Quality
and the Cobra Effect: When Your Organization’s Incentives Create the
Exact Behavior They Were Designed to Eliminate
The Dead
Cobras That Were Worth More Than Live Ones
During British colonial rule in India, the government grew concerned
about the number of venomous cobras in Delhi. Their solution seemed
rational: offer a cash bounty for every dead cobra brought to an
official. And at first, it worked. People killed cobras, collected the
reward, and the cobra population declined.
Then something unexpected happened. Enterprising citizens began
breeding cobras specifically to kill and sell for the bounty. When the
government discovered this and canceled the program, the breeders
released their now-worthless snakes into the streets. Delhi ended up
with more cobras than before the intervention.
The British had created a textbook case of perverse incentives — a
phenomenon now known as the Cobra Effect. And if you think this is just
an amusing historical footnote with no relevance to your quality system,
I invite you to look more closely at your own organization’s KPIs.
Because the dead cobras on your shop floor have different names.
They’re called “rework hours reduced by not reporting defects.” They’re
called “audit findings that magically dropped the month before
recertification.” They’re called “on-time delivery achieved by shipping
parts that should have been held.”
And just like colonial Delhi, your organization probably has more of
them than anyone is willing to admit.
How Perverse
Incentives Infect Quality Systems
The Cobra Effect occurs whenever a well-intentioned incentive
produces the opposite of its intended result. In quality management,
this happens with alarming regularity because most organizations
construct their measurement systems around a dangerous assumption: that
people will respond to metrics the way management intends them to.
They won’t. They’ll respond to metrics the way the metrics are
structured.
Consider the manufacturing plant that tied its quality inspectors’
bonuses to the defect detection rate — the percentage of defects caught
before shipment. The logic seemed sound: reward inspectors for catching
more defects, and they’ll catch more defects. What actually happened was
that inspectors began classifying borderline conditions as defects
during incoming inspection so they could “catch” them, artificially
inflating their detection rates while flooding the corrective action
system with trivial findings. The real defects — the subtle, dangerous
ones — received less attention because they were harder to identify and
didn’t improve the metric as efficiently.
The inspectors weren’t malicious. They weren’t incompetent. They were
responding rationally to the incentive structure management had created.
And that structure was breeding cobras.
The
Quality Manager’s Taxonomy of Perverse Incentives
Over twenty-five years of auditing and consulting across automotive,
aerospace, and pharmaceutical manufacturing, I’ve catalogued the most
common ways quality incentive systems backfire. Here are the patterns I
see repeatedly:
The Target-Setting Cobra
An automotive supplier sets a target of zero customer complaints per
quarter. The plant manager’s bonus depends on it. When a customer calls
to report a concern, the customer service representative is coached to
classify it as an “inquiry” rather than a “complaint.” The complaint
count stays at zero. The manager gets the bonus. The customer’s
unresolved issue festers until it becomes a full-blown warranty claim
costing ten times what early intervention would have.
The target didn’t eliminate customer complaints. It eliminated the
reporting of customer complaints. These are not the same thing.
The Efficiency Cobra
A pharmaceutical company implements a lean initiative to reduce cycle
time in its quality control laboratory. Test turnaround time becomes the
primary performance metric. The lab responds by batching samples less
frequently, testing the easy ones first, and pushing complex or
ambiguous results to the back of the queue. Average turnaround time
improves. But out-of-specification results — the results that matter
most — now take longer to reach production, increasing the risk that
suspect product gets released.
The organization optimized the metric it measured at the expense of
the outcome it actually needed.
The Cost-of-Quality Cobra
A medical device manufacturer launches a cost reduction program
targeting the cost of quality. The quality department is tasked with
reducing inspection hours, scrap costs, and rework expenses. They
respond by reducing inspection frequency, accepting borderline parts
rather than rejecting them, and reclassifying scrap as “rework” to move
it to a different cost center. The cost of quality metric improves. The
actual cost — warranty claims, customer returns, the quiet erosion of
brand trust — increases invisibly.
