Quality
and the Cobra Effect: When Your Incentives Designed to Fix Quality
Actually Create the Defects They Were Supposed to Prevent
A
Story About How the Perfect Bonus Plan Built the Perfect Defect
Factory
In colonial Delhi, the British government grew concerned about the
population of venomous cobras. Their solution was elegant in its
simplicity: offer a cash bounty for every dead cobra. Problem meets
incentive. Incentive meets action. Action meets result.
At first, it worked. Cobra carcasses piled up. The administrators
congratulated themselves on a policy triumph. Then something unexpected
happened. Enterprising citizens began breeding cobras in their homes —
killing them at a steady clip and collecting the bounty. When the
government discovered this and cancelled the program, the breeders
released their now-worthless snakes into the streets. Delhi ended up
with more cobras than before the bounty.
The Cobra Effect — when an incentive produces the exact opposite of
its intended result — is not a colonial curiosity. It is alive and well
on your shop floor, hiding inside your bonus structures, your KPI
dashboards, and your performance review criteria. And it is quietly
manufacturing defects that your quality system was designed to
eliminate.
The Anatomy of a Perverse
Incentive
Every organization designs incentives. Some are formal — bonuses,
promotions, scorecards. Others are informal — praise, attention, the
simple relief of being left alone. Both types send messages. And the
message that lands is rarely the one you intended.
Consider the manufacturing manager whose bonus is tied to first-pass
yield. The metric is clear: parts that pass inspection without rework
divided by total parts produced. The intent is equally clear: encourage
processes that produce good parts the first time.
Here is what actually happens.
Operators discover that parts with borderline defects can be nudged
past inspection with a bit of creative interpretation. Inspectors,
feeling the collective pressure of the team’s bonus, begin applying the
most generous reading of the specification. The borderline becomes
acceptable. The acceptable becomes standard. First-pass yield climbs to
98.7%. The bonus is paid. Everyone celebrates.
Meanwhile, your customer begins returning parts at three times the
previous rate. Because the defect didn’t disappear — it migrated. It
left your factory dressed as a conforming part and arrived at your
customer’s dock wearing its true colors.
The incentive worked. The metric improved. The quality degraded. This
is the Cobra Effect in manufacturing.
The Gallery of Good
Intentions
You do not have to look hard to find these dynamics. They are
everywhere, and they follow predictable patterns.
The Scrap Target
That Increased Total Waste
A plant manager, frustrated by high scrap rates, set a hard cap: no
line could scrap more than 2% of its output. The intent was to force
discipline. The result was more creative.
Operators began hiding scrap in rework. Rework counts didn’t count
against the scrap metric, so parts that should have been discarded were
instead reprocessed — often poorly — and sent downstream. The rework
station became a bottleneck. Throughput dropped. Energy consumption per
good part increased. And the total waste, when you counted rework
materials, labor, energy, and the downstream defects from poorly
reworked parts, actually exceeded the original scrap rate.
The scrap metric looked beautiful. The reality was ugly. The cobras
were breeding in the rework cell.
The OEE Race That Broke
Maintenance
A site linked shift bonuses to OEE — Overall Equipment Effectiveness.
Suddenly, every minute of downtime was money out of operators’ pockets.
Predictive maintenance windows were postponed. “It can run another
shift” became the rallying cry. Operators began performing minor repairs
themselves rather than calling maintenance and eating into the
availability metric.
OEE numbers soared past 90%. The dashboard was magnificent.
Then, in a single week, three critical machines failed
catastrophically. The kind of failure that takes weeks to repair, not
hours. Total OEE for the month collapsed to 61%. The maintenance team,
when asked what happened, could only point to months of deferred
preventive work orders — every one of them signed off with the notation
“operator request — continue running.”
The incentive to maximize uptime had systematically dismantled the
system that made uptime possible.
The Zero Defect
Month That Buried Problems
An automotive supplier launched an ambitious campaign: the first
production line to achieve a full calendar month with zero recorded
defects would receive a team bonus and a company-wide celebration.
