Hidden Factory: Microstops and Lost Capacity (2026)

By the editors at Reliable

How We Evaluated

Independent editorial analysis based on a published interview with the term’s originator, a government reference page, a nonprofit lexicon, a standards body’s public specification, one vendor’s formula documentation, and one open-access, peer-reviewed conference case study.

Reliable Magazine does not sell production monitoring software or consulting and has no commercial interest in routing readers toward any particular vendor. Reliable does not accept payment for rankings. Vendors may sponsor enhanced listings with additional detail, but editorial rankings are independent. Read our editorial policy.

Two rules applied throughout. Sourced figures retain the scope of their references; worked examples and calculations are labeled. The one vendor source is identified as a vendor where it is used.

Where the Term Comes From

Armand Feigenbaum, the quality engineer behind Total Quality Control, used the phrase in a two-part IndustryWeek interview published July 4, 1994.

He said the cost of quality in major American companies could be 25% or more of sales, with most of that being failure cost. He attributed it to what he called the “hidden organization or hidden factory,” the part of an organization that exists to do bad work because the process drives people into it.

Asked about the effect on American industry, he said that from an operating point of view about 20% to 40% of the total capacity of many American companies was tied up in that hidden organization.

Three cautions apply to anyone quoting the range.

It is about quality-related waste. Feigenbaum was discussing quality-related waste and failure-related work across organizations. The interview does not measure microstops or isolate equipment losses.

It is a stated estimate. The interview gives no supporting dataset for the capacity figure.

It is dated and scoped. The statement is from 1994 and refers to many American companies.

From Quality Rework to Equipment Losses

The equipment side of the idea comes from Total Productive Maintenance (TPM).

The U.S. Environmental Protection Agency’s reference page on TPM says the method aims at the total elimination of losses, and it lists breakdowns, setup and adjustment losses, idling and minor stoppages, reduced speed, defects and rework, and startup and yield losses. The same page gives the TPM index for those losses: Overall Equipment Effectiveness (OEE), calculated by multiplying availability, performance, and quality rate.

The three OEE factors describe the share of planned time available for production, the speed achieved during run time, and the share of output that meets requirements:

• Availability = run time / planned production time.
• Performance = (ideal cycle time x total count) / run time.
• Quality = first-pass good count / total count. Parts requiring rework are excluded from first-pass good count.

These formulas follow the calculation documentation published by Vorne, a vendor of production monitoring systems. The OPC Foundation’s machine tool specification gives the same three-factor structure and says its KPI definitions are taken from ISO 22400-2.

The Lean Enterprise Institute’s lexicon adds that the performance rate covers running below design speed and stoppages lasting a few seconds.

This guide uses “hidden factory” in a broader sense that includes these equipment losses. Feigenbaum’s 20% to 40% estimate was not based on that definition.

What a Microstop Is

The sources reviewed here give no single definition. They describe the same thing at different scales.

• The Lean Enterprise Institute describes stoppages lasting a few seconds.
• The case study below covers a broader category of operator-cleared minor stoppages, from about ten seconds to 10 to 15 minutes.
• Vorne, a production monitoring vendor, assigns stops long enough to justify tracking a reason, typically several minutes, to availability. It assigns small stops to performance.

A plant sets its own cutoff in its OEE system, and the cutoff decides where a stop lands. A stop long enough to be logged as downtime reduces availability. A shorter stop shows up as fewer parts than the ideal cycle time allows, which reduces performance.

An unlogged stop is different from an unmeasured loss. With valid counts, a sound cycle-time reference, and correct planned time, the OEE score still reflects the output shortfall. Missing event detail can hide the causes.

Why Microstops Stay Hidden

The case study’s literature review sorts OEE losses into two types. Sporadic losses, such as breakdowns, happen suddenly and get noticed. Chronic losses happen often, are usually considered small, and tend to persist.

The authors summarize Seiichi Nakajima’s description of chronic problems. They are usually latent. Each incident causes negligible loss. They occur frequently. Operators restore them easily. They rarely reach supervisors. They are difficult to quantify.

Three mechanisms can make these losses harder to identify, according to the same review:

• Habit. Chronic losses become part of the daily routine and stop being experienced as losses.
• Manual logging. Research cited in the paper finds that manual data collection has low accuracy because minor stoppages are often forgotten.
• Cycle-time references. OEE studies have reported performance rates of 100% and above, which the authors read as a sign that minor stoppages and reduced speed were not fully logged. A performance rate above 100% calls for checking the cycle-time reference and input data. A cycle-time allowance that is too generous can conceal losses; missing event codes alone do not establish why the score exceeds 100%. Vorne’s documentation says a performance figure above 100% usually indicates an ideal cycle time that is set incorrectly.

