When Normality Changes: Exploring ANF in Smoking Behavior and Aerosol Research

When Normality Changes: Exploring ANF in Smoking Behavior and Aerosol Research

From individual smoking patterns to experimental aerosol deposition and exposure systems

What Is Normal for a Smoker?

Imagine a person who has smoked approximately 15 cigarettes per day for many years. Their consumption is remarkably consistent, fluctuating between 12 and 18 cigarettes daily. From a statistical perspective, this behavior might be considered normal for that individual.

Now suppose consumption gradually increases to 20, then 25 cigarettes per day. At what point should we conclude that the person’s normal behavior has changed? And what if the number of cigarettes remains unchanged, but the smoker begins taking longer, deeper, or more frequent puffs?

These questions illustrate a potentially interesting application of the Area of Normal Functioning (ANF), a framework I have been developing from my earlier research in industrial robotics, statistical exception detection, and computer performance management.

The central idea is to monitor not only individual measurements but also the boundaries of the region within which a system normally operates.

Statistical normality does not imply safety or health. A stable smoking pattern may be statistically normal for an individual while remaining harmful. There is no safe level of cigarette smoking. This distinction between observed normality and independently established acceptable limits is fundamental to ANF.

1. From Industrial Robotics to the Area of Normal Functioning

The ANF concept originated in my research on industrial robot grasping. A robot could successfully grasp an object within a certain range of deviations from its nominal position. Beyond that range, the grasping operation failed. By experimentally varying object positions and observing the outcome, it was possible to establish the region within which the operation remained successful.

Later, I extended related ideas to computer performance and capacity management, developing the Statistical Exception and Trend Detection System (SETDS), Exception Value (EV), and IT-Control Charts.

For a monitored variable X, the statistical ANF can be expressed as:

ANFₓ(t) = [Lₓ(t), Uₓ(t)]

Here Lₓ(t) and Uₓ(t) represent the lower and upper boundaries of observed normal behavior. These boundaries are not necessarily fixed. They may evolve with workload, environmental conditions, system configuration, or other drivers. ANF therefore offers a way to distinguish isolated exceptions from sustained trends, structural changes, and transitions between operating regimes.

Could similar ideas be useful in tobacco and aerosol research?

2. ANF for Smoking Behavior

Smoking behavior is multidimensional. Potential measurements include:

·         Cigarettes consumed per day

·         Time between cigarettes

·         Time to the first cigarette after waking

·         Puff volume and duration

·         Number of puffs per cigarette

·         Selected biomarkers of exposure

A smoker may maintain a stable cigarette count while changing puffing behavior or nicotine intake. This raises the possibility of establishing separate but interconnected ANF regions for consumption, smoking topography, and exposure:

ANFsmoking(t) = {ANFᴄ(t), ANFᴛ(t), ANFᴇ(t)}

Here C, T, and E represent consumption, topography, and exposure. A change in one region may occur without an obvious change in another.

One possible application is investigating compensatory smoking, where changes in product characteristics may be accompanied by changes in smoking intensity. However, compensation is not universal and should not be assumed. It is an empirical question requiring measurements of actual behavior and exposure. The purpose of ANF would be to identify and characterize changes, not to determine whether a particular level of tobacco consumption is safe.

3. From Smoking Behavior to Experimental Aerosol Research

While considering smoking behavior, I realized that an even more promising application might be found in the laboratory systems used to characterize tobacco and electronic nicotine delivery system aerosols.

In particular, air-liquid-interface (ALI) exposure systems allow researchers to expose biological cell cultures to aerosols under controlled experimental conditions. A major challenge is understanding the relationship between aerosol generation, particle transport, deposition efficiency, spatial uniformity, and the dose actually delivered to biological samples.

This problem is closely related to a fundamental question in ANF modeling: What is the normal operating region of an experimental system, and how can we determine when its behavior has changed?

Unlike an individual’s smoking behavior, laboratory systems may offer the opportunity to establish ANF boundaries experimentally under controlled conditions. This connects directly to the original robotics interpretation of ANF.

4. Existing Research on Aerosol Deposition and Exposure Systems

Several publications involving aerosol researcher Pasha Kosachevsky and colleagues provide a useful starting point.

Deposition efficiency and uniformity in the Vitrocell 24/48 system

Oldham and colleagues investigated the deposition efficiency and uniformity of monodisperse solid particles in the Vitrocell 24/48 ALI exposure system. Their 2020 paper in Aerosol Science and Technology compared experimental measurements with computational fluid dynamics (CFD) predictions.

The study examined particles with aerodynamic diameters ranging from approximately 0.5 to 3.3 micrometers. The researchers found that deposition efficiency and spatial uniformity varied with particle size. Their experimental design also included repeated measurements on different days.

These observations suggest a possible ANF application: characterizing the expected deposition distribution under specified experimental conditions. Rather than considering only average deposition efficiency, ANF could represent the statistical variability of deposition across different locations.

Experimental measurements versus CFD predictions

In another 2020 study, Oldham, Kosachevsky, and colleagues investigated particle deposition in the Vitrocell AMES 48 ALI exposure system. They compared experimentally measured deposition efficiency and uniformity with CFD predictions.

The reported average deposition efficiencies ranged from approximately 0.007% to 0.43%, depending on particle size. The study found good agreement between average measured deposition efficiencies and CFD predictions, while deposition uniformity remained an important consideration.

This provides an interesting opportunity for extending ANF to model validation. Instead of monitoring only measured values, we could analyze differences between experimental measurements and predictions:

e(s,c) = Dmeasured(s,c) − DCFD(s,c)

Here s represents spatial position, c represents experimental conditions, and D represents deposition efficiency or another deposition-related measurement. An ANF could be established for these residuals, taking experimental uncertainty into account. Changes in the residual distribution might indicate differences between expected and observed system behavior. Such analysis would complement, rather than replace, established CFD validation and uncertainty-quantification methods.

Characterization of e-vapor exposure systems

In 2022, Zhang and colleagues published a study on the fit-for-purpose characterization of ALI exposure systems for e-vapor aerosols. The research evaluated commercially available exposure systems under specified experimental conditions. This suggests another ANF application: characterizing a multidimensional operating region using measurements associated with aerosol generation, transport, and exposure.

Rapid condensate collection systems

Also in 2022, Lalonde and colleagues investigated a rapid condensate collection apparatus for electronic nicotine delivery systems. Such experimental systems raise questions about collection performance, repeatability, and the effects of operating conditions. ANF-based monitoring could potentially provide an additional statistical framework for evaluating repeated measurements and identifying systematic changes.

These published studies do not establish ANF or demonstrate its superiority. They provide experimental contexts in which the proposed methodology might be tested.

5. A Multidimensional ANF for Aerosol Deposition

My earlier work on exception-based modeling treated system performance as a multidimensional data space. Different projections of that space could be used to construct control charts and identify unusual behavior. A similar approach might be applicable to aerosol deposition:

ANFᴅ(s,c,t)

Here D is a deposition-related measurement, s identifies spatial location, c represents experimental conditions, and t represents time or experimental sequence, when available.

We could investigate three complementary types of ANF:

·         Spatial ANF: Characterizes the normal distribution of deposited particles across measurement locations.

·         Model-residual ANF: Characterizes expected differences between experimentally measured deposition and CFD predictions.

·         Temporal ANF: Characterizes variability and changes across repeated experiments or measurement periods.

The first two can be studied without a conventional time series. The third is especially suitable for SETDS and IT-Control Chart methods when sufficient sequential data exist.

An important methodological challenge is distinguishing normal spatial differences from statistically significant changes. If deposition varies systematically across an exposure plate, those differences may be normal characteristics of the system rather than anomalies. The model must therefore account for spatial position and experimental conditions before interpreting deviations.