Line 7 won. Champagne corks popped. Photos were taken.
Six weeks later, an auditor discovered that Line 7’s shift leads had
been quietly reclassifying defects as “rework opportunities” and
repairing them before they entered the tracking system. The defects were
real. The records said otherwise. The team had not eliminated defects —
they had eliminated the data about defects.
The celebration photos stayed on the wall for months. Nobody had the
heart to take them down.
Why Smart People Build
Broken Incentives
The Cobra Effect is not a symptom of stupidity. It is a symptom of
single-variable thinking — the belief that you can optimize a complex
system by measuring and rewarding one dimension of it.
Quality is multidimensional. It lives at the intersection of speed,
cost, safety, morale, customer satisfaction, long-term reliability, and
institutional learning. When you pull on one thread, the entire fabric
shifts. Most incentive systems ignore this reality. They select one
metric, amplify it, attach money to it, and hope for the best.
This fails for three structural reasons.
First, people optimize for the metric, not the
intent. This is not cynicism — it is rational behavior. If you
are measured on first-pass yield, you will maximize first-pass yield. If
that means shipping borderline parts, you will ship borderline parts.
The metric is the message, and the message is all that matters when rent
is due.
Second, metrics are abstractions, and abstractions
leak. Every metric is a proxy for reality, and every proxy
leaves something out. First-pass yield leaves out the severity of the
defects that do escape. OEE leaves out the long-term health of the
equipment. Scrap rate leaves out rework. When you incentivize the proxy,
you incentivize the gap between the proxy and reality.
Third, the Cobra Effect is self-reinforcing. Once
perverse incentives take hold, they create constituencies. The people
who benefit from the current system — and that includes everyone whose
bonus depends on it — will defend the system. They will resist changes
to the metric. They will explain away anomalies. They will become the
cobras.
The
Systemic View: How to Design Incentives That Don’t Breed Cobras
The solution is not to abandon incentives. People respond to rewards
— that is a feature, not a bug. The solution is to design incentives
that are aligned with the full picture of quality, not a single
pixel.
Balance Competing Metrics
Never tie significant rewards to a single metric. Instead, build
balanced scorecards that include both leading and lagging indicators,
both efficiency and effectiveness measures, both short-term results and
long-term health.
A plant bonus might include first-pass yield alongside customer
complaint rate, maintenance compliance, safety incidents, and training
hours completed. No single metric can dominate. No single optimization
can game the system.
Include the Customer’s Voice
The most powerful antidote to perverse incentives is the customer’s
unfiltered feedback. Build customer complaints, returns, warranty
claims, and satisfaction surveys directly into performance metrics. It
is much harder to hide a defect from the customer than from your
internal tracking system.
When a team knows that the parts they ship will be evaluated by the
end user — and that evaluation will affect their bonus — the incentive
to game internal metrics diminishes dramatically.
Audit the Metrics, Not
Just the Numbers
Every metric is a system. Every system can be gamed. The question is
not whether gaming is happening — it is whether you are looking for
it.
Build regular metric audits into your quality system. These are not
audits of the numbers themselves but of the behaviors the numbers are
driving. Walk the floor. Watch how operators respond to borderline
conditions. Ask inspectors what happens when they find defects. Listen
for the phrase “that doesn’t count against us” — it is the sound of a
cobra being bred.
Design for the Second Order
Before implementing any incentive, run a premortem. Gather the team
and ask: “If we were determined to game this metric — to make the number
look good while actually making quality worse — how would we do it?”
The answers will be uncomfortable. They will also be predictive. The
people who will game the system are often the same people who can most
clearly see how it could be gamed. Use that knowledge to close the
loopholes before they open.
Make It Safe to Report Bad
News
Many perverse incentives are born from fear. When honest reporting
leads to punishment — lost bonuses, public criticism, career damage —
people will find ways to avoid honest reporting. Thecobra Effect often
begins not with greed but with self-preservation.
