During colonial times, India’s British rulers became acutely annoyed by the country’s snake menace. So they hatched what they thought was a bright idea—pay people to catch and kill snakes. The people, in turn, responded vigorously by catching all the snakes they could. But soon enough, the idea ran aground. As snakes became rarer in the wild, a few enterprising everyday Indians had an even brighter idea: They began to farm snakes so they could sell them to the British rulers for a profit.

When a metric becomes a target, it ceases to be a good metric. So goes Goodhart’s law. But it's not just the colonial British who suffered from this ungodly folly. Examples abound. Closer home, there is an urban legend about a project to build toilets. A few decades ago, somewhere in the mid-western hills, a foreign-funded project helped build toilets for people. The project operators knew that people, accustomed as they were to habit, wouldn’t begin using the toilets just yet. So they came up with a solution. They started paying people to inculcate a toilet habit but their solution proved clever-by-half. People relieved themselves in the toilet until they were paid. Once the payments stopped, they returned to their old ways under the open sky. The toilets found better use as a store of firewood.

Systems and programmes follow their design and implementation, not the desire and intent of those who create them. Well-designed systems make the metrics they optimise for and the outcomes they chase fairly congruent with the goals they were designed for. Poorly designed systems display a rift between the metrics they optimise for and the intended goals. Our public policy-scape is replete with these examples. In fact, the Nepali health and social sector is a veritable smorgasbord of ill-conceived and poorly designed programmes and systems.

The example of health insurance in Nepal is poignant. In 2014, the government passed the current National Health Insurance Policy. As designed and implemented, health insurance was primarily optimised to protect people from financial ruin. The primary metric our health insurance system cares about is the number of people that are financially protected from catastrophic health expenditures, and not necessarily the number of people its use makes healthier. Up to a certain annual limit, the insurance system pays for people to access health services. The hope was that this would reduce financial hardship due to accessing health services, and by extension, maybe that would make people healthier.

Turns out, that was too giant a leap of logical faith. Although we have little formal evaluation of such insurance programmes in Nepal—based on experience from other similar countries—it is unlikely that health insurance programmes, implemented as they are, have made people any healthier. Worse, evidence suggests that insurance programmes may even have failed to financially protect people from the ruinous effects of health expenditures.

Appropriately provided health services have the potential to make one healthy, but getting an unnecessary CT scan and some multivitamin tablets just because they have been paid for is not the way to do it. If anything, an insurance system that pays for the consumption of health services primarily benefits the people who sell the services.

But health insurance is not the only financial programme in the health sector, where systems and programmes are designed to chase a metric without much consideration for what that metric achieves. A recent Bagmati province government decision (now scuttled) to buy CT scanners for district hospitals was an example. Merely increasing the number of CT scanners does not even help people get better diagnostic services, let alone make them healthier. Upgrading primary health centres to 15-bed hospitals without consideration for the actual services that people need might only help increase the number of hospital beds. It would, however, neither provide appropriate services to people nor make them healthier.

Another futile metric chase in the health policy scape has been in the implementation of the free drug programme. The free drug programme as implemented by the government is especially egregious since it is designed just to increase the number of beneficiaries who obtain a selected set of drugs, with scant regard to whether those are the drugs that the local community really needs or even whether the drugs are any good. And in the latter measure, the free drugs have been especially problematic. Several companies that supply drugs to the government severely underquote their prices and supply shoddy products. Bidders have been known to supply normal saline bags for less than Rs5 when a bottle of mineral water costs about Rs20. Optimising for the easiest metrics often produces systems that chase futile targets and achieve meaningless and harmful goals.

Then how should public systems be designed? How should public programmes and systems that are designed to solve difficult problems be handled? Is there no method here? There is, and still relies on designing systems around measurable metrics that can be achieved, but they need to be the right metrics. A constellation of metrics—rather than a standalone—that better reflect the goals of the programme are more likely to create a robust system that optimises for the stated goals. While designing complex systems, it is always desirable to make them more resilient by introducing local adaptability, fault tolerance and modularity. Also, it is helpful to remain cognizant that even well designed systems are only as good as the people that run it.

So, when designing a health insurance system, it should not be simply designed to increase the demand for health services, but to help people get the kind of health services they need. That means if someone needs a home health nurse, do not necessarily pay a hospital to provide CT services. If the goal is to increase people’s access to necessary drugs, the local community should be able to buy the kind of medicines they actually need. No bales full of scabies potions while people mostly need blood pressure pills.

The solution to badly designed systems that follow irrelevant metrics aren’t those that follow no metrics at all. Rather, the solution is to design systems that best align their goals with the metrics they measure. Therein lies the way around the rub of Goodhart's law.