Quality
and the Jevons Paradox: When Your Organization’s Efficiency Improvements
Increase Its Total Defect Output — and the Faster Process You Were So
Proud Of Became the Machine That Produced More Defects Per Hour Than
Your Old Slow One Ever Did

In 1865, a British economist named William Stanley Jevons made an
observation that annoyed everyone who believed efficiency was the answer
to scarcity. England was burning through coal at an alarming rate, and
the conventional wisdom said: make steam engines more efficient, and
we’ll burn less coal. Jevons looked at the data and said the opposite
would happen. He was right. When James Watt’s improved steam engine made
coal cheaper to use, industries didn’t conserve — they consumed more.
Efficiency didn’t reduce demand. It increased it. The better the engine,
the more coal England burned.

This became known as the Jevons Paradox, and it has haunted every
optimization effort since. Including yours.

Because here’s what happens in quality organizations all over the
world: a team improves a process, reduces cycle time, increases
throughput, celebrates the efficiency gain — and six months later
discovers that total defect output has gone up. Not the defect rate. The
defect count. The process is better, faster, and leaner, and it
is producing more defective units per shift than the old “inefficient”
process ever did.

Nobody saw it coming. Everyone was looking at the wrong metric.

The Mechanics of the Trap

Let me paint you a picture. A manufacturing plant produces electronic
control units for automotive applications. The line runs at 400 units
per shift with a defect rate of 2.5%. That means the team deals with
roughly 10 defective units per shift — a manageable number that the
rework cell can absorb.

Management decides to invest in automation. New equipment,
streamlined material flow, optimized changeover procedures. The line now
runs at 650 units per shift. The team celebrates because the defect rate
dropped to 1.8%. Everyone high-fives in the monthly review. Charts are
presented. Bonuses are discussed.

But nobody did the arithmetic.

650 units at 1.8% is 11.7 defects per shift. The defect rate
improved by 28%, but the total number of defective units
increased by 17%. The rework cell, sized for 10 units, is now
drowning. The scrap costs, calculated per unit, look better but the
total scrap bill is higher. The customer, receiving more units total, is
also receiving more defective units. And the quality team, measuring
percentages on dashboards, thinks everything is fine.

This is the Jevons Paradox in quality. Efficiency gains increase the
volume of output enough that even improved rates
produce worse absolute outcomes. The paradox isn’t that quality
got worse per unit. The paradox is that quality improvement, measured
properly, went backward in total.

Why This Keeps Happening

The Jevons Paradox thrives in organizations for three reasons, and
all three are self-inflicted.

First, organizations optimize what they measure, and they
measure rates, not totals. Your dashboard shows defect rate,
scrap rate, first-pass yield, and process capability indices. These are
all proportions. They tell you the quality of a unit, not the quality of
your output. When throughput increases, a stable or even improving rate
masks an increasing absolute defect count. Your KPIs are literally
designed to hide the problem.

I worked with a medical device manufacturer that reduced its defect
rate from 3.2% to 1.9% over eighteen months while simultaneously
tripling production volume. The quality director was promoted based on
the rate improvement. The customer complaint department, meanwhile, saw
a 78% increase in complaint volume. Nobody connected the two because
nobody was looking at total defects reaching customers. The rate said
“improvement.” The reality said “crisis.”

Second, efficiency improvements are almost always evaluated
in isolation. The process engineering team optimizes cycle
time. The quality team tracks defect rates. The production team
maximizes throughput. Nobody owns the intersection — the total number of
defective units the organization produces in a given period. This is a
governance gap, not a technical one. You have the data. You just don’t
have anyone responsible for looking at it this way.

Third, organizations assume that rate improvements will
outpace volume increases. They won’t, because the Jevons
Paradox is structural. When you make a process more efficient, the
economic response is to run more volume. Customers order more. Sales
targets increase. The plant runs extra shifts. The very efficiency that
was supposed to reduce waste creates the conditions for more waste. Not
because anyone is incompetent, but because the system is working exactly
as economic incentives dictate.

The Rebound Effect in
Quality

Economists have a name for what happens when efficiency gains get
consumed by increased usage: the rebound effect. In quality, the rebound
effect shows up in several specific patterns, and recognizing them is
the first step to managing them.

The Throughput Rebound: This is the classic Jevons
scenario I’ve been describing. Faster process → more volume → more total
defects despite lower defect rate. This is the most common and the most
invisible, because your rate-based metrics will never flag it.

The Complexity Rebound: Efficiency improvements
often enable product proliferation. If you can change over faster, you
run more variants. More variants mean more opportunities for error, more
specifications to control, more training gaps. The efficiency gain in
changeover time is consumed by the complexity cost of managing twenty
product configurations instead of five. I watched this happen at a
consumer electronics plant that reduced changeover time from 45 minutes
to 12 using SMED methodology. Within a year, they had added eleven new
product variants because the marketing team realized they could now
fulfill custom orders. Defect rates tripled on the variants because the
quality system, designed for five configurations, couldn’t scale to
sixteen.

The Relaxation Rebound: When a process becomes more
automated or more efficient, operators and supervisors tend to reduce
their vigilance. This is the quality version of the Peltzman Effect —
the safety improvement that makes people behave less safely. If the
machine is “better,” human attention drifts. Inspection becomes cursory.
Anomaly detection degrades. The process efficiency that was supposed to
catch more defects actually causes people to catch fewer.

The Scale Rebound: Efficiency improvements often
lead to capacity expansion. A plant that could serve one market segment
now has the capacity to serve three. New customers, new specifications,
new regulatory requirements, new failure modes. The quality system that
was adequate for one context breaks under the strain of three. The
efficiency gain creates the conditions for its own undoing.

