Micro-stops and Their Impact on Performance

What Is a Micro-stop? Operational Definition and MECE Taxonomy

Micro-stops are the silent cancer of productivity. They are brief interruptions, lasting between 1 second and 5 minutes, that do not get registered as formal stops in CMMS systems. However, their cumulative effect is devastating. To diagnose them, we must first define them precisely.

A micro-stop is any unplanned interruption of the production flow that meets three conditions:

• Duration: Between 1 second and 5 minutes.
• Resolution: Can be solved by the operator themselves without needing specialized maintenance.
• Nature: It is transitory and recurring.

Its main characteristic is that it escapes conventional recording systems. An operator is not going to open a work order for a 45-second jam. They simply resolve it and move on. But if that jam occurs 20 times per shift, we are talking about 15 minutes of pure loss.

The Importance of a MECE Classification

For effective diagnosis, we need a taxonomy that is Mutually Exclusive, Collectively Exhaustive (MECE). This means that each micro-stop must be classifiable in one and only one category, and that together all categories cover 100% of cases.

This is our field taxonomy, empirically validated:

Category	Technical Description	Concrete Examples	Relative Frequency (Empirical Average)
Machine adjustment	Minor readjustment of process parameters	Pressure, temperature, speed, tool position adjustment	22%
Material jam	Blockage in feeding, accumulation, or transfer of material	Piece jammed in feeder, poorly wound coil, hopper obstruction	28%
Quality failure	Detection of a defect requiring stop for inspection or correction	Stop to verify measurement, sensor cleaning, piece rejection	15%
Supply	Waiting for material, component, tool, or necessary information	Waiting for next batch, key search, waiting for quality OK	18%
Ergonomic	Interruption caused by momentary fatigue or uncomfortable posture	Operator stretches, changes position, briefly rests due to tiredness	10%
Informational	Search or consultation of instructions, drawings, or clarifications	Drawing consultation, work instruction search, clarification with supervisor	7%

This classification allows us not only to know how many micro-stops occur, but also to understand their nature to design specific countermeasures. A dominant category such as "Material jam" (28%) suggests feeding design or raw material quality problems. A high proportion in "Ergonomic" signals workstation design problems.

Critical Distinction: Micro-stop vs. Slow Cycle

A frequent and costly methodological mistake is confusing micro-stops with slow cycles. The distinction is essential for correct diagnosis:

• Micro-stop: The process stops completely. Instantaneous speed = 0. The operator is actively resolving the problem.
• Slow cycle: The process continues but at a speed lower than nominal. There is no stop, but a reduction in cadence.

In practice, a slow cycle can be more insidious because it is less visible. An operator who "takes their time" does not generate a stop, but reduces overall performance. The Work Sampling methodology with Snap Reading allows capturing both phenomena if the observation protocol is well designed.

Work Sampling Methodology: Statistical Foundation for a Reliable Diagnosis

Work Sampling is not a simple casual observation. It is a rigorous statistical technique based on inferential statistics. Its power lies in the fact that, from a representative sample of random observations, we can estimate the real proportion of time a process spends in each state with a defined confidence level.

Theoretical Basis: From Bernoulli to the Gauss Curve

Each observation in Work Sampling is a Bernoulli experiment: either we capture the event (micro-stop) or we do not. This is modeled with a binomial distribution. However, when the sample size is sufficiently large (typically N > 30), the Central Limit Theorem allows us to approximate the distribution of the sample proportion using a normal distribution (Gauss Curve).

This is the statistical magic that gives validity to the method. We do not need to observe continuously for hours. We need a sufficient number of random and independent observations.

The confidence interval formula for a proportion is:

$\hat{p} \pm Z_{\alpha/2} \cdot \sqrt{\frac{\hat{p}(1-\hat{p})}{N}}$

Where:

• $\hat{p}$ is the sample proportion (e.g., 0.08 for 8% micro-stops)
• $Z$ is the value of the normal distribution for the desired confidence level
• $N$ is the sample size

Sample Size Calculation (N): Precision vs. Cost

The sample size is not chosen at random. It is calculated based on two parameters the engineer must decide:

1. Confidence Level (Z): How confident do I want to be? (95% is the industrial standard)
2. Margin of Error (e): What precision do I need? (±2% is usually sufficient for diagnosis)

The formula for determining the minimum N is:

$N = \frac{Z^2 \cdot p(1-p)}{e^2}$

Examples applied to micro-stop diagnosis (estimated p = 8%):

Confidence Level	Z Value	Margin of Error (e)	Observations Required (N)	Practical Interpretation
95%	1.96	±2%	~707	Standard for diagnosis. ~3-4 working days of observation.
95%	1.96	±1.5%	~1,253	For high-precision or benchmarking studies.
99%	2.576	±2%	~1,216	Maximum confidence, for high-investment decisions.
90%	1.645	±2%	~496	Preliminary or quick study. 2 working days.