The review also cites earlier OEE studies in which performance losses were larger than availability losses.

A Documented Example: 319 Hours Estimated, 106 Logged

A documented example is a 2022 open-access, peer-reviewed conference case study by Marcus Bengtsson, Peter Alm, and Bo Tjulin, published by IOS Press in the proceedings of the Swedish Production Symposium.

The setting. The plant is a discrete manufacturer supplying components for the automotive industry, with about 400 manufacturing machines. It had measured OEE for more than ten years, using a manual system in which operators log their own losses. The machine studied was a turning machine installed in 2007.

The method. Two team leaders and two operators listed the minor stoppages they had seen in the past year, with a frequency and duration for each. Where they gave a range, the researchers used the low end. The estimates were then compared with a year of data from the OEE system.

What the operators described. They named eight chronic losses. One was a door that fails to close on the first attempt four to five times per shift, at about ten seconds each. Others involved chips and dirt on sensors and conveyors losing their signal.

The result. The authors reported an annualized estimate of 319.4 hours, compared with 106.2 hours recorded in the relevant OEE loss category. The estimate included 10 to 15 minute interruptions and used a three-shift model for a cell operating a four-shift strategy. The logs also excluded some machine losses when another machine was the bottleneck, while including downstream losses. This comparison differs in schedule and reporting scope.

The authors’ tool expressed the estimate as almost 40 lost shifts. They note a difference of more than 200 hours between the estimate and the OEE system, and that the operators and team leaders were themselves surprised by the total.

For scale, the same OEE system logged 355.2 hours of breakdowns and long stops on that machine.

The authors give two reasons for the difference. Logging minor stoppages by hand is difficult, and some stops are no longer seen as losses. The plant also logged OEE on one fixed machine even when a different machine was the bottleneck.

The limits. This is one machine in one plant, based on estimates from four people. The authors describe it as a problem-filled machine compared with others in the same cell.

How to Estimate Microstop Losses on One Machine

The case study’s tool uses basic arithmetic, and the same steps work on any machine.

1. Sit down with the operators and team leaders who run the machine. List every recurring short stop that would not happen if the machine ran without problems.
2. For each stop type, estimate how many times it occurs per shift and how many minutes it takes to clear. Where the answer is a range, use the low end.
3. Multiply occurrences per shift by minutes per stop by the number of shifts worked in a year. Convert the total to hours and to lost shifts.
4. Pull a year of data from the OEE system for the matching loss category.
5. Compare the estimate with recorded losses for the same machine, period, operating schedule and loss definitions. Investigate the difference as a possible reporting gap. Validate the estimates through observation or event data before treating the difference as unreported loss.

A worked example with hypothetical inputs. Assume a jam that occurs four times per shift and takes 30 seconds to clear, on a machine running three eight-hour shifts, five days a week, 46 weeks a year. Those calendar assumptions match the case study’s tool.

• Loss per shift: 4 stops x 0.5 minutes = 2 minutes
• Shifts per year: 3 x 5 x 46 = 690
• Annual loss: 2 x 690 = 1,380 minutes, or 23 hours
• Lost shifts: 23 / 8 = about 2.9

One 30-second stop type adds up to nearly three shifts of machine time a year in this example. The inputs are illustrative and are not drawn from any plant.

This calculation estimates machine time lost. Converting that time into additional line output requires checking the production constraint and the effect of buffers and other losses. Use the actual operating calendar and avoid counting overlapping events twice. Do not add an estimated event total to an existing OEE loss total without reconciling what each already includes.

The 85% Benchmark

An OEE of 85% is often used as a reference point when lost capacity is discussed, so its basis matters.

The Lean Enterprise Institute’s lexicon illustrates the OEE calculation with 90% availability, 95% performance, and 99% quality, which multiply to 84.6%.

That multiplication does not establish an industry average or prove a universal target. This guide did not trace the origin of the 85% benchmark to an original source. The sources reviewed here give no measured industry average for OEE, and this guide does not publish one.