6. Experimentally Established ANF Boundaries

One of the most interesting aspects of this application is the opportunity to establish ANF boundaries experimentally. In robotics, I investigated the limits of successful grasping by systematically changing object positions. In aerosol research, a related approach might involve controlled variation of particle sizes, flow conditions, exposure configurations, or other relevant experimental parameters.

The objective would be to characterize the region within which the system satisfies independently specified performance criteria. This is distinct from estimating the statistical distribution of routine observations.

·         Observed ANF: The region representing statistically typical measurements under specified conditions.

·         Experimental ANF: The region of operating conditions demonstrated through controlled experiments to meet defined performance requirements.

·         Acceptability limits: Independent scientific, technical, regulatory, or health-related criteria.

These three concepts should not be confused. A system might exhibit highly repeatable deposition measurements while failing to achieve the uniformity required for a particular experiment. Statistical stability alone does not establish experimental suitability.

7. Can SETDS Add Value?

The practical research question is whether ANF and SETDS can provide useful capabilities beyond existing statistical methods. Possible applications include:

·         Exception detection: Identifying unusual measurements relative to condition-specific reference regions.

·         Change-point detection: Recognizing systematic changes across repeated experimental runs.

·         Trend detection: Identifying gradual changes in measurement behavior.

·         Multidimensional analysis: Detecting changes in relationships among deposition efficiency, spatial uniformity, and operating conditions.

·         Driver-based modeling: Explaining observed variability through particle characteristics, flow parameters, or other experimental factors.

For example, a laboratory system could maintain approximately the same average deposition efficiency while developing a less uniform spatial distribution. A single average metric might not adequately characterize that change. A multidimensional ANF might provide a more informative representation.

However, this remains a hypothesis. Its value must be demonstrated through comparison with conventional statistical process control, residual analysis, uncertainty quantification, and established reproducibility measures.

8. A Possible Research Study

A practical starting point would be to analyze an existing experimental dataset. Ideally, it would include:

·         Deposition measurements at multiple spatial locations

·         Repeated measurements under comparable conditions

·         Corresponding CFD predictions, where available

·         Experimental settings and relevant metadata

·         Measurements under different operating conditions

We could then establish baseline ANF regions and investigate their ability to detect meaningful deviations. Depending on the available data, the study could focus on spatial deposition variability, CFD prediction residuals, or temporal changes in experimental performance.

For a rigorous evaluation, we would need to separate baseline and evaluation data and quantify detection performance, including false alarms and sensitivity to known changes. An important question would be whether ANF provides additional insight compared with existing methods, rather than simply offering a different visualization of the same statistical information.

If the results are promising, this could form the basis of an interdisciplinary conference paper.

9. When Normality Changes

The broader significance of ANF extends beyond tobacco research. In computing systems, a stable workload may conceal changes in resource consumption. In human behavior, stable activity may conceal changes in physiological response. In experimental aerosol systems, stable average measurements may conceal changes in spatial distribution or relationships among measured variables.

In all these cases, the essential question is not merely whether an individual measurement is unusual. It is whether the underlying region of normal functioning has changed. This is the central idea behind ANF.

The potential application to aerosol deposition and exposure systems is still exploratory, but the existing experimental literature suggests several testable research questions. I would welcome feedback from researchers in aerosol science, laboratory measurement systems, computational fluid dynamics, and statistical modeling.

Can monitoring the normal operating region itself provide useful information that conventional measurement-by-measurement analysis overlooks? That is the question I would like to investigate.

References

1. Oldham, M. J., et al. (2020). Deposition efficiency and uniformity of monodisperse solid particle deposition in the Vitrocell 24/48 Air-Liquid-Interface in vitro exposure system. Aerosol Science and Technology, 54(1), 52–65. https://doi.org/10.1080/02786826.2019.1676877

2. Oldham, M. J., et al. (2020). Comparison of experimentally measured and computational fluid dynamic predicted deposition and deposition uniformity of monodisperse solid particles in the Vitrocell AMES 48 air-liquid-interface in-vitro exposure system. Toxicology in Vitro, 67, 104870. https://doi.org/10.1016/j.tiv.2020.104870

3. Zhang, J., et al. (2022). Fit-for-purpose characterization of air-liquid-interface (ALI) in vitro exposure systems for e-vapor aerosol. Toxicology in Vitro. https://doi.org/10.1016/j.tiv.2022.105352

4. Lalonde, G., et al. (2022). Characterization of a rapid condensate collection apparatus for in vitro assays of electronic nicotine delivery systems. Toxicology in Vitro. https://doi.org/10.1016/j.tiv.2022.105434

5. National Cancer Institute. “Light” Cigarettes and Cancer Risk. https://www.cancer.gov/about-cancer/causes-prevention/risk/tobacco/light-cigarettes-fact-sheet

Author’s note: This article presents a proposed extension of the Area of Normal Functioning methodology. ANF has not yet been validated by the author for aerosol deposition or tobacco exposure systems. The proposed applications are intended as research hypotheses.

Igor Trubin  |  TruTech Development, LLC  |  November 2026

When Normal Moves: The Overton Window Through the Lens of ANF

 Political Normality, Hidden Degradation, and the Drivers of Change

Igor Trubin | October 2026

Introduction: Is Normality a Political Concept?

Recently, while discussing my ongoing book project on the Area of Normal Functioning (ANF) with my friend Eric, an unexpected question emerged: Can the concept of normality also be applied to politics?

After years of studying normal functioning in industrial robotics, computer performance, and more recently human and organizational systems, I had not seriously considered political science as another possible application of ANF.

Eric introduced me to the Overton Window, a concept developed by Joseph P. Overton in the 1990s. It describes the range of political ideas and policies that are considered acceptable or politically feasible at a particular time.

The boundaries of this window are not fixed. They may shift, expand, or contract as public attitudes and social expectations evolve.

That immediately reminded me of ANF.

Our subsequent discussions raised two additional questions:

• What happens when the boundaries of normality change in a way that conceals a gradual deterioration in system performance or behavior?

• Can we identify the independent factors—or drivers—associated with such changes and use them to forecast future shifts?

These questions connect political normality with my earlier work on statistical exception detection, trend analysis, and business-driver modeling.

This article explores those connections as possible directions for further research.

1. The Overton Window: Political Normality Is Dynamic

The Overton Window describes the range of policy ideas that political decision-makers can support without unacceptable political risk.

An idea considered politically impossible today may eventually become mainstream. Conversely, a previously acceptable policy may move outside the range of politically feasible options.

The window may move in different directions, and its width may change.

This resembles the basic idea behind the Area of Normal Functioning.

In ANF, normality is not necessarily a fixed condition. It is a region whose boundaries may evolve as system behavior, workload, environment, or other operating conditions change.

The important question is therefore not only what is normal today, but how the boundaries of normality change over time.

However, an important distinction must be recognized.

The Overton Window concerns political acceptability and feasibility, while ANF was originally developed to describe the operating conditions under which a technical system functions normally.

These ideas are related but not identical. Their connection requires careful investigation rather than a simple transfer of terminology.

2. Similar Concepts in Other Disciplines

Eric also shared three interesting analogies from ecology, psychology, and safety engineering.

Shifting Baseline Syndrome — Ecology

In 1995, Daniel Pauly described how successive generations of fisheries researchers could accept increasingly depleted fish populations as their normal reference conditions.

Each generation begins with a different baseline, potentially losing sight of earlier environmental conditions.

As a result, long-term degradation may become less visible.

Adaptation-Level Theory — Psychology

Harry Helson's Adaptation-Level Theory describes how human judgments depend on reference levels shaped by previous experiences.

Something initially perceived as unusual may eventually become ordinary as people adapt.

