Why You Can't Fix a System by Fixing Its Parts

In the 1960s, the city of Hanoi had a rat problem. The French colonial government, in a burst of bureaucratic ingenuity, offered a bounty for every rat tail turned in. Citizens could bring rat tails to government offices and collect a cash reward per tail. Elegant, efficient, incentive-aligned. The kind of solution that looks brilliant on a whiteboard.

Within weeks, officials noticed something strange. The bounty payments were climbing, but the rat population wasn't shrinking. People were collecting tails — lots of them — but there were just as many rats on the streets as before.

Then they discovered what was actually happening. Enterprising residents had started cutting the tails off live rats and releasing them back into the sewers. Tailless rats bred more rats, which grew more tails, which meant more bounties. Some citizens had gone further — they were breeding rats in their homes. The bounty program hadn't reduced the rat population. It had created a rat farming industry.

The government was trying to fix a system by intervening in one of its parts. The system responded the way systems always do: by rearranging itself around the intervention until the intervention became the problem.

The Most Expensive Mistake in Problem-Solving

There's a way of thinking about problems that feels so natural, so self-evidently correct, that most people never question it. It goes like this: find the broken part, fix the broken part, problem solved.

Your car won't start? Replace the battery. Website is slow? Upgrade the server. Employee underperforming? Send them to training. Revenue is down? Cut costs. Each of these is a linear intervention — identify the faulty component, repair or replace it, expect the system to improve.

This works beautifully for machines. A car is a complicated system — lots of parts, but each part has a predictable relationship with every other part. Swap the alternator and the electrical system behaves exactly as the manual says it will.

But most of the systems that actually matter in your life — your health, your team, your company, the economy, your relationships — are not complicated. They're complex. And complex systems play by fundamentally different rules.

Complicated vs. Complex: The Distinction That Changes Everything

A Boeing 747 is complicated. It has six million parts. But every part is knowable, every interaction is predictable, and if you follow the manual, you can build another one that works identically.

A rainforest is complex. It might have fewer "parts" than a 747, but the interactions between those parts are nonlinear, adaptive, and context-dependent. A change in one species doesn't just affect the species it directly interacts with — it ripples through food chains, soil chemistry, pollination patterns, and microclimate conditions in ways that no manual could predict.

Complicated systems can be understood by understanding their parts. Complex systems can only be understood by understanding their relationships.

When you try to fix a complex system the way you'd fix a complicated one — isolate the broken part, swap it out — the system doesn't cooperate. It adapts. It compensates. It finds a new equilibrium that might be worse than where you started. Just ask Hanoi about its rats.

The Bathtub You Can't See

Here's a mental model that will change how you see almost everything: the stock and flow model, which is the backbone of systems dynamics.

Imagine a bathtub. The water level in the tub is the stock — the current state of the thing you care about. The faucet is the inflow — whatever adds to the stock. The drain is the outflow — whatever subtracts from it. The water level at any moment is the cumulative result of all the water that's flowed in minus all the water that's flowed out.

This is obvious with a bathtub. It's invisible with everything else.

Employee Turnover

Your company is losing good people. The stock is your headcount. The inflow is hiring. The outflow is attrition. Most companies, when they see the stock dropping, crank up the inflow — more recruiters, bigger signing bonuses, faster hiring pipelines. And it works, temporarily. The tub fills back up.

But nobody looked at the drain. The outflow — the reason people are leaving — hasn't changed. Toxic manager, no growth path, below-market pay, broken culture. So the new hires leave too, at the same rate. You're now spending twice as much on hiring to maintain the same water level. You've doubled your cost without improving your system.

The fix wasn't more water. The fix was plugging the drain. But drains are structural. They require you to look at the system, not just the stock.

Your Health

You want to lose weight. The stock is your body weight. The inflow is calories consumed. The outflow is calories burned. Simple, right? So you cut calories — reduce the inflow. It works for two weeks.

Then the system adapts. Your metabolism slows down. Your hunger hormones increase. Your energy drops, so you move less — reducing the outflow. The system has responded to your intervention by adjusting every other variable to maintain its current state. Biologists call this homeostasis. Systems thinkers call it a balancing feedback loop. Either way, you've learned the hard way that the bathtub has a mind of its own.

Every system you interact with has stocks, inflows, and outflows. The moment you start seeing them, you'll understand why most interventions fail — they're turning up the faucet when they should be examining the drain.

Feedback Loops: The Invisible Hand That Runs Everything

The reason complex systems are so hard to manage is that they're governed by feedback loops — circular chains of cause and effect where the output of the system becomes its own input.

There are two types, and they do opposite things.

Reinforcing Loops: The Snowball

A reinforcing loop amplifies whatever is already happening. Small changes become big changes become enormous changes.

You post something on social media. A few people share it. More people see it. More people share it. The algorithm notices the engagement and pushes it to more feeds. More people see it. The thing goes viral. A tiny initial signal has been amplified by a loop that feeds itself.

This is the same mechanism behind compound interest, arms races, bank runs, and the growth phase of every startup. Reinforcing loops are why small things become big things faster than anyone expects.