Practical tip: When the real proportion (p) is unknown, p = 0.5 is used as a conservative estimator, which maximizes the required N. A pilot study of 50-100 readings allows obtaining a preliminary p and optimizing the sampling design, saving time and resources.

Tippett Method: Beyond the Proportion

The Tippett method, developed in 1934 for the British textile industry, takes Work Sampling one step further. It not only estimates how much time is lost, but also allows reconstructing the distribution of durations of the micro-stops.

Tippett method protocol:

1. Perform random observations according to the calculated temporal design.
2. For each observation, record the state: PRODUCTIVE / MICRO-STOP / MAJOR STOP / INACTIVE.
3. Additionally, when a micro-stop is captured, record its estimated or timed duration.
4. Build a frequency histogram of durations.
5. Fit to a theoretical distribution (typically negative exponential or log-normal).

Key empirical finding: In discrete production environments, the distribution of micro-stop durations often follows a log-normal distribution. This means that most are very brief (median of 15-45 seconds), but there exists a "long tail" of less frequent events that can extend up to 3-5 minutes. This long tail is what has the most impact on overall performance.

Snap Reading Observation Protocol: Minimizing the Hawthorne Effect

The statistical validity of Work Sampling depends critically on the quality of the observations. The greatest methodological enemy is the Hawthorne Effect: the tendency of workers to modify their behavior when they know they are being observed.

The Snap Reading protocol is specifically designed to mitigate this effect. The idea is simple but powerful: the observer captures a "snapshot" of the process state at a random moment, without interacting.

Practical Guide for Field Observers

1. Strict Temporal Randomization:

• Generate an observation time sequence using a random number generator.
• Example: Observe at 9:07, 9:23, 9:51, 10:15, etc. Avoid patterns such as "every 15 minutes."
• Use tools like Cronometras to schedule and manage these random sequences efficiently.

2. Spatial Randomization:

• Vary the order of visits to workstations.
• Do not always observe the same line or machine in the same order.

3. Settling Period:

• Discard the first 2-3 working days of observation. Operators initially "behave well," but return to their normal behavior after a few days.

4. Discrete Observation Technique:

• The observer must be "invisible": do not interact, do not ask, do not comment on findings.
• Position yourself at a point with clear vision but without interfering.
• Record the state at the exact instant of the observation, without averaging.

5. Minimum Study Duration:

• A reliable study must capture variability between shifts, days of the week, and different operators.
• Minimum: 5 full working days. Recommended: 10 working days for high confidence.

6. Observer Training:

• The observer must be able to distinguish in milliseconds between "machine in productive cycle" and "machine in micro-stop with operator present."
• The distinction between "micro-stop" and "slow cycle" requires specific training.

Impact on OEE: Connecting Micro-stops with Key Metrics

OEE (Overall Equipment Effectiveness) is the gold standard for measuring productivity in manufacturing. It is broken down into three factors:

$OEE = Availability \times Performance \times Quality$

Micro-stops primarily impact the Performance factor, which is calculated as:

$Performance = \frac{Ideal\ cycle\ time \times Unit\ count}{Effective\ operating\ time}$

Quantifying the Impact on Performance

If our Work Sampling reveals that 8% of the observations capture micro-stops, this translates directly into a performance loss. The calculation is more subtle than simply subtracting that 8%.

Corrected formula for the Performance factor considering micro-stops:

$Performance = \frac{Real\ unit\ count}{Theoretical\ count\ based\ on\ ideal\ cycle\ time}$

Where the "theoretical count" must consider that during the micro-stop time no unit is produced. Therefore, a line with 8% micro-stops can never reach 100% theoretical performance.

Domino Effect: When Micro-stops Affect Availability

There is a critical cross effect that many analyses overlook. When micro-stops are extremely frequent or of variable duration, they can:

• Accumulate and generate bottlenecks downstream.
• Lead to major stops (e.g., a poorly resolved jam breaks a component).
• Generate accelerated wear due to frequent start/stop cycles.

In these cases, micro-stops begin to also impact the Availability factor, multiplying their negative effect.

OEE Without Sensors: The Power of Visual Diagnosis

One of the great advantages of Work Sampling is that it allows estimating OEE without needing to install additional sensors on machinery. Through random observations, we can estimate:

• Availability: Proportion of observations where the machine is in "major stop."
• Performance: Proportion of observations where the machine is in "micro-stop" or "slow cycle."
• Quality: Proportion of observations where "quality failure" is detected or inspections are carried out.

Platforms like Induly can complement this visual diagnosis with real-time production control data, validating Work Sampling estimates with transactional data.

