves from 65.06% to 73%, and the median of each row's own rise is 8.44 points.
Forecasts are kept apart. Two rows give a future figure, and we label it as a forecast because no such result was measured.
How we built it
OpenAlex and Europe PMC for open access records on overall equipment effectiveness, then thesis repositories.
We read the full text, or the abstract where only that was open, and never a search snippet or a summary.
The value as the authors reported it, its scope, its timing and the data method. Empty stays empty: we estimated nothing.
A second reader reopened 25 random rows and every row at 85% or more against the source, and corrected what was wrong.
At the time of our search, OpenAlex returned 1,200 open access records before its rate limit stopped us, and Europe PMC added 136. We did not save the query log, so read these counts as approximate.
This is a selection, not a census. We read the hits from outside Indonesia first, then a sample of the large Indonesian group, and 16 rows still come from there.
The publishers that refused our automated reader are named at the top of this page. Their papers are missing because we could not open them, not because of their quality.
What we left out
- Reviews and papers that repeat another study's value, like those that quote Hedman or Cheh.
- Papers that give only a gain, such as "13% higher", with no OEE level.
- Simulations and model outputs, and values shown only in charts.
- Hospital and laboratory equipment, where one lab study also had a vendor co-author.
- Duplicates, including one paper indexed twice.
- Papers we could not read, like Raju et al. 2022 on Emerald, whose abstract gives no OEE value.
The same checks explain the Not in the median notes: subgroups, possible overlap, a different OEE definition, conflicting values, or a series with no overall figure. Standard OEE here means availability x performance x quality, as in Overall Equipment Effectiveness (R. C. Hansen, 2001) and our guide to what OEE is.
"Journal (peer review claimed)" means the journal or the paper says it is peer reviewed. That is the journal's own wording, so it is weak evidence, and we did not audit it.
Where we could not find that claim, the row says "other". Some of these journals are small.
How to cite this table
License: our compilation (table, CSV, charts and text) is CC BY 4.0: reuse it, also commercially, with a credit and a link to this page. The studies remain their authors' work, so cite them directly when you use one value.
The CSV is UTF-8 with one header row, one row per value above, and gives authors, title, venue, links, access date and the reason a row is out of the median.
Use it on your own numbers
Comparing your line with a row here says little until your method matches. Check your OEE data collection method, look for OEE calculation mistakes, then measure with the OEE calculator.
When you need a goal, set OEE targets from your own trend. The OEE benchmark report covers the wider picture.
Most of these studies do not say how the data was caught. 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, so you can see how your OEE per line and shift was measured. See OEE per line and shift in Fabrico.
What people ask next
What OEE do published studies report?
They show what got published, not what plants achieve. Half of all rows come from Indonesia or Bangladesh. In this table the median starting value is 63%, across 47 rows from 38 studies, with a range of 14% to 87%. The OEE benchmark report covers other sourced figures.
Is this a benchmark for my industry?
No. Most rows are single machines or lines, and many sectors have one or two rows. Compare your line with its own trend first, as in how to set OEE targets.
What is a good OEE according to these studies?
These studies cannot answer that, because they show what got published and not what plants achieve. The 85% figure comes from Nakajima's TPM work, and is 85% OEE realistic covers it. For your own goal, see how to set OEE targets.
Why do the Swedish studies disagree?
Hedman 2016 reports a 65% average and Cheh 2014 reports 43.5%, for the same groups, with data from the same software firm. Cheh's dates differ within the thesis and neither explains the gap, so we cannot say how far the data overlap or which is closer. How the time base and defaults were handled can move the result, and downtime data accuracy covers why capture matters.
Can I reuse the table and the CSV?
Yes. Our compilation is CC BY 4.0, so reuse it with a credit and a link to this page. Cite the original study when you use a single value.
Why are the after values nearly all higher?
Authors mostly publish projects that worked, and the after period often has no control group. Read those gains as reported cases. To plan your own, start from how to improve OEE.
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.