The reference point used to define normality changes.

Normalization of Deviance — Safety Engineering

Diane Vaughan's research on the 1986 Space Shuttle Challenger disaster examined how repeated departures from established standards could gradually become accepted within an organization.

When deviations repeatedly occur without immediate failure, they may begin to appear safe or routine.

Eventually, exceptional behavior can become standard practice even when the underlying risk remains.

These examples illustrate different mechanisms through which normality changes:

• In ecology, the historical baseline changes.

• In psychology, the perceptual reference level changes.

• In organizations, the boundaries of tolerated behavior change.

• In politics, the boundaries of acceptable policies and practices change.

The common feature is a moving reference point.

Could these processes be studied through a common analytical framework?

3. From the Overton Window to Political ANF

In my earlier research, the Area of Normal Functioning was represented as a region bounded by lower and upper limits.

For a measurable variable X:

ANF_X(t) = [L_X(t), U_X(t)]

Here, L and U represent the lower and upper boundaries of normal functioning at time t.

Rather than being fixed thresholds, these boundaries can be estimated from observations and updated as the system evolves.

In subsequent work on the Statistical Exception and Trend Detection System (SETDS), related methods were used to detect exceptions, identify trends and change points, and distinguish isolated anomalies from persistent changes in system behavior.

Could a similar approach help study political normality?

Consider a specific political issue whose policy alternatives can reasonably be represented along a numerical scale.

Repeated surveys might allow researchers to estimate the range of positions considered acceptable at different times.

A preliminary representation could be:

ANF_P(t) = [L_P(t), U_P(t)]

where the boundaries describe an estimated range of political acceptability for issue P.

This is a proposed adaptation of ANF, not an established political-science measurement method.

Its practical usefulness depends on the data, the definition of acceptability, and whether an interval is an adequate representation.

Importantly, political scientists have already developed quantitative approaches to policy acceptability.

For example, Matthew Jenkins and Daniel Gomez, in Unveiling the Multidimensional Nature of Policy Acceptability: A Cross-National Approach (first published in 2025), distinguish political, affective, and deliberative dimensions of policy acceptability.

Such work could provide a valuable measurement foundation for investigating how acceptance changes over time.

ANF's potential contribution would not be to invent a measurement of political acceptability, but to investigate its changing boundaries, exceptions, structural transitions, and possible future trajectories.

4. Detecting Changes in Political Normality

In computer performance monitoring, a system may remain within a relatively stable operating region for an extended period.

Then something changes.

A new workload, software release, hardware modification, or environmental condition may produce a gradual trend or a sudden shift.

SETDS and IT-Control Charts were developed to identify statistical exceptions and changing behavior in such data.

Political processes may exhibit analogous statistical patterns.

Imagine repeated surveys concerning the public acceptability of AI-assisted government services.

Initially, acceptance is limited. Over several years, it may increase gradually or change abruptly following technological or institutional developments.

The data could reveal several distinct phenomena:

Stable periods: Public attitudes fluctuate within a relatively consistent range.

Exceptions: Individual observations differ significantly from the established statistical pattern.

Trends: Acceptance increases or decreases persistently.

Change points: The statistical properties of the time series change.

Boundary migration: The estimated region of acceptable positions moves.

Boundary expansion or contraction: The range of politically acceptable alternatives becomes wider or narrower.

These processes resemble those investigated in technical performance analysis.

There is also a connection to Punctuated Equilibrium Theory, associated with Frank Baumgartner and Bryan Jones, which examines periods of relative policy stability interrupted by substantial change.

ANF might complement such research through explicit estimation and monitoring of dynamic boundaries.

5. When the New Normal Hides Degradation

This connection leads to a particularly important question, extending far beyond politics:

What happens when a system adapts to deteriorating conditions and begins treating them as normal?

Consider a computer application whose response time gradually increases.

An adaptive statistical monitoring system may detect the trend and eventually establish a new baseline.

The application could then appear statistically stable again.

However, its response time may now exceed the service level required by the business.

In other words, the system has established a new statistical normal without necessarily restoring acceptable performance.

The same problem can arise elsewhere.

In ecology, the baseline used to evaluate environmental conditions may gradually shift toward a degraded state.

In organizational management, recurring procedural violations may become tolerated because they have not yet produced visible consequences.

In politics, repeated behavior may become increasingly familiar or accepted even when its relationship to established institutional standards has not changed.

This suggests the importance of distinguishing two kinds of boundaries.

Observed ANF

The region estimated from the system's actual measured behavior.

It describes what is statistically typical under observed conditions.

Reference operating limits

The region defined by independent requirements, safety criteria, experimental results, service-level objectives, or other externally justified standards.

These limits describe what the system is expected or required to achieve.

Unlike the observed ANF, reference limits should not automatically move merely because the observed behavior changes.

They may themselves require revision, but such revision should be based on independent justification.

The difference between these regions can provide useful evidence of possible degradation.

It is important to emphasize that a moving ANF does not by itself prove deterioration. A shift may reflect adaptation, improvement, environmental change, or a new operating mode.

Degradation requires an independent definition of what constitutes an undesirable outcome.

This leads to a potentially general principle:

Statistical stability is not the same as healthy functioning.

I believe this distinction may be relevant to many dynamic systems within the broader scope of ANF.

6. Why Normality Changes: From SETDS to Model Factory

So far, we have considered how to identify changes in normal behavior and how to distinguish adaptation from possible degradation.

But another question remains:

Why do the boundaries of normality move?

This question brings us back to an important component of my earlier SETDS research.

In a 2018 blog post on anomaly and novelty detection, I described three related levels of SETDS.

The first detects statistical anomalies and outliers.

The second uses trend and change-point analysis to identify persistent changes that may indicate a new operating regime.

The third relates the observed performance data to external demand indicators, or business drivers, to develop statistically supported forecasting models.

I implemented this modeling direction in an approach called Model Factory.

Related methods are also described in my work on computer resource modeling, including US Patent 10,437,697, Systems and Methods for Modeling Computer Resource Metrics, on which I am a co-inventor.

That patent describes automated construction and evaluation of regression models connecting resource utilization with business transaction volumes, including statistical diagnostics, model selection, and forecasting.

The key methodological principle is that the behavior of a system should be analyzed in relation to measurable external demand rather than exclusively as an isolated time series.

Could this idea be extended beyond computer systems?

Political normality and independent drivers

Suppose we observe a persistent change in public acceptance of a particular policy.

Instead of merely noting that acceptance changed around a certain historical event, we could examine whether the change is statistically associated with independently measured factors, such as:

• economic conditions;

• adoption of relevant technologies;

• levels of institutional trust;

• changes in demographic composition;

• exposure to information or public discussion.

The purpose would be to identify measurable relationships and determine whether they improve explanation and forecasting.

For example, the adoption of AI-assisted government services might be examined alongside measurements of technological familiarity, direct experience with such services, and institutional trust.

The selected drivers would need to be defined and measured independently of the outcome under investigation.

Statistical validation would be essential.

It would include examining model fit, stability, time lags, competing explanations, and forecasting performance using observations not employed in model construction.

And an important limitation remains:

Statistical correlation does not automatically establish causation.

Political processes are complex, and relationships may be affected by hidden variables, feedback effects, and changing measurement conditions.

Nevertheless, driver-based modeling could provide a more systematic approach than simply associating observed changes with historical events.

7. A Driver-Conditioned ANF

The preceding discussion suggests a further possibility.

In the simplest representation, ANF depends on time:

ANF_X(t) = [L_X(t), U_X(t)]

But some boundaries may also depend on external operating conditions.

We could therefore explore a conditional representation:

ANF_X(t | B(t))

where B(t) is a vector of measured external drivers.

This notation means that the estimated normal region for X depends on the values of relevant drivers at time t.