They didn’t reduce the cost of quality. They reduced the cost of
measuring quality. The real costs simply migrated to a different
ledger.
The Audit Cobra
An organization ties management bonuses to the number of
nonconformities found during internal audits. The logic: more findings
mean a more rigorous audit program. Auditors respond by documenting
minor administrative issues — missing signatures, outdated forms, slight
documentation formatting problems — while overlooking systemic process
failures that would be harder to document and more uncomfortable to
report. The finding count increases. The audit program’s actual
effectiveness decreases.
They were breeding cobras in the audit schedule.
Why Smart
Organizations Keep Building Cobra Farms
If perverse incentives are so well-documented and their consequences
so predictable, why do intelligent, well-meaning organizations keep
creating them?
The answer lies in what psychologists call the “intentionality bias”
— our tendency to assume that people will interpret our intentions
correctly and act accordingly. When a manager sets a quality target,
they assume everyone understands the spirit behind the number. They
don’t. People understand the number. They optimize for the number. The
spirit is irrelevant to the metric.
There are also structural reasons:
Measurement fixation. Organizations measure what’s
easy to measure, not what matters most. It’s easy to count customer
complaints. It’s hard to measure customer frustration that hasn’t yet
been articulated. So organizations optimize the measurable and ignore
the meaningful.
Time horizon misalignment. Most incentive systems
reward short-term performance. Quality improvements — and quality
failures — often manifest over long time horizons. The manager who
suppresses defect reporting this quarter gets a bonus. The recall that
happens eighteen months from now is someone else’s problem.
Siloed metrics. When each department is measured on
its own KPIs without cross-functional accountability, each department
optimizes locally. Quality reduces inspections to cut costs. Production
speeds up lines to hit delivery targets. Purchasing buys cheaper
materials to reduce unit cost. Each silo hits its numbers. The customer
receives a worse product.
The reviewability trap. Organizations prefer metrics
that are objective, auditable, and defensible. These tend to be
quantitative and narrow. The richer, more nuanced assessments of quality
— customer relationships, process robustness, organizational learning —
are harder to reduce to a number, so they get measured less and
therefore managed less.
The Anatomy of a
Cobra-Proof Quality Incentive
Preventing the Cobra Effect doesn’t mean abandoning metrics or
targets. It means designing incentive systems with the same rigor you’d
apply to designing a process — anticipating failure modes, understanding
variation, and building in checks.
Here’s what cobra-proof incentive design looks like in practice:
Balance competing metrics. Never incentivize a
single metric in isolation. If you’re measuring defect detection rate,
also measure false positive rates and time-to-resolution. If you’re
measuring on-time delivery, also measure first-pass yield and customer
return rates. The tension between competing metrics prevents any single
metric from being gamed.
Measure outcomes, not activities. Count customer
satisfaction scores, not customer complaint counts. Measure product
performance in the field, not inspection pass rates at final audit.
Track warranty costs over twelve months, not scrap rates this week.
Outcomes are harder to manipulate than process metrics because they
reflect reality rather than reporting.
Separate measurement from accountability. The people
who collect quality data should not be the same people whose performance
is judged by that data. This is basic segregation of duties, and it’s
remarkable how often organizations violate it in their quality
measurement systems. When the plant manager is responsible for both
hitting quality targets and reporting quality results, the results will
always look better than reality.
Build in second-order measurement. For every metric
you incentivize, ask: what behavior would this metric reward if someone
wanted to improve it without actually improving quality? Then measure
that behavior directly. If you’re rewarding reduced audit findings, also
track customer complaints and field failures. If the audit findings drop
but field failures increase, you have a cobra problem.
Use ranges, not targets. Instead of “reduce customer
complaints to zero” — a target that incentivizes suppression — use
“maintain customer complaints within a statistically expected range and
demonstrate systematic improvement over time.” This acknowledges
variation and rewards genuine improvement rather than gaming.