The most powerful incentive you can design is psychological safety:
the confidence that reporting a problem will lead to help, not harm.
When people believe that the system wants the truth, they are less
likely to manufacture lies.
The Leadership Trap
There is a particular danger for leaders reading this. The trap is
the comfortable belief that the Cobra Effect is something that happens
in other organizations — less disciplined, less enlightened, less
committed to quality than yours.
It is not.
The Cobra Effect thrives in organizations with strong cultures,
talented people, and sincere commitments to excellence. Why? Because
those organizations are most likely to believe their own metrics. They
have dashboards that look impressive, trends that point upward, and
teams that celebrate wins. The environment of success makes it harder to
see the cobras breeding in the corners.
The leader’s job is not to trust the metrics. It is to trust the
system that produces the metrics — and to continuously verify that the
system is producing truth, not theater.
A Practical Test
Here is a simple test you can run this week. Take your most important
quality metric — the one tied to bonuses, displayed on dashboards,
discussed in every meeting. Then ask three questions:
-
If someone wanted to make this number look better without
actually improving quality, how would they do it? Write down
every method you can think of. Be creative. Be cynical. -
Which of these methods are currently possible in your
system? Not which are happening — which are possible. If the
pathway exists, assume someone has walked it. -
What would you need to change to make those methods
impossible? This is your to-do list. Not the metric itself —
the pathways around it.
If you cannot think of any way to game the metric, you are not
thinking hard enough. Every metric can be gamed. The question is whether
you have made it difficult enough that gaming requires more effort than
genuine improvement.
The Deeper Lesson
The Cobra Effect teaches us something uncomfortable about quality
management. It teaches us that our tools — metrics, incentives,
dashboards, targets — are not neutral. They shape behavior. They shape
culture. They shape reality itself.
Every KPI is a lens. It focuses attention on some things and blurs
others. Every target is a magnet. It pulls behavior toward itself — and
away from everything else. Every bonus is a bet. You are betting that
the behavior the incentive drives will align with the outcome you
actually want.
The history of quality is littered with bets that lost. Plants that
optimized their way into obsolescence. Teams that celebrated their way
into customer defection. Companies that dashboarded their way into
bankruptcy.
The antidote is not complexity. More metrics do not fix the problems
caused by the wrong metrics. The antidote is humility — the recognition
that no single number captures the full picture of quality, and that
every incentive system carries within it the seeds of its own
perversion.
The Cobra Handler’s Checklist
Before you design your next incentive, review this list:
- Have you balanced the metric with at least two competing
measures? If not, stop and redesign. - Have you included the customer’s direct voice?
Returns, complaints, and warranty data are your reality check. - Have you run a premortem with the frontline team?
They know how to game it. Ask them. - Is it safe to fail? If honest reporting hurts,
dishonest reporting will flourish. - Will you audit the behavior, not just the number?
Build metric audits into your calendar — quarterly at minimum. - Are you prepared to change the metric when it stops serving
quality? The best metric today may be the worst metric
tomorrow. Let go when it is time.
The Final Count
Back in colonial Delhi, the cobra bounty was cancelled. The breeders
released their snakes. The problem was worse than before. And somewhere
in the archives of British India, there is probably a report that says
“Initial results promising.”
Your dashboards probably say something similar.
The question is whether you will believe them — or whether you will
walk the floor, talk to your operators, inspect your rework station, and
count the cobras yourself.
The metric is not the territory. The map is not the reality. And the
incentive is not the outcome.
It never was.
Peter Stasko is a Quality Architect with over 25
years of experience transforming manufacturing quality systems across
automotive, industrial, and electronics industries. He specializes in
bridging the gap between theoretical quality frameworks and practical
shop-floor implementation — helping organizations build systems where
good quality is the natural outcome, not a constant struggle against
perverse incentives and unintended consequences.