For example, normal computer CPU utilization may depend on transaction demand.

Normal physiological measurements may depend on activity level, age, and environmental conditions.

Normal organizational workload may depend on staffing, incoming demand, and operational complexity.

Likewise, observed political attitudes might depend on economic conditions, technology adoption, and other measurable factors.

A useful model could distinguish between changes explained by measured drivers and changes that remain unexpected even after accounting for those drivers.

However, two different modeling tasks must be kept separate:

1. Explaining or forecasting the observed value of a metric using independent drivers.

2. Estimating whether the boundaries of its normal operating region also change with those drivers.

The first is closely related to established Model Factory methods.

The second is a proposed ANF extension requiring additional mathematical formulation and experimental verification.

Together, they offer a possible path from exception detection to more explanatory and predictive modeling of dynamic normality.

8. What If There Is No Single Political Normal?

Political normality introduces another challenge: different groups may have different definitions of what is acceptable.

A single average can conceal substantial differences among populations.

Consequently, a political ANF may need to be multidimensional or represented by several distinct regions.

Relevant dimensions might include:

• political issue;

• population group;

• geographic region;

• type of acceptability;

• time and external conditions.

For strongly polarized or multimodal distributions, one interval with lower and upper boundaries may be misleading.

In my earlier work on exception-based modeling, the normal functioning region could be represented in a multidimensional space, with different projections used for analysis.

That perspective may be useful here as well.

The challenge is not necessarily to find one universal political norm, but to characterize the structure and evolution of multiple observed normal regions.

9. Testing Political ANF with Real Data

Fortunately, political science provides historical datasets that could support exploratory experiments.

One possibility is Bright Line Watch, which publishes surveys of experts and the public concerning democratic principles and institutional performance.

A research experiment could be organized into five stages.

Stage 1 — Measurement

Choose a clearly defined indicator measured consistently across time. Determine whether it represents acceptability, preference, institutional performance, or another construct.

Stage 2 — Dynamic ANF estimation

Construct the time series, estimate the normal operating region, and detect exceptions, trends, and change points.

Stage 3 — Degradation assessment

Compare the observed ANF with independently defined criteria or preserved historical reference conditions where justified.

This step is essential when the research question concerns possible deterioration rather than merely statistical change.

Stage 4 — Driver analysis

Identify independently measured candidate drivers and test their statistical relationships with the observed metric or ANF boundaries.

Model Factory principles could guide automated model selection, validation, and forecasting.

Stage 5 — Evaluation

Compare the results against conventional statistical methods and existing political-science interpretations.

Account for uncertainty, survey sampling, question wording, time-series dependence, and possible feedback between variables.

An important practical limitation is that some political surveys are conducted only periodically. Their sampling frequency may be insufficient to identify short-lived exceptions or accurately estimate certain change points.

Such an experiment would not establish a universal theory of political normality.

It could, however, help determine whether ANF and SETDS offer a useful complementary perspective.

10. What This Means for the ANF Book

My ongoing book project explores extending the Area of Normal Functioning from its origins in industrial robotics to technical, AI, biological, human, organizational, and other dynamic systems.

Political normality was not part of my original plan.

Yet the discussion with Eric has suggested several broader research directions.

The first is the relationship between adaptation and degradation.

A dynamic system may adjust to changing conditions and establish a new stable operating region. But adaptation alone does not guarantee that the new state is desirable or safe.

The second is the relationship between normality and its drivers.

Understanding how a system changes is only part of the problem. Identifying the independent factors associated with its behavior may help explain those changes and forecast future developments.

These topics are sufficiently general to deserve separate treatment in the book, extending beyond politics.

For now, political ANF remains an exploratory concept. Its scientific usefulness must be established through careful definitions, appropriate measurements, and empirical comparison with existing methods.

But the discussion raises three questions that seem increasingly central to the broader ANF project:

How do we detect when normality changes?

How do we recognize when a new normal conceals degradation?

And can we identify the drivers that explain and help predict those changes?

These questions connect industrial robotics, computing, organizational management, human behavior, and now political science.

They also bring me back to a possible title for the book:

Everything Is Normal… Until It Isn't.

Perhaps the most intriguing possibility is that the system may continue to appear normal—even after it has begun moving toward failure.

---

References and Further Reading

1. Mackinac Center for Public Policy. The Overton Window of Political Possibility.

2. Pauly, D. (1995). Anecdotes and the Shifting Baseline Syndrome of Fisheries. Trends in Ecology & Evolution, 10(10), 430.

3. Helson, H. (1948). Adaptation-Level as a Basis for a Quantitative Theory of Frames of Reference. Psychological Review, 55(6), 297–313.

4. Vaughan, D. (1996). The Challenger Launch Decision: Risky Technology, Culture, and Deviance at NASA. University of Chicago Press.

5. Baumgartner, F. R., & Jones, B. D. (1993). Agendas and Instability in American Politics. University of Chicago Press.

6. Jenkins, M., & Gomez, D. (2025, online; 2026 journal issue). Unveiling the Multidimensional Nature of Policy Acceptability: A Cross-National Approach. Policy Studies Journal, 54(1).

7. Helmke, G., & Rath, J. (2025). Defining and Measuring Democratic Norms. Annual Review of Political Science, 28, 233–251.

8. Trubin, I. (2018). Anomaly Detection vs. Novelty Detection: SETDS Method Detects and Separates Both. System Management by Exception.

9. Trubin, I., Schutt, M., & Robinson, J. (2019). Systems and Methods for Modeling Computer Resource Metrics. US Patent 10,437,697.

10. Trubin, I. (2026). Starting a Book Project: The Area of Normal Functioning.

Building the Area of Normal Functioning: From Industrial Robotics to Continuous Performance Testing

 By Igor Trubin

Can Normal Functioning Be Determined Experimentally?

In my recent work on the Area of Normal Functioning (ANF), I have focused primarily on statistical methods for identifying and tracking the boundaries of normal system behavior.

However, revisiting my earlier research in industrial robotics led me to another approach: ANF boundaries can also be determined experimentally.

This is not a new engineering practice. Engineers have been testing operational limits of technical systems for decades. What may be new is interpreting these experiments as a systematic way of constructing and maintaining the Area of Normal Functioning.

Even more interestingly, combining experimental testing with statistical monitoring creates an opportunity for continuously evolving ANF models.

1. The Original Idea: Experimental ANF in Industrial Robotics

During my early research with industrial robots, I investigated how a robotic gripper could reliably grasp an object even when the object's actual position differed from its expected position.

The traditional approach assumed that an object should be positioned accurately enough for successful grasping.

But what happens if it is displaced?

I investigated this question experimentally using a real industrial robot.

Starting from the ideal object position, I manually introduced positional deviations and attempted to grasp the object. By progressively increasing these deviations in different directions, I could identify the region within which the grasping operation remained successful.

Beyond certain limits, grasping failed.

The resulting region represented what I now describe as an Area of Normal Functioning for the robot-object interaction.

Importantly, the boundaries were not obtained by analyzing historical operational statistics. They were established through controlled experiments.

The approach could also accommodate several dimensions, including positional displacement, orientation, and other relevant conditions.

This was an early example of constructing an ANF directly from a system's observed capabilities.

2. The Same Principle Applies to Computer Systems

Now consider a modern computing application.

Its performance depends on multiple factors:

• Workload intensity and transaction arrival rates

• Number of concurrent users

• CPU, memory, storage, and network resources

• Application and infrastructure configuration

• Response-time and throughput requirements

For a particular system configuration, there is a range of workloads within which the application functions acceptably.

As the workload increases, the system may eventually violate response-time requirements, experience unacceptable error rates, or become unstable.

This is exactly what conventional load and stress testing investigates.