Amazon's flywheel is a famous example. Lower prices attract more customers. More customers attract more sellers. More sellers create more competition, which drives prices lower. Lower prices attract more customers. Jeff Bezos drew this on a napkin in 2001, and it has powered every strategic decision since. He didn't design a product. He designed a reinforcing loop.

Balancing Loops: The Thermostat

A balancing loop resists change. It pushes the system back toward a target, like a thermostat that turns on the heat when the room gets cold and turns it off when the room gets warm.

Your body temperature is governed by a balancing loop. So is the population of predators and prey in an ecosystem. So is the price of a commodity — when prices rise, demand drops and supply increases, pushing prices back down.

Most of the frustrating "nothing changes" situations in your life are balancing loops working exactly as designed. The corporate culture that resists every reform. The personal habit you can't seem to break. The industry that absorbs every disruption and returns to normal. These aren't failures of willpower or leadership. They're balancing feedback loops maintaining an equilibrium — and the equilibrium is the problem.

The Trap

Here's what makes this treacherous. Reinforcing loops and balancing loops are often operating simultaneously, on different timescales, in the same system. You apply pressure, and the reinforcing loop gives you early results — the thrill of momentum. Then the balancing loop kicks in, slowly, invisibly, and claws back every inch of progress.

This is the story of almost every diet, every corporate transformation initiative, every government reform. Early wins followed by a slow, demoralising reversion to the mean. Not because the intervention was wrong, but because it only addressed one loop while the other loop was quietly doing its job.

So if you can't push harder and you can't push faster, the question becomes: where do you push?

Leverage Points: Where to Actually Push

Donella Meadows, one of the greatest systems thinkers who ever lived, spent decades studying this question. In 1999, she published a list of leverage points — places within a system where a small shift can produce big changes. She ranked twelve of them, from least effective to most effective.

The least effective leverage points are the ones most people reach for first: numbers. Adjusting parameters. Tweaking budgets. Changing the tax rate by two percent. Hiring five more salespeople. These are the equivalent of turning the faucet up slightly. They change the flow, but they don't change the structure that created the flow.

The most effective leverage points are structural:

The rules of the system. Who has power? What's incentivised? What's punished? Changing the rules changes everything downstream. This is why constitutional amendments matter more than executive orders, and why a single change in your company's promotion criteria can reshape behaviour more than a hundred motivational speeches.
The goals of the system. A healthcare system whose goal is "maximise hospital revenue" will produce fundamentally different outcomes than one whose goal is "maximise patient health" — even with the same doctors, the same equipment, and the same funding. The goal shapes every decision, and most systems have goals that nobody consciously chose.
The paradigm. The deepest leverage point is the mindset out of which the system arises. The shared assumptions, the unquestioned beliefs, the "that's just how things work." When you shift a paradigm — from "the Earth is the centre of the universe" to "it isn't," from "markets are always efficient" to "sometimes they're insane" — you don't fix the system. You replace it.

Why CEOs Keep Failing at Transformation

This framework explains a pattern that baffles business journalists: why smart, well-resourced CEOs keep failing to transform their companies.

They restructure. They rebrand. They hire consultants. They launch initiatives with names like "Project Phoenix" and "Vision 2025." They're pulling lever after lever — but they're pulling the low-leverage ones. Tweaking numbers, moving boxes on the org chart, adjusting incentives by a few percentage points.

They almost never touch the goals or the paradigm. They never ask: "What is this system actually optimised for? And is that the right thing?" Because asking that question might reveal that the system is optimised for something embarrassing — like protecting the executive team's status, or avoiding short-term risk at the cost of long-term survival.

The system isn't broken. It's doing exactly what it was built to do. You just don't like the results.

Seeing the System You're Standing In

The hardest part of systems thinking isn't learning the vocabulary. It's not feedback loops or stocks and flows or leverage points. Those are tools, and tools can be taught in an afternoon.

The hard part is seeing the system when you're inside it. A fish doesn't know it's in water. You don't see the feedback loops in your own organisation because you're a node in the loop. You don't notice the balancing mechanisms in your own habits because the balancing feels like "normal." The system is invisible precisely because you are part of it.

This is why outsiders see things insiders can't. It's why a new employee spots dysfunction in their first week that veterans have been blind to for years. It's why travel changes your thinking — not because other countries are better, but because leaving the system lets you see its shape for the first time.

Donella Meadows once said that the most important thing about a system is that no one can see all of it. She didn't mean that systems are mysterious or unknowable. She meant that the act of seeing always involves a point of view, and every point of view has a blind spot.

The Hanoi government couldn't see its own rat bounty as part of a system because it was thinking in straight lines: rats are bad, bounties create incentives to kill rats, therefore bounties reduce rats. Each step in the chain was logical. The chain itself was wrong — because it assumed the system would sit still while you poked it.

Systems don't sit still. They adapt, compensate, and reorganise. They absorb your intervention and metabolise it into something you didn't intend. The only way to work with a system instead of against it is to stop thinking about parts and start thinking about patterns. Stop asking "what's broken?" and start asking "what's connected?"

The bathtub doesn't care how hard you turn the faucet. It cares about the relationship between the faucet and the drain. And until you see both, you're just making the water more expensive.