How to Break the Paradox

The Jevons Paradox in quality is not inevitable. It is manageable,
but only if you change how you think about measurement, governance, and
system design. Here is a framework that works.

1. Track Absolute
Defects Alongside Defect Rates

This is the simplest and most powerful change you can make. Every
dashboard, every report, every management review should include both
rate metrics and absolute counts. Track total defects per shift, per
day, per week. Track total customer complaints, not just complaints per
million. Track total scrap cost in dollars, not just scrap as a
percentage of revenue.

When you present both metrics side by side, the paradox becomes
visible. “Our defect rate dropped from 2.5% to 1.8%, and our total
defects increased from 10 to 11.7 per shift” tells a very different
story than “Our defect rate dropped by 28%.” The first statement prompts
investigation. The second prompts celebration. One of those leads to
better quality. The other leads to a promotion and a worse problem.

2. Establish
Volume-Adjusted Quality Targets

Static quality targets are dangerous in a dynamic production
environment. Instead of targeting a fixed defect rate, target a maximum
total defect count adjusted for production volume. If you run 400 units,
your target might be 8 defects maximum. If you run 650 units, your
target might be 10 defects maximum — not 650 × 1.8% = 11.7. The
adjustment should be sub-linear, reflecting the reality that some volume
increase is healthy but quality costs scale with absolute numbers, not
rates.

This creates a natural tension between throughput and quality that
forces collaboration between production and quality teams. The
production team can’t simply crank up volume and point to the stable
defect rate. They have to demonstrate that the increased volume doesn’t
push total defects beyond the threshold.

3.
Size Your Quality Infrastructure for Total Output, Not Rates

Your inspection capacity, your rework cell, your complaint handling
team, your supplier quality engineers — all of these should be scaled to
the absolute number of defects you expect to handle, not the defect
rate. When throughput increases, these resources must scale with it.
This seems obvious, but in practice, efficiency improvements are almost
always accompanied by headcount reductions in quality support functions.
The logic is: “We have fewer defects per unit, so we need fewer quality
people.” The Jevons Paradox says the opposite is true.

I consulted for an aerospace supplier that reduced its defect rate by
40% through a Six Sigma project and immediately eliminated two
inspection positions. Within six months, the remaining inspectors were
overwhelmed by the total defect volume from the increased production
rate, escapes to customers increased, and the company faced a major
corrective action from its OEM customer. The quality team had done its
job. The finance team had undone it.

4. Model the Paradox
Before You Improve

Before any efficiency improvement project, run the numbers. Model the
expected throughput increase, calculate the projected total defect count
at the improved rate, and compare it to the current total defect count.
If the total goes up, the project needs to include quality improvements
that compensate for the volume effect. This is not optional. This is the
difference between improvement and self-deception.

The model doesn’t need to be complex. A simple spreadsheet that asks:
“At the projected new throughput and the projected new defect rate, how
many defective units will we produce per period? Is that more or fewer
than we produce today?” If the answer is “more,” the project plan needs
additional quality countermeasures before it proceeds.

5. Implement
Throughput-Dependent Controls

As volume increases, control systems should tighten, not stay static.
This means implementing adaptive inspection frequencies that increase
with throughput. It means statistical process control limits that
account for the increased sensitivity that higher volume provides. It
means escalation triggers based on absolute defect counts, not just
rates.

A practical implementation: set up control charts that track total
defects per shift alongside traditional X-bar and R charts. When the
total defect count exceeds a threshold — even if the rate is within spec
— trigger a review. This catches the Jevons effect in real time, before
it compounds into a customer-facing problem.

The Strategic Implication

The Jevons Paradox is not just a measurement problem. It is a
strategic one. Organizations that understand this paradox make
fundamentally different decisions about capacity, growth, and quality
investment.

The organizations that get it right treat quality capacity as a
strategic asset that must grow proportionally with production capacity.
They don’t see quality as a cost to be optimized away. They see it as
infrastructure that must scale. They understand that a plant running at
90% capacity with a robust quality system will outperform a plant
running at 120% capacity with a quality system designed for 90% — not
because the quality system is worse, but because the total defect output
is higher.

These organizations also make different decisions about automation.
Instead of automating to increase throughput, they automate to increase
capability — to do things that humans can’t do reliably, like
in-line measurement, real-time SPC, automated defect detection. The
throughput increase is a byproduct, not the goal. And when throughput
does increase, the quality system is already scaled to handle it because
it was designed for capability, not efficiency.

A Personal Observation

In twenty-five years of quality work across automotive, aerospace,
and pharmaceutical manufacturing, I have seen the Jevons Paradox play
out more times than I can count. It is one of the most reliable patterns
in manufacturing quality. And it is one of the most consistently
ignored.

The reason it gets ignored is not stupidity. It is measurement
design. We build dashboards that track rates because rates are
comparable across periods and products. We build incentive systems
around rates because rates are easy to target. We build organizational
structures around rates because rate improvement fits neatly into
project portfolios and performance reviews.

But quality is not a rate. Quality is an outcome. And outcomes are
measured in absolutes: how many defective products reached customers,
how many patients were affected, how many vehicles were recalled, how
many lives were impacted by the work we did or failed to do.

The Jevons Paradox reminds us that the path to better quality
outcomes sometimes requires slower, not faster. That efficiency without
proportionality is not improvement — it is acceleration toward the wrong
destination. And that the most dangerous moment in any quality
improvement journey is not when things go wrong. It is when the metrics
tell you everything is getting better while the reality is getting
worse.

Watch your totals. Not just your rates. The paradox is patient, and
it is always waiting.

Peter Stasko is a Quality Architect with 25+ years
of experience transforming organizations across automotive, aerospace,
and pharmaceutical industries. He specializes in helping companies see
what their dashboards are hiding — and fix what their metrics are
missing.