A load test can establish the workload range within which the application meets its operational objectives. A stress test can explore the limits beyond that range and reveal failure modes.

Why not interpret these experiments as constructing the Area of Normal Functioning?

For example, imagine an application with a response-time objective of less than 500 milliseconds at a specified percentile.

A series of load tests might establish that this objective is consistently met for workloads between 100 and 1,200 transactions per second.

Beyond that range, performance becomes unacceptable under the tested conditions.

The experimentally established workload interval can then be represented as:

ANF = [100, 1200] transactions/second

This interval is illustrative, and its validity depends on the test configuration, workload mix, performance objectives, and measurement uncertainty.

It is important to recognize that this example describes ANF in the workload dimension. The corresponding acceptable region in response-time, resource-utilization, and other dimensions may have a different shape.

In general, ANF is better understood as a multidimensional region than as a single interval.

3. Rebuilding ANF After Application Releases and Infrastructure Changes

Modern IT organizations routinely perform performance testing after significant changes, including:

• Application releases

• Hardware upgrades

• Infrastructure migrations

• Database configuration changes

• Cloud instance resizing

• Major architecture modifications

These tests usually answer familiar questions:

Did performance improve or deteriorate? Can the application handle the expected workload? Did the change introduce a regression?

But the same test results could answer a broader question:

How has the Area of Normal Functioning changed?

Suppose an application upgrade increases its sustainable transaction rate from 1,200 to 1,600 transactions per second while preserving the required response time.

Conventionally, we report a performance improvement.

From the ANF perspective, we can describe this as an expansion of the operating region.

Alternatively, an inefficient application release might reduce sustainable throughput from 1,200 to 900 transactions per second.

This is not merely a performance regression. It is a contraction of ANF.

Such changes could be tracked across application versions, hardware generations, or infrastructure configurations.

The resulting history would describe how the system's operating capabilities evolve over time.

4. What Happens Between Performance Tests?

Experimental determination of ANF is valuable, but it has a practical limitation.

We cannot continuously stress every production system to its operational limits.

Even when load tests are performed regularly, system behavior may change between experiments.

Workload composition may evolve. Infrastructure may experience degradation. Resource contention may increase. Seasonal demand patterns may emerge.

This is where my Statistical Exception and Trend Detection System (SETDS) becomes particularly relevant.

SETDS uses statistical analysis to identify exceptions, trends, and changes in system behavior.

Instead of experimentally pushing the system toward failure, SETDS observes its actual operating behavior and evaluates how that behavior changes.

For a measured performance metric X, the statistically estimated ANF can be represented by time-dependent boundaries:

ANF_X(t) = [L_X(t), U_X(t)]

Here L and U represent the lower and upper boundaries of the normal operating region estimated from historical observations.

These boundaries can change as the system evolves.

The key idea is to combine two complementary approaches:

Experimental ANF construction: Establish or validate acceptable operating limits using controlled performance testing.

Statistical ANF monitoring: Use SETDS to estimate normal operating patterns, detect exceptions, identify changes, and forecast the evolution of those patterns between experiments.

An important distinction must be maintained: a statistically normal operating region is not necessarily the complete region of acceptable operation.

For instance, an application may normally operate at 30–50% CPU utilization while its experimentally validated capacity allows substantially higher utilization without violating performance requirements.

Comparing these two regions could reveal unused capacity, emerging risks, and changes in operational headroom.

5. From Periodic Testing to Continuous ANF Reconstruction

There is another interesting possibility.

Some modern systems undergo automated performance testing continuously, particularly as part of software development and continuous integration processes.

My colleagues and I explored a related approach in our ICPE 2022 paper:

"Change Point Detection for MongoDB Time Series Performance Regression."

The research investigated performance measurements from automated tests triggered by changes to the MongoDB software repository.

We developed and evaluated a hybrid change-point detection approach that included the Perfomalist method and machine-learning techniques.

The objective was to identify significant performance changes without relying solely on manual analysis or fixed thresholds.

The reported change-point detection accuracy was 83%.

At the time, we described the problem primarily in terms of identifying performance regressions.

But now I see an opportunity to reinterpret and extend this work within the ANF framework.

Every automated performance test contributes evidence about a system's behavior under particular conditions.

When tests cover a sufficiently broad and controlled range of workloads, their results can contribute to reconstructing the ANF boundaries.

Meanwhile, change-point detection can identify moments when the previously established ANF model may no longer be valid.

For example:

1. A new software commit triggers automated performance tests.

2. Performance metrics are collected and compared with historical results.

3. SETDS or another change-point detection method identifies a significant behavioral change.

4. The system evaluates whether the change affects ANF boundaries.

5. If sufficient experimental evidence exists, the ANF model is updated; otherwise, additional targeted testing is initiated.

Thus, continuous performance testing could evolve from a mechanism for detecting regressions into a mechanism for continuous ANF reconstruction.

A single regression or change point would not necessarily establish a new boundary. It would provide evidence that a boundary may have changed and could trigger further validation.

This distinction is essential for making the approach scientifically sound.

6. ANF as a Multidimensional, Evolving Model

Another lesson from my robotics research is that normal functioning is rarely one-dimensional.

For robotic grasping, the successful operating region may depend on displacement along multiple axes, orientation, and grasping conditions.

For computer systems, the region may depend on workload, response time, resource availability, application configuration, and many other factors.

We can represent this more generally as:

ANF(t) = {x : the system satisfies defined functioning criteria under conditions x at time t}

The vector x describes the relevant operating conditions and system measurements.

Experimental testing samples this multidimensional space and helps identify the boundaries separating acceptable and unacceptable functioning.

Statistical monitoring examines system behavior within that space over time.

The combined approach makes it possible to investigate not only whether a metric is abnormal, but also:

• Whether the ANF region is expanding or contracting

• Whether its center is moving

• Whether its shape is changing

• Whether previously normal behavior is approaching a validated failure boundary

• Whether a structural change requires a new experimental assessment

In earlier work, I explored multidimensional performance models using a data-cube representation, with different projections producing control charts and other operational views.

The ANF concept offers a way to connect that work with experimentally established operating regions and their evolution.

7. Toward a Closed-Loop ANF Management Process

Combining experimental and statistical approaches suggests a practical ANF management lifecycle.

Step 1 — Construct: Establish operating boundaries through controlled experiments, including load, stress, and other relevant tests.

Step 2 — Monitor: Use SETDS to observe actual operational behavior and estimate statistical ANF boundaries.

Step 3 — Detect: Identify exceptions, trends, and change points that may indicate changes in the system's functioning.

Step 4 — Predict: Forecast whether system behavior is approaching a statistical or experimentally validated boundary.

Step 5 — Validate and Rebuild: Perform new experiments when necessary to verify or reconstruct ANF boundaries following significant changes.

This creates a feedback loop in which experimental testing and statistical monitoring reinforce one another.

The same principle may apply well beyond IT systems, including manufacturing, robotics, transportation, and other adaptive technical systems.

8. Conclusion: A Different Interpretation of Familiar Engineering Practices

The central idea is straightforward.

Performance testing determines what a system can do. Statistical monitoring describes what the system normally does and how that behavior changes. ANF provides a framework for connecting the two.

The individual methods are not new.

Robotic experiments, load testing, statistical process control, and change-point detection have long histories.

The proposed contribution is to integrate them into a common model of normal functioning, where the operating region itself becomes an object of measurement, analysis, forecasting, and reconstruction.

Looking back at my early industrial robotics experiments, I find it interesting that the same fundamental question remains relevant decades later:

How far can a system deviate from its expected operating conditions while still functioning successfully?

Today, with automated performance testing, statistical monitoring, and advanced analytics, we have the opportunity to answer that question continuously.