At a Glance

Honest Limitations

• No industry-wide microstop figure exists in these sources. Nothing here supports a percentage of capacity lost to microstops across manufacturing, and this guide does not construct one.
• The Feigenbaum range is a stated estimate. It comes from a 1994 discussion of quality-related waste, and the interview gives no supporting dataset.
• The case study is one machine. Its 319.4 hours rest on estimates from four people at one plant, on a machine the authors call problem-filled. The estimate included 10 to 15 minute interruptions and a three-shift model for a four-shift cell, and the OEE log had scope limits of its own.
• Operator estimates can miss losses too. The authors note that losses perceived as normal may not surface in interviews, so the estimate is not a ceiling.
• Microstop definitions vary. The sources describe stops from a few seconds to several minutes, and no universal cutoff appears in them.
• Nakajima is cited secondhand. His descriptions reach this guide through the case study. This guide did not verify these passages against Nakajima’s original books.
• The worked example is hypothetical. Its inputs were chosen to show the arithmetic, and it estimates machine time, which is different from recoverable output.
• The ISO standard was not reviewed. The OPC Foundation specification says its KPI definitions are taken from ISO 22400-2. This guide relies on that public specification and did not review the ISO standard itself.
• One vendor source is used. The formula wording follows Vorne’s public documentation. Vorne sells production monitoring systems.

Frequently Asked Questions

What is the hidden factory in manufacturing?

It is production capacity consumed by recurring losses that may go unrecognized or underexplained in plant reporting. Armand Feigenbaum used the phrase in a 1994 IndustryWeek interview for the part of an organization that exists to do failure-related work. This guide also applies the term to equipment losses such as minor stoppages and reduced speed.

How much capacity does the hidden factory consume?

Feigenbaum said in 1994 that about 20% to 40% of the total capacity of many American companies was tied up in it. That is his stated estimate from a discussion of quality-related waste, and the interview gives no supporting dataset. No source reviewed for this guide gives a measured percentage for equipment losses or microstops across manufacturing.

What is a microstop?

A microstop, also called a minor stoppage or small stop, is a brief interruption that an operator clears without calling maintenance. The Lean Enterprise Institute describes stoppages lasting a few seconds. The case study in this guide covers a broader category of operator-cleared minor stoppages, some lasting 10 to 15 minutes. A plant sets its own cutoff in its OEE system.

Do microstops count against availability or performance in OEE?

Usually performance. Availability is run time divided by planned production time, so a stop long enough to be logged as downtime reduces it. Shorter stops reduce performance, which compares the ideal cycle time multiplied by total count against run time. The Lean Enterprise Institute’s lexicon places stoppages lasting a few seconds in the performance rate.

Why are microstops underreported?

They are frequent, short, and fixed by operators, so they rarely reach supervisors. A 2022 peer-reviewed conference case study adds that chronic losses become routine and that manual logging can miss them. On the machine it studied, the authors reported an annualized estimate of 319.4 hours against 106.2 hours recorded in the relevant OEE loss category, a comparison that differs in schedule and reporting scope.

How do you calculate lost capacity from microstops?

For each recurring stop type, multiply occurrences per shift by minutes per stop by shifts worked per year, then convert to hours. In a hypothetical example, a 30-second stop that occurs four times per shift on a three-shift, five-day, 46-week schedule adds up to 23 hours a year, or about 2.9 eight-hour shifts. Compare the estimate with recorded losses for the same machine, period, operating schedule and loss definitions. Investigate the difference as a possible reporting gap. Validate the estimates through observation or event data before treating the difference as unreported loss.

Related Guides

• How to Calculate OEE
• Best MES Platforms for Manufacturing
• Best CMMS Software
• Best APM Software
• Maintenance and Reliability Glossary

Sources

• IndustryWeek: Dr. Armand Feigenbaum on the Cost of Quality and the Hidden Factory (interview published July 4, 1994)
• U.S. Environmental Protection Agency: Lean Thinking and Methods – TPM (TPM loss categories and the OEE calculation)
• Lean Enterprise Institute: Overall Equipment Effectiveness (lexicon definition of the three OEE rates)
• IOS Press: Visualizing the Effects of Chronic Versus Sporadic Losses in Manufacturing Industries: A Case Study (open-access, peer-reviewed 2022 conference paper by Bengtsson, Alm, and Tjulin)
• OPC Foundation: OPC UA for Machine Tools, Annex C: KPI Calculation (informative annex with the OEE calculation; names ISO 22400-2 as its KPI source)
• Vorne, OEE.com: Calculate OEE (vendor documentation of the OEE formulas)