Include qualitative assessment. Not everything worth
measuring can be counted. Supplement quantitative metrics with
qualitative reviews: management discussions, customer relationship
assessments, process walk-throughs, and cultural evaluations. These are
harder to game precisely because they resist reduction to a single
number.
A Real-World Cobra Hunt
I once worked with an aerospace components manufacturer that had
implemented a sophisticated balanced scorecard for its quality
department. One of the key metrics was CAPA closure rate — the
percentage of corrective and preventive actions closed on time. The
target was 95%.
For two consecutive years, the quality department hit this target.
The VP of Quality presented the results at every management review with
understandable pride. During my assessment, however, I noticed something
odd: the CAPA closure rate was excellent, but the recurrence rate for
similar nonconformances was climbing. The same types of problems kept
appearing in different areas.
When I dug into the CAPA records, the pattern became clear.
Investigators were defining root causes narrowly — identifying proximate
technical causes rather than systemic ones — because narrow root causes
led to narrow corrective actions, which were faster and easier to close.
A broad systemic investigation might take three months and involve
multiple departments. A narrow technical fix could be closed in two
weeks.
The CAPA closure rate metric was excellent. The actual corrective
action effectiveness was poor. The organization was breeding cobras in
its corrective action system and calling them closed.
The fix was straightforward: we added a secondary metric measuring
corrective action effectiveness — specifically, the recurrence rate of
similar nonconformances within twelve months of CAPA closure. We also
implemented a sampling program where a cross-functional team reviewed
closed CAPAs for root cause depth and action effectiveness. The closure
rate initially dropped as investigators took on more thorough analyses.
But within six months, the recurrence rate had fallen by 60%, and the
organization was solving problems permanently rather than closing them
quickly.
The Second-Order Test
Here’s a practical tool I use with every organization I consult for:
the Second-Order Test. Before implementing any quality metric or
incentive, ask the team three questions:
- What behavior does this metric intentionally
reward? - What behavior could this metric unintentionally
reward? - How would we detect if the second behavior was
happening?
If you can’t answer the third question convincingly, you’re not ready
to implement the metric.
This is essentially an FMEA for your incentive system — anticipating
failure modes and building in detection controls. And just like process
FMEAs, this exercise reveals risks that aren’t obvious until you
systematically look for them.
I’ve run this exercise with dozens of leadership teams. In every
single session, someone has had the uncomfortable realization that a
current metric is incentivizing the wrong behavior. Every time. The
Cobra Effect isn’t a risk that might happen. Given enough time and
pressure, it’s a certainty.
The Leadership
Responsibility
Ultimately, preventing the Cobra Effect is a leadership
responsibility because it requires resisting the temptation of simple,
clean metrics that tell a positive story. It requires the discipline to
measure what matters rather than what’s flattering. And it requires the
courage to accept that real quality improvement is messy, incremental,
and resistant to the kind of clean quarterly progress reports that
executives prefer.
The organizations with the strongest quality cultures I’ve
encountered share one characteristic: they treat their measurement
systems with the same skepticism they apply to their manufacturing
processes. They assume metrics will drift, incentives will be gamed, and
people will optimize for what’s measured rather than what’s meant. They
build in countermeasures — balanced metrics, cross-functional reviews,
qualitative assessments, second-order checks — not because they don’t
trust their people, but because they understand human behavior.
Trust your people. Design your systems for the behavior they’ll
actually produce, not the behavior you wish they’d produce.
In colonial Delhi, the cobra bounty didn’t fail because the people
were dishonest. It failed because the incentive was poorly designed. The
cobras were always going to be bred. The only question was whether the
system would be designed to anticipate that fact.
Your quality metrics are breeding something right now. The question
is whether you know what it is.
Peter Stasko is a Quality Architect with 25+ years
of experience transforming organizations across automotive, aerospace,
and pharmaceutical industries. He has helped companies on three
continents design quality systems that anticipate human behavior rather
than fight against it, and he still hasn’t found an organization that
isn’t accidentally breeding cobras somewhere in its metrics.