That may be an important next step in developing the Area of Normal Functioning framework.

---

References

1. Trubin, I. — 1986. Issledovanie protsessov zakhvatyvaniya obektov manipulirovaniya promyshlennym robotom [Research of Industrial Robot Grasping Processes]. Ph.D. dissertation (in Russian).

2. Leznik, M., Iqbal, M. S., Trubin, I., Lochner, A., Jamshidi, P., and Bauer, A. (2022). Change Point Detection for MongoDB Time Series Performance Regression. Companion of the ACM/SPEC International Conference on Performance Engineering (ICPE '22), pp. 45–48. https://doi.org/10.1145/3491204.3527488

3. Trubin, I. — Exception Based Modeling and Forecasting. CMG, 2008.

4. Perfomalist — Statistical exception, trend, and change-point analysis. https://perfomalist.com

Managing by Exception: Extending the Area of Normal Functioning to Corporate Teams

Exploring how ANF and SETDS can complement Agile, Kanban, and established management practices through statistical analysis and capacity forecasting

Introduction: From Technical Systems to Human Organizations

Throughout my career in IT Performance and Capacity Management, I have worked on a recurring problem: how to identify meaningful changes in complex systems without overwhelming administrators with unnecessary information.

This problem led me to develop the Statistical Exception and Trend Detection System (SETDS), along with methods for exception detection, change-point analysis, trend recognition, and forecasting.

The underlying management principle was relatively simple:

Do not intervene merely because something changes. Intervene when the change is meaningful.

This principle became central to my approach to Exception-Based Management.

More recently, while developing the broader concept of the Area of Normal Functioning (ANF), I began considering whether the same statistical principles could help us better understand corporate teams and organizational workloads.

What if we treated a team's workload as a dynamic system with its own normal operating range?

Could we distinguish ordinary fluctuations from meaningful structural changes? Could we anticipate emerging capacity problems? And could we reduce unnecessary managerial intervention while preserving team autonomy?

Initially, these possibilities appeared to offer a substantially different approach to organizational management.

However, examining established management frameworks reveals an important fact: many of these ideas already have significant precedents.

Agile emphasizes self-management and adaptation. Kanban provides quantitative flow metrics and forecasting. Statistical Process Control distinguishes common variation from exceptional conditions. PRINCE2 explicitly incorporates management by exception.

This raises a more interesting research question.

Rather than attempting to replace these approaches, could ANF provide an additional analytical framework that integrates statistical exception detection, structural change recognition, business-demand modeling, and proactive capacity forecasting?

I believe this possibility deserves further investigation.

1. Not Every Change Requires Management Intervention

Consider a software engineering team.

During one week, the team completes 25 tasks. During another, it completes only 19.

A manager looking at a conventional productivity dashboard might immediately question the decline.

But is there really a problem?

Perhaps the second week involved more complex tasks, unexpected technical difficulties, or necessary architectural improvements.

A single measurement rarely tells the whole story.

In technical systems, distinguishing normal variation from meaningful change has been studied for decades.

Statistical Process Control (SPC), originating in Walter Shewhart's pioneering work, introduced the distinction between common-cause and special-cause variation.

Thomas Nolan and Lloyd Provost, in their 1990 article Understanding Variation, emphasized the importance of this distinction for management decisions. They also explained why unnecessary adjustments to stable processes can actually increase variation.

The same principle applies to corporate teams.

Not every temporary decline in output represents a productivity problem. Not every increase in workload means that a team is approaching exhaustion.

We need to understand behavior over time.

This is where the ANF concept becomes relevant.

Instead of comparing every measurement against a fixed target or simple historical average, we can estimate a range of normal functioning based on observed behavior.

Suppose a team historically completes between 18 and 28 comparable work items per week.

Completing 19 items may be entirely consistent with normal variation.

A sustained decline to 12 items, however, could indicate a meaningful change requiring investigation.

The management principle becomes:

Normal variation should generally be tolerated. Meaningful deviations should be investigated, not automatically punished or corrected.

This is not a new management philosophy. What interests me is whether ANF and SETDS could provide an integrated statistical mechanism for implementing it.

2. What Existing Management Frameworks Already Offer

Before discussing the potential contribution of ANF, it is important to recognize the existing approaches.

Agile and Scrum: Autonomy and Adaptation

Agile principles emphasize sustainable development, teamwork, and responsiveness to change.

Scrum builds on transparency, inspection, and adaptation. It also recognizes teams as self-managing.

This philosophy is compatible with the objectives of ANF-based management.

Both approaches recognize that continuous external intervention is not necessarily productive.

However, Scrum does not prescribe a particular statistical method for calculating normal workload boundaries, detecting structural changes, or forecasting when team capacity constraints may emerge.

Such analytical methods could complement Scrum without changing its underlying philosophy.

Kanban: Managing the Flow of Work

Kanban is particularly relevant to the proposed application.

The May 2025 Kanban Guide identifies four fundamental flow metrics:

• Work in Progress (WIP).

• Throughput.

• Work Item Age.

• Cycle Time.

Kanban also supports probabilistic Service Level Expectations based on historical observations.

These metrics already provide valuable information about workload, predictability, and bottlenecks.

Furthermore, tools such as Jira include control charts displaying cycle times, rolling averages, and variability.

Therefore, introducing statistical charts into team management is not, by itself, a new contribution.

The research opportunity lies in investigating whether ANF/SETDS could add value through the integration of change-point detection, dynamically estimated operating boundaries, and forecasting of future operational constraints.

PRINCE2: Management by Exception

Another particularly relevant predecessor is PRINCE2.

Its Manage by Exception principle allows authority to be delegated within agreed tolerances.

When a project is expected to exceed these tolerances, the appropriate management level becomes involved.

This is conceptually close to the approach I have described as Exception-Based Management.

PRINCE2 also recognizes the importance of forecasting potential tolerance breaches before they occur.

An important distinction, however, is that project management tolerances are generally defined according to agreed objectives and constraints.

The ANF concept introduces a different type of boundary: one statistically estimated from observed system behavior.

These two types of boundaries should not be confused.

Statistical ANF boundaries describe what has historically occurred. Management tolerances describe what is considered acceptable.

Combining these perspectives could provide useful additional information for management decisions.

Statistical Process Control: A Common Foundation

Perhaps the closest methodological predecessor is Statistical Process Control.

SPC already provides methods for recognizing significant deviations from stable process behavior.

My SETDS work belongs to this broader tradition of statistical performance analysis, with particular emphasis on exceptions, change points, trends, and forecasting.

The proposed organizational application should therefore be understood as an extension and integration of established statistical ideas rather than a claim that statistical management itself is new.

The central research question is whether this particular combination can provide useful capabilities beyond those already available in conventional management analytics.

A Note About Terminology

There is also a terminology issue worth clarifying.

I have used Exception-Based Management to describe my management approach. However, Scrum.org already uses the abbreviation EBM for its distinct Evidence-Based Management framework.

Although the concepts are related through their reliance on information for management decisions, they are not the same methodology.

To avoid confusion, I will use ANF-based management or the full phrase Exception-Based Management when discussing my proposed approach.

3. From Reactive Management to Dynamic ANF Analysis

Having recognized these predecessors, we can formulate the proposed ANF application more precisely.

I envision it as an analytical layer that could complement existing organizational management frameworks.

Its potential value would come from combining four capabilities.

First: Estimating the Area of Normal Functioning.

Historical observations can be used to estimate a typical range of operating behavior.

Depending on the analytical method, boundaries might be represented by statistical control limits or distribution quantiles, such as P05 and P95, with P50 representing the median.

These methods have different statistical interpretations and should not be treated as interchangeable.

The purpose is to understand expected variation rather than using a single fixed target as the only acceptable value.