Practical Case: Performance Loss Calculation from Observation Data

Let's apply the theory with a real numerical example. Suppose an assembly line with an ideal cycle time of 0.5 minutes per unit (120 units/hour).

Work Sampling study data (5 working days, 800 observations):

• "PRODUCTIVE" observations: 680 (85%)
• "MICRO-STOP" observations: 64 (8%)
• "MAJOR STOP" observations: 32 (4%)
• "INACTIVE" observations: 24 (3%)

Confidence interval calculation for the proportion of micro-stops:

• Sample proportion: $\hat{p} = 64/800 = 0.08$
• Standard error: $\sqrt{\frac{0.08(1-0.08)}{800}} = 0.0096$
• For a 95% confidence level (Z=1.96): Margin of error = $1.96 \times 0.0096 = 0.0188$
• Confidence interval: $0.08 \pm 0.0188$ → [6.12%, 9.88%]

Interpretation: With 95% confidence, the real proportion of time lost in micro-stops is between 6.12% and 9.88%.

Performance loss calculation:

• Effective operating time = 100% - 4% (major stops) = 96%
• But during that 96%, only 85% of the time is productive (8% is micro-stop, 3% is inactive)
• Real performance = 85% / 96% = 88.5% (compared to a theoretical performance of 100%)
• Loss attributable to micro-stops and inactivity = 11.5%

In an 8-hour shift, this equals 55.2 minutes lost per shift, or approximately 66 units not produced (at 120 units/hour).

Critical Differentiation: Micro-stop vs. Slow Cycle

We reinforce this distinction because it is the most common mistake in field studies and leads to completely wrong diagnoses.

Characteristic	Micro-stop	Slow Cycle
Process state	Completely stopped	In motion but at reduced speed
Instantaneous speed	0	>0 but < Nominal speed
Typical cause	Blockage, jam, adjustment	Fatigue, inexperience, demotivation, machine wear
Visibility in Snap Reading	High (stopped machine is clearly seen)	Low (requires comparison with standard cycle time)
Impact on OEE	Mainly on Performance	Exclusively on Performance
Typical countermeasure	Root cause solution (design, maintenance)	Training, motivation, parameter adjustment

How to capture both phenomena in a study:

1. For micro-stops: Classic Snap Reading is perfect. If at the instant of observation the machine is stopped and the operator is present, it is a micro-stop.
2. For slow cycles: A variant of the method is required. The observer must measure the time between two complete cycles and compare it with the standard cycle time. If significantly longer, there is a slow cycle.

Conclusions and Action Steps

Micro-stops represent a significant and frequently missed improvement opportunity in manufacturing. Their "invisible" nature to conventional systems should not be confused with a minor impact.

Key Findings:

1. Magnitude: Micro-stops typically consume between 3% and 12% of gross operating time in discrete production lines.
2. Reliable diagnosis: Work Sampling with random observations (Tippett method) is the statistically valid tool to quantify them without hardware investment.
3. Impact on OEE: They mainly affect the Performance factor, but can lead to Availability losses if frequent or poorly managed.
4. Essential classification: A MECE taxonomy allows going beyond quantification and understanding the causes to design effective solutions.

Steps to Start a Diagnostic Study:

1. Define the scope: Select the line, process, or study area. Ideally, one with unexplained performance problems.
2. Design the sampling: Calculate the sample size (N) based on the desired confidence level and precision. A study of 700-800 observations is a good starting point.
3. Train observers: Ensure they understand the MECE taxonomy and the micro-stop/slow cycle distinction.
4. Execute the Snap Reading protocol: With strict temporal and spatial randomization for at least 5 working days.
5. Analyze data: Calculate confidence intervals, duration distribution, and MECE categorization.
6. Quantify impact: Translate the percentages to minutes and units lost per shift/day/month.
7. Prioritize actions: Focus first on the most frequent or impactful MECE categories.
8. Establish follow-up: Repeat the study periodically to measure the effect of implemented countermeasures.

These methods engineering and time study techniques, far from being obsolete, are the quantitative foundation on which the most significant productivity improvements are built. Digital tools like Cronometras for time analysis or WorkSamp for work sampling are making these methodologies more accessible and accurate than ever.

The hidden productivity is there, waiting to be discovered. It just needs to be measured with statistical rigor to be managed.

---

Resources and Tools

To deepen these methodologies and find specialized tools:

• ASETEMYT Directory: The complete directory of industrial time study tools and services.
• Cronometras: Specialized software for time and motion analysis with automatic generation of random sequences.
• Induly: Production Control and Industrial Time Clock platform that complements Work Sampling studies with real-time data.
• ASETEMYT Blog: Technical articles and case studies on industrial productivity.
• Add to Directory: If you know a tool or service that should be in the directory.

Have you conducted Work Sampling studies in your plant? What percentage of micro-stops have you identified? Share your experiences in the comments.