The cobra effect in depth: when the fix becomes the problem

The previous article in this series introduced the cobra effect briefly, as an example of what happens when you don't think through how people actually respond to a rule. It deserves its own, deeper treatment – because the pattern shows up far more often than most people assume, and it can be learned to recognize before it strikes.

The story behind the name

The story that gave the phenomenon its name takes place under British colonial rule in Delhi. Authorities wanted to reduce the number of venomous cobras in the city and offered a bounty for every dead snake turned in. At first it worked: the cobra population dropped. But locals soon discovered it was easier to breed cobras for the bounty than to hunt wild ones. A small cobra-breeding industry emerged.

When authorities realized what was happening, they scrapped the bounty. The breeders, now stuck with worthless snakes, did the only rational thing: released them. The result was more cobras in Delhi than before the program even started.

What makes the story worth remembering isn't the snakes. It's the structure: a measure targeted the wrong thing (dead cobras, not the wild cobra population), and that measure could be satisfied in a way that undermined its entire purpose.

Worth noting: the Delhi origin story itself is debated among historians. Some argue it rests on loose accounts rather than verified archives, and that a similar bounty scheme more likely first appeared in the Madras Presidency. That doesn't change the point of the analogy – the term and the pattern it describes are well established – but it's worth knowing before retelling the story as undisputed fact.

More examples of the same pattern

The Delhi story isn't unique. The same structure recurs in several documented cases:

French colonial Vietnam, Hanoi, early 1900s. Authorities paid a bounty per rat tail turned in, as proof of a killed rat, to fight a rat infestation. Residents soon started simply cutting off the tail and releasing the rat, allowing it to keep breeding and generating more tails to collect bounties for. Some are said to have started breeding rats specifically for that purpose.

Soviet factory quotas under the planned economy. Factories given quotas measured in number of nails produced churned out enormous quantities of tiny, practically useless nails to hit the volume target. When the quota was switched to weight instead, factories shifted to producing a small number of enormous, heavy nails – technical target compliance again, without the underlying need (usable nails in a range of sizes) being met.

What all three examples share: the target was a proxy for what was actually wanted, and the proxy could be satisfied more cheaply than the actual goal.

A contemporary example: patient satisfaction in healthcare

You don't need to go back to colonial-era snakes and rats to find the same pattern today. A clear example exists in healthcare.

Many healthcare organizations want to raise patient satisfaction and introduce measurements for it – post-visit surveys, NPS-style scores, follow-up calls. The intent is good: understanding and improving the patient's actual experience. The problem lies in how the measurement gets carried out. Sending out, collecting, and analyzing surveys takes resources, and in many organizations that time is partly drawn from doctors and nurses, who end up spending a growing share of their working hours on administration around the measurement instead of on patients. At the same time, healthcare has seen a marked increase in administrative staff in many places, partly tied to exactly this kind of tracking and reporting – roles that in turn create new processes, documentation requirements, and reporting steps for clinical staff to deal with, not always clearly connected to the patient's actual care.

The result is exactly the pattern the cobra effect describes: measuring patient satisfaction competes for the same limited resource – clinical time – that actually drives patient satisfaction. The more time goes into measuring and reporting, the less time is left for what the surveys are actually meant to capture the experience of: being seen, listened to, and treated without unnecessary waiting. Queues grow, and satisfaction falls – not despite the good intention to measure and improve, but partly because of how that intention was implemented in practice.

It's the same structure as the nail quotas and the rat tails, just in modern dress: a proxy (survey scores, reporting rates) that can be "satisfied" in a way – spending administrative resources on the measurement itself – that erodes the thing the proxy was set up to reflect.

The same pattern shows up in policing. Centralized reporting, especially for more sensitive cases, often means officers in major metro areas have to travel to a central unit to write up reports or process cases that used to be closed out closer to the field. The time spent traveling and on administration is time not spent on the streets, which is the actual safety and visibility citizens are asking for. The structure is the same: an administrative requirement – meant to ensure quality and traceability – competes for the same resource as the police presence it was meant to support.

The conditions that make the cobra effect likely

The cobra effect doesn't emerge at random. In practice it requires three things at once:

1. A measurable target that replaces the actual purpose. Counting dead snakes instead of the wild snake population, closed support tickets instead of customers whose issue was actually resolved, lines of code instead of working software.
2. A cheaper way to satisfy the measure than to satisfy the purpose. There has to be a shortcut – otherwise the rational path and the intended path would coincide.
3. No monitoring of the shortcut. If someone actually checks whether the cobras are wild or bred, or whether the support ticket is genuinely resolved, the incentive to take the shortcut disappears.

That combination is what makes an apparently reasonable rule or bonus model suddenly generate exactly the behavior it was meant to prevent. And the harder the pressure placed on the measure – the more heavily the quota or bonus is weighted – the stronger the incentive to find the shortcut, regardless of how well the measure once correlated with the real purpose.

Seeing it before it happens

The practical value in all of this isn't in the stories themselves, but in the question you can ask before introducing a new rule, KPI, or bonus model: is this a goal in itself, or a proxy for something else – and if it's a proxy, what's the cheapest way to satisfy the proxy without satisfying what it was meant to measure?

That question doesn't solve everything. But it shifts the focus from "does this rule do what we want" to "what does this rule make it possible to win by doing instead," which is a different – and far more useful – question.

What determines whether this kind of effect spreads, amplifies, or dampens within an organization is rarely a single decision in isolation. It's about how different parts of the system – incentives, behaviors, outcomes – feed back into each other over time. That's the subject of the next part in this series.