Second: Detecting meaningful changes.

A team may experience an abrupt shift following a product launch, organizational restructuring, or significant change in business demand.

Change-point analysis could help identify when the underlying behavior has changed.

The normal operating range may then need to be reassessed.

Third: Recognizing developing trends.

Not all important changes happen suddenly.

Incoming work, cycle time, or backlog may increase gradually while individual measurements remain within their historical ranges.

Trend analysis can help distinguish persistent movement from short-term fluctuations.

Fourth: Forecasting future constraints.

When a persistent trend is identified, forecasting methods may estimate when an operational threshold will be approached or crossed.

This is an extension of the time-to-threshold forecasting approach I have used in technical Capacity Management.

Together, these capabilities could support a more proactive form of workload analysis.

Importantly, none of these techniques individually constitutes a new invention in organizational management.

The proposed contribution lies in their integration through ANF and SETDS, and in evaluating whether that integration improves management decisions.

3.1. A Practical Illustration: IT-Control Chart of a Team's Weekly Workload

To illustrate this idea, I created a hypothetical example using synthetic workload observations for a software engineering team.

The example covers 60 weeks: 52 observed weeks and an eight-week forecasting horizon.

Rather than using completed story points as a proxy for workload, I use estimated effort-hours associated with incoming work.

This distinction is important.

Completed story points describe delivery estimates, not necessarily incoming demand, actual employee working hours, or sustainable team capacity.

Estimated effort-hours are also imperfect measurements, but they allow us to illustrate how changing demand might be compared with a separately estimated processing-capacity limit.

Figure 1. ANF-based IT-Control Chart of a hypothetical software engineering team's weekly incoming workload. The chart shows observed demand (black), retrospectively estimated statistical ANF boundaries (P05 in blue, P50 in green, P95 in red), a shaded normal-functioning range, a simulated change in operating conditions, and a linear forecast. The orange dashed line represents an independently assumed processing-capacity limit of 185 estimated effort-hours per week. All observations are synthetic.

Figure 1. ANF-based IT-Control Chart of a hypothetical software engineering team's weekly incoming workload. The chart shows observed demand (black), retrospectively estimated statistical ANF boundaries (P05 in blue, P50 in green, P95 in red), a shaded normal-functioning range, a simulated change in operating conditions, and a linear forecast. The orange dashed line represents an independently assumed processing-capacity limit of 185 estimated effort-hours per week. All observations are synthetic.

The chart illustrates three periods.

Period 1: Normal Functioning (Weeks 1–20)

During the initial period, incoming workload fluctuates around a relatively stable level.

The retrospective statistical baseline is:

• Lower ANF boundary (P05): approximately 127.5 effort-hours per week.

• Median (P50): approximately 133.8 effort-hours per week.

• Upper ANF boundary (P95): approximately 140.4 effort-hours per week.

Most observations remain within the estimated normal-functioning range.

This variation does not necessarily require corrective intervention.

The team can continue operating autonomously while managers maintain regular communication, support, and oversight.

Period 2: A New Operating Regime (Weeks 21–40)

Beginning around Week 21, the synthetic example introduces an increase in incoming demand.

This represents a hypothetical business event, such as a product launch or the acquisition of additional customers.

The team begins operating at a higher workload level.

Using a sample from the new regime (Weeks 24–39), the retrospectively estimated ANF becomes:

• P05: approximately 143.4 effort-hours per week.

• P50: approximately 150.7 effort-hours per week.

• P95: approximately 160.2 effort-hours per week.

This illustrates an important feature of dynamic systems: their normal operating conditions can change.

However, we must distinguish a statistically established new baseline from a healthy or sustainable one.

A manager should investigate whether the team has adapted successfully through automation, process improvements, or additional resources—or whether it is absorbing the extra demand through excessive effort.

The statistical change does not answer that question by itself.

Period 3: Increasing Demand and Capacity Forecasting (Weeks 41–60)

After Week 40, the synthetic workload begins increasing more rapidly.

Several observations exceed the previously estimated upper ANF boundary.

The repeated deviations suggest that the earlier operating regime may no longer adequately describe current conditions.

For illustration, I fitted a simple linear trend to the observations from Weeks 41–52.

The resulting trend increases by approximately 1.84 estimated effort-hours per week.

If that trend continues without intervention, the forecast reaches the assumed processing-capacity threshold of 185 effort-hours per week around Week 57.

This is a conceptual example of Time-to-Threshold Forecasting applied to an organizational system.

Instead of waiting until backlogs accumulate or delivery deadlines are missed, management could investigate the trend and consider adjusting priorities, improving processes, or increasing available resources.

However, several methodological limitations must be emphasized.

First, the change in demand was deliberately introduced into the synthetic dataset. It was not independently discovered by SETDS.

Second, the ANF boundaries are empirical percentiles calculated retrospectively from selected operating periods. They are not classical Shewhart control limits, nor do individual percentile exceedances automatically establish statistical significance.

Third, the forecast is a simple linear regression extrapolation. It does not include uncertainty intervals, future changes in demand, or the effects of management intervention.

Finally, the 185-hour threshold is an assumed operational planning limit. It is not a medically or psychologically defined burnout threshold.

What Does This Example Demonstrate?

Despite its simplified nature, the example illustrates three different analytical questions:

Normal variation: Is the observed workload consistent with an established operating regime?

Structural change: Has the behavior of the system changed enough to justify reassessing its statistical baseline?

Future constraints: Does the developing trend suggest that operational demand may exceed available processing capacity?

The example also highlights why analyzing incoming workload alone is insufficient.

A practical organizational application would need to examine incoming demand together with completed throughput, backlog, cycle time, available resources, and relevant quality indicators.

The real opportunity for ANF-based management is to integrate these perspectives, not to replace them with a single workload chart.

4. Connecting Business Demand to Team Capacity

Another possible extension comes directly from my earlier work in IT Capacity Management.

In technical systems, resource requirements often depend on measurable business drivers.

For example, we may model how transaction volume affects CPU utilization, memory consumption, or response time.

The objective is to establish empirical relationships between business activity and technical resource demand.

A similar approach could be investigated for corporate teams.

Consider a customer onboarding department.

Its workload may depend on several factors:

• Number of new customers.

• Complexity of onboarding requirements.

• Number of products or services involved.

• Amount of manual processing.

• Availability of automation.

• Experience and skills of the team.

Suppose historical observations suggest that onboarding ten additional customers is associated with approximately 15 extra hours of team effort under comparable conditions.

Such a relationship could help estimate future resource requirements.

However, organizational systems are more complex than technical infrastructure.

The relationship may change when new software is introduced, employees gain experience, processes are redesigned, or customer requirements become more complicated.

Therefore, a useful model should not assume that historical relationships remain constant indefinitely.

This is where change-point detection and dynamic model reassessment could become particularly valuable.

Instead of relying on a permanently fixed relationship between business demand and team effort, we could investigate how that relationship evolves.

I see this as a possible continuation of my earlier exception-based capacity modeling work.

The objective would not be to mechanically determine how many employees a company must hire.

It would be to provide quantitative evidence to support resource planning and managerial judgment.

5. The Human Difference: Statistical Normality Is Not Sustainability

At this point, an essential distinction must be made.

Corporate teams are not server clusters.

Although statistical methods may be applied to observable organizational behavior, people cannot be managed as interchangeable technical resources.

Human performance depends on motivation, skills, fatigue, interpersonal relationships, organizational culture, and many other factors.

A team might maintain high output because its members have become more experienced and efficient.

The same output might also be sustained through excessive overtime and growing exhaustion.

Operational statistics alone may not distinguish these situations.

An additional problem arises when historical behavior itself is unhealthy.

Imagine a department whose employees have consistently worked excessive hours for several years.

