Why Quality Is Inversely Proportional to Variability

This essay grew out of a recent LinkedIn discussion about a deceptively simple question:

What is quality?

I offered Montgomery’s statement that “quality is inversely proportional to variability.” The responses raised a worthwhile challenge: Is this a complete definition of quality, or does it describe only one important property of it?

That distinction matters. Reducing variability can make a product or service more consistent and predictable, but consistency alone does not guarantee that the result meets customer needs or even targets the right outcome. To understand the role of variability, we must first distinguish quality of design from quality of conformance.

As consumers, we recognize the importance of quality in the products and services we use every day. We may know quality when we experience it, but defining it precisely—and improving it systematically—is more difficult. For manufacturers, service providers, and other business stakeholders, an intuitive understanding is not enough. Quality must be defined in measurable terms. After all, we cannot systematically improve what we cannot measure.

A traditional definition of quality is “fitness for use.” This definition considers whether a product or service satisfies the needs of the people who use it, which is very difficult to quantify from an engineering perspective. Montgomery divides fitness for use into two general aspects: quality of design and quality of conformance (Montgomery, 2013, p. 6).

Quality of Design (Variety in Design)

Quality of design refers to the intentional choices made when defining a product’s features, materials, performance, and intended market. For example, the differences between an iPhone Pro model with three cameras and an iPhone SE with one camera are deliberate design decisions. The two phones represent different combinations of features, performance, and price.

These differences do not necessarily mean that one product has poor quality. Each product may be well designed for a different group of customers. The relevant question is whether the chosen design is fit for its intended use.

Quality of Conformance (Variability in Production and Delivery)

Quality of conformance describes how closely the finished product matches its intended design and specifications. Once the design has been established, the production process must reproduce it consistently.

Suppose a shaft is designed to have a particular diameter. In practice, the diameters of manufactured shafts will not be perfectly identical. Differences will arise because of material properties, machine wear, measurement error, environmental conditions, operator methods, and other factors. These differences among nominally identical outputs constitute variation.

Statistical methods help organizations understand and reduce unwanted variations. The goal is not merely to produce some acceptable units, but to create a process that repeatedly produces units close to the intended target.

Conformance alone, however, is not sufficient. A product may conform perfectly to outdated design or poorly chosen specifications and still fail to meet customer needs. High quality therefore requires both an appropriate design and consistent conformance to that design.

A Modern Definition of Quality

Montgomery offers a concise modern definition:

“Quality is inversely proportional to variability” (2013, p. 6).

In this context, variability means unwanted or harmful variability—not intentional product variety or differences that customers value.

Consider a process that produces 10,000 shafts. If the shaft diameter is a critical quality characteristic, the units should remain close to both the specified target and one another. A process with a wide spread of diameters is more likely to produce parts that do not fit properly during assembly, fail prematurely, or require rework.

Reducing variability provides several benefits:

Products perform more consistently.
Assembly problems and defects become less frequent.
Customers have a more predictable experience.
Performance claims become more credible.
Problems become easier to diagnose and correct.

Low variability does not automatically mean high quality. A process can consistently produce the wrong result. If every shaft has nearly the same diameter but that diameter is different from the design target, the process is precise but inaccurate. Effective quality improvement must therefore address three questions:

Is the design appropriate for the customer’s needs?
Is the process centered on the correct target?
Is unwanted variability sufficiently small?

Eliminating all variation is neither practical nor necessary. The objective is to understand its sources and reduce it to a level that allows the product or service to perform reliably.

Case Study 1: Shoe Manufacturing Plants

Consider two plants producing the same model of shoe for the same company.

The first plant produces shoes with highly variable durability. Some pairs last for years, while others wear out after only a few months. The plant occasionally produces excellent, long-lasting shoes, but its performance is unpredictable.

The second plant produces shoes with much more consistent durability: nearly every pair lasts approximately six months.

Which plant would you prefer to own?