A statistical model might identify this workload as normal because it has become the established operating pattern.

But statistically normal does not necessarily mean healthy, desirable, or sustainable.

This distinction is fundamental to extending ANF into human systems.

I would therefore separate two types of boundaries.

Statistical ANF boundaries describe the range of observed behavior under a particular operating regime.

Organizational sustainability boundaries represent conditions considered acceptable based on available resources, operational goals, working conditions, employee feedback, and relevant research.

These boundaries are not necessarily identical.

This also connects ANF to existing research in organizational psychology.

For example, Job Demands–Resources (JD-R) theory examines relationships among workplace demands, available resources, burnout, and engagement.

Similarly, the SPACE framework for developer productivity emphasizes that productivity cannot be represented adequately by a single metric. It includes dimensions such as satisfaction, well-being, performance, communication, and efficiency.

These frameworks provide perspectives that a purely statistical ANF model would not capture on its own.

A meaningful organizational application should therefore complement established human-centered research.

It should also prioritize team-level measurements where practical, protect employee privacy, and avoid turning exception detection into individual surveillance.

6. A Different Role for the Manager

What would this mean for a corporate manager?

I do not believe the goal should be automated management in which statistical models decide when human intervention is permitted.

A manager has responsibilities that cannot be reduced to quantitative monitoring.

These include coaching, resolving conflicts, supporting professional development, understanding employee concerns, and creating an effective working environment.

However, statistical analysis could help managers devote less attention to routine fluctuations and more attention to meaningful issues.

Three activities become particularly important.

Understanding Exceptions

When a significant deviation occurs, the first response should be investigation rather than blame.

Is the change caused by business demand, staffing, task complexity, process inefficiency, or an improvement in productivity?

An exception should initiate a question, not automatically trigger a corrective action.

Anticipating Constraints

Trend analysis and forecasting could help identify emerging problems before they become operational crises.

If incoming demand consistently grows faster than throughput, a manager may have time to adjust priorities or request additional resources.

Supporting Adaptation

When structural changes occur, teams may need to adapt.

Management should help determine whether the team can establish a sustainable new operating regime or whether the organization must change its expectations, resources, or processes.

In this interpretation, Exception-Based Management is not about reducing leadership to statistical alerts.

It is about using statistical evidence to support better judgment while preserving team autonomy.

7. Toward an Integrated Organizational ANF Framework

The organizational extension of ANF remains a research proposal.

My previous work provides a technical foundation in statistical exception detection, change-point analysis, trend recognition, and capacity forecasting.

The next challenge is to investigate whether combining these methods adds practical value to existing organizational management approaches.

A possible framework could contain five stages.

Stage 1 — Select meaningful operational measurements.

Use existing team-level information such as incoming demand, throughput, backlog, cycle time, and available processing capacity.

Avoid treating a single productivity metric as a complete description of team performance.

Stage 2 — Establish a statistical ANF.

Estimate expected operating behavior from historical observations, taking into account staffing changes, work complexity, seasonality, and other relevant conditions.

Stage 3 — Detect exceptions and structural changes.

Apply statistical methods to distinguish ordinary variation from potentially significant changes in system behavior.

Investigate whether a detected change represents improvement, deterioration, or adaptation to new conditions.

Stage 4 — Forecast operational constraints.

Analyze developing trends and estimate when relevant planning limits might be approached.

Keep these limits distinct from the statistical ANF boundaries.

Stage 5 — Support decisions and evaluate outcomes.

Combine statistical evidence with managerial experience, employee feedback, and organizational context.

Measure whether the approach improves the quality or timeliness of decisions.

The final stage is particularly important.

It would not be sufficient to demonstrate that an ANF model produces visually interesting charts.

A meaningful study should compare its results with existing methods.

For example, we could evaluate conventional fixed thresholds, established SPC techniques, Kanban-based flow analysis, and an integrated ANF/SETDS implementation using identical datasets.

Potential evaluation criteria could include detection delay, false-alert frequency, forecast accuracy, and the usefulness of the resulting information for management decisions.

Only such comparisons could establish whether the proposed integration offers measurable advantages.

8. What Might Be the Actual Contribution of ANF?

After examining existing management practices, I would formulate the research opportunity more carefully than I initially did.

Management by exception is not new.

Statistical control charts are not new.

Agile autonomy, Kanban flow metrics, and workload forecasting are not new either.

However, these concepts are frequently implemented in different contexts and for different purposes.

The possibility I would like to investigate is whether an ANF-based analytical layer can bring together:

1. Statistical estimation of normal operating behavior.

2. Detection of meaningful exceptions and structural changes.

3. Recognition of evolving workload trends.

4. Modeling relationships between business demand and operational capacity.

5. Forecasting the time remaining before relevant constraints may be reached.

The specific contribution would depend on the methods implemented and their demonstrated performance compared with existing alternatives.

The central hypothesis is that integrating these analytical capabilities could improve the recognition of meaningful workload changes and provide earlier, more actionable information for management decisions.

This remains a research hypothesis, not an empirically validated conclusion.

The broader significance may lie in exploring how methods developed for technical performance management can be adapted to more complex organizational systems without ignoring their human characteristics.

Conclusion: ANF as a Complement to Existing Management Practices

My work on the Area of Normal Functioning began long before its recent applications to IT performance analysis.

The original concept emerged from my research into robotic grasping and passive adaptation.

Later, my work on SETDS and IT-Control Charts provided statistical approaches for examining normal behavior, detecting exceptions, identifying structural changes, and analyzing trends in technical systems.

Today, I am exploring whether this accumulated experience can contribute to understanding human and organizational systems.

Reviewing existing management practices shows that many of the underlying ideas already have well-established foundations.

Rather than diminishing the relevance of ANF, I believe this provides a more meaningful direction for further work.

ANF does not need to replace Agile, Kanban, PRINCE2, or Statistical Process Control.

It may instead provide a complementary analytical perspective, particularly where operating conditions change over time and future capacity constraints are important.

The objective is not to turn people into measurable components of a mechanical system.

It is to help leaders distinguish routine variation from meaningful change, recognize emerging constraints, and support organizational adaptation.

The principle remains consistent with my earlier technical work:

Do not react to every fluctuation. Understand what is normal, recognize when conditions meaningfully change, and anticipate future constraints before they become serious problems.

Whether an integrated ANF approach can improve corporate workload management remains an open research question.

But it is one I believe is worth investigating.

---

References and Further Reading

1. Nolan, T. W., & Provost, L. P. (1990). Understanding Variation. Quality Progress, 23(5), 70–78. American Society for Quality.

2. Schwaber, K., & Sutherland, J. (2020). The Scrum Guide. Official Scrum Guide.

3. Kanban Guides. (2025). The Kanban Guide. Official Guide.

4. Atlassian. View and Understand the Control Chart. Jira Software Documentation.

5. PRINCE2. Understanding Tolerances in PRINCE2 Project Management. PRINCE2 Resource.

6. Scrum.org. (2024). Evidence-Based Management Guide. Official Guide.

7. Forsgren, N., Storey, M.-A., Maddila, C., Zimmermann, T., Houck, B., & Butler, J. (2021). The SPACE of Developer Productivity: There's More to It Than You Think. ACM Queue, 19(1). DOI: 10.1145/3454122.3454124.

8. Bakker, A. B., Demerouti, E., & Sanz-Vergel, A. (2023). Job Demands–Resources Theory: Ten Years Later. Annual Review of Organizational Psychology and Organizational Behavior, 10, 25–53. DOI: 10.1146/annurev-orgpsych-120920-053933.

---

This article is part of my ongoing research and book project on the Area of Normal Functioning (ANF), exploring its applications across technical, AI, human, and hybrid systems.