From a customer’s perspective, neither plant necessarily produces a high-quality shoe. If the expected lifetime is two years, the second plant consistently misses the target. Nevertheless, it may provide a better foundation for improvement because its output is predictable. Engineers can focus on shifting durability toward the desired target while attempting to preserve the process’s consistency.

The first plant presents a more difficult problem. Because its output varies so widely, the company cannot make a credible promise about durability. Some customers may receive excellent shoes, while others receive disappointing ones. The company must first identify and control the sources of variation before it can reliably improve average performance.

This example illustrates an important distinction: consistency is not the same as excellence, but excellence cannot be delivered reliably without consistency.

Case Study 2: Software Engineering

In software systems, critical-to-quality (CTQ) characteristics may include response time, data-processing speed, availability, reliability, or prediction accuracy.

Because a line of deterministic code produces the same output when given the same inputs under the same conditions, it may appear that variability is irrelevant to software. In practice, however, software operates within a larger system. Its inputs, users, hardware, network conditions, configurations, and operating environments can all vary.

Consider two online forms that collect information for an algorithm that predicts whether a user needs coaching.

Form 1 asks vague or inadequate questions that users interpret differently. As a result, the form produces missing, inconsistent, or inaccurate data. Even users with similar profiles may provide widely different responses. The same user may also answer differently when completing the form on separate occasions. This input variability can produce unreliable predictions.

Form 2 uses clear questions, appropriate response options, validation rules, and a carefully designed user interface. These features reduce ambiguity and missing data, resulting in more consistent inputs and more reliable predictions.

However, consistency alone is still not enough. A form can consistently collect biased or irrelevant information. Its questions must also measure the factors that genuinely matter to the prediction. Therefore, the quality of the system depends on both the consistency and the validity of the collected data.

Case Study 3: Flight Delays

In service industries, critical-to-quality characteristics are often represented through key performance indicators. Airline departure delay is one example.

Suppose two airlines are both delayed by an average of 13 minutes. For Airline A, individual delays remain close to 13 minutes, with a standard deviation of one minute. For Airline B, the average is also 13 minutes, but individual flights vary widely: some depart early, while others are delayed by 30 minutes or more.

Both airlines need to reduce their average delay. However, Airline A operates more predictably. Its consistent 13-minute delay may point to a recurring scheduling mismatch, a systematic bottleneck, or another persistent feature of the process. Once the cause is understood, the airline can work to shift the average toward zero.

Airline B faces two problems: its average delay and its high variability. Because its performance changes substantially from flight to flight, multiple conditions may be influencing the process. Passengers also have greater difficulty planning around such unpredictable performance.

We cannot conclude from these summary statistics alone that either process is statistically controlled or that Airline A’s problem will be easy to fix. We can conclude, however, that lower variability makes service performance more predictable and usually makes systematic improvement easier.

Conclusion

Low variation in products and services is not accidental. It usually reflects deliberate design, well-understood processes, reliable measurement, appropriate training, and continued monitoring.

Quality is not demonstrated by producing an excellent result occasionally. It is demonstrated by delivering the intended result repeatedly. A high-quality process combines an appropriate customer-focused target with low unwanted variability around that target.

Think Box!

Use the modern definition of quality to compare a beginner with an expert musician, magician, or athlete.

Choose one measurable aspect of performance and imagine that each person repeats the same task many times. Consider the following questions:

How do their average performances differ?
Whose performance varies more from one attempt to another?
Could the beginner occasionally perform as well as the expert?
How does the expert combine a high level of performance with consistency?
Which forms of variation are harmful, and which might represent valuable creativity or adaptability?

Written by Javad Seif

Author’s note: I originally wrote this essay in December 2023 as part of my textbook on quality engineering. I revised it on September 22, 2026.

Reference

Montgomery, D. C. (2013). Introduction to statistical quality control (7th ed.). John Wiley & Sons.