I. The Ladder Essay 5
The Shower Tap and the Thermostat
Loop thinking
You get into the shower and it is too cold. You turn the handle up. It gets hotter, too hot. You turn the handle down. It gets cold again. You turn it up. It gets hot. This goes on for a while, and eventually you step out and declare that the shower is broken, and you are quite sure the shower is broken, because nothing you do produces the state you want.
The shower is not broken. The shower has a delay in it. The water takes time to travel from the pipe to the head, and the water takes time to get up to temperature, and your hand takes time to read the temperature and move the handle. When the delay between the action and its effect is long enough, the corrections you make are based on information that is already out of date. You are steering a car with a three second delay in the steering.
Almost every pattern that looks mysterious turns out to be a loop with a delay in it, once you know to look. The shower, the thermostat, the queue that fills up on Friday afternoons, the person who is never late except when they are almost always late. Each is a circle where something comes back around, and each is a circle where the thing coming back is a bit old by the time it arrives.
Take the shower properly, as a loop. You want the water at 40 degrees. You feel 35. You open the tap by one notch. That is the action.
The water at the head takes a few seconds to change, so the result you get is not the result of your action. It is the result of your action, arriving late. You feel the water is still cold, because the water you are feeling has not yet been affected by what you just did.
So you open the tap one more notch. Now you have made two corrections based on one piece of information, and the second correction was made for a reason that has already stopped being true. The water climbs past 40 and goes to 50. Now you feel it is too hot and turn the tap down. By the time the cooler water arrives, the water has already overshot downward. The next reading is colder than reality, so you turn it up again, and the swing continues.
Now the second part, which is the part that makes this a loop rather than a story. Every swing feeds the next one. The turn-down produces cool water, which makes you feel cold, which makes you turn up, which produces hot water, which makes you turn down. The thing that comes back around is not the water. It is your own reaction to the water. That circularity is the loop, and the delay is why the loop swings instead of settling.
Notice what is not in that story. There is no villain and no broken part. Every single action is correct given what you believed at the moment you took it. The failure is not in any one of the decisions. It is in the timing between them, and timing is a property of the arrangement, not of the person.
Polarity: plus and minus
Every loop has a polarity, and polarity tells you what the loop does when you leave it alone. It is the sign of the feedback, and it is worth getting right because the two signs behave completely differently.
A reinforcing loop has a plus sign. The output of the loop comes back as an input in a way that makes the next round bigger. More sales leads to more reach, more reach leads to more sales. A customer tells a friend, the friend buys, the friend tells three more people, and the loop turns faster. Reinforcing loops are how things compound: savings that earn interest, reputation that brings work, a garden that grows better each year because the compost it makes feeds the soil it grows in. Left alone, a reinforcing loop runs away in one direction. That is not always bad. Sometimes you want a runaway. The point is to know which kind you have.
A balancing loop has a minus sign. The output comes back to cancel the gap it came from. A thermostat senses a temperature difference and corrects for it. You feel too hot and you turn the tap down, which makes you feel less hot. A stock of inventory falls below a reorder point and a new order is placed, which brings the stock back up. Left alone, a balancing loop settles at its target and stays there.
Most real loops contain both, in some proportion. That is why the same system can look stable for months and then change character.
Chains and loops
The first thing to ask about any puzzling pattern is whether it is a chain or a loop, because they have completely different futures. A chain is a sequence of causes with a beginning and an end. The effect at the end is final. You can trace it back, and if you find the first link, you have found the cause. A loop has no beginning, because the output comes back round to become an input. Trace it back and you arrive where you started, and you learn that the thing you were looking for is not an initial cause but the shape of the circle.
The practical consequence of the difference is enormous. If the problem is a chain, you look for the first link and you change it, and the problem stops. If the problem is a loop, changing the first link may do nothing at all, because the loop will regenerate whatever you break. The place to intervene in a loop is usually in the link that changes the shape of the circle: the delay, the gain, the size of the stock, or the target the loop is aiming at.
Delay is what turns a settle into a swing
Here is the most useful thing in this essay. Take a balancing loop with no delay. You feel 35, you open the tap, and the water at the head is now 40 and stays there, because there is no lag between what you do and what you feel. You open the tap once and you are done. The loop has a target, your correction removes the gap, and the gap stays removed.
Your action caused the result, your result was what you wanted, and your reading of the situation was accurate because it was current. Every part of the loop is working. It works because there is nothing between the cause and the effect to be out of step with.
Now put a delay in the same loop. Now your correction is always based on old information, and each correction overshoots, and the overshoot is corrected by the next correction, which also overshoots. The system that was stable becomes unstable. Nothing about the goal changed. Nothing about the strength of the correction changed. What changed is the time between cause and effect, and time was doing more work than you thought.
This is worth saying plainly because it is counter-intuitive. Delays are usually treated as pure loss. A delay in a supply chain is bad, a delay in a response to a customer is bad. But a delay is also what makes a system stable, because it damps the swing. A thermostat that responds instantly to the current room temperature would hunt all day, switching on and off in response to its own heating. The delay in the measurement is what stops it. The shower has the delay in the wrong place, at the point where you make the correction rather than the point where the effect happens, and that is exactly why it hunts.
So a delay is not a fault. It is a parameter. The question is where in the loop it sits and how long it is. The shower delay is in the wrong place, between the reading and the effect, so the correction is based on stale information. The thermostat delay is in the right place, between the effect and the reading, so the reading is stale but the correction is correctly sized for the effect that is coming.
Loops that make things grow and loops that make things stop
A reinforcing loop in a small business usually looks like this. The work you do well is the work you get asked for more of. The work you get asked for more of is the work you get better at. Being better at it means you do it more cheaply. Doing it more cheaply means you can price it lower or take on work you would otherwise turn down. This loop turns over years, and it is the reason a small firm can end up doing one thing extremely well without any plan to get there.
The same shape runs backwards as a trap, and the trap is commoner than the engine. The work you do badly is the work you are offered least of, so you get less practice at it, so you stay bad at it, so you are offered less. The loop is: no practice, no skill, no work, no practice. A reinforcing loop is neutral about direction. It will happily carry you the way you are already going, and going the wrong way, it does so faster.
Avoidance loops are the common failure. The thing you keep not doing is the thing you never get better at, and the not-doing gets more comfortable each time, and the next time it is even easier not to. In a school, the student who never asks a question in class stops feeling the cost of asking, and by the time a question would matter most, asking feels impossible. The loop is: not asking, less practice at asking, more avoidance, less asking.
There is one more shape worth knowing, and it is very common in organisations. A balancing loop with a target that keeps moving. The target moves faster than the system can respond, so the system never catches up and never settles, and everybody in it is permanently a little behind. Endless review cycles, targets raised every quarter, a team that is always nearly there. The system is not malfunctioning. The target is simply running.
What if the shower had a thermostatic valve instead of a manual tap? The valve does the loop for you, and it does it properly. It measures the water temperature at the head, compares it with a target you set, and adjusts the flow to close the gap. Crucially, it measures at the point where the water arrives, so its information is about the effect it is about to produce, and the correction is correctly sized. The valve also has a built-in delay, and the valve's designers gave it a small gain so that it settles rather than hunts.
Now change one condition. What if the valve's sensor were moved to the pipe, upstream of where the water actually reaches you? Nothing about the valve changes. The loop is identical in structure, but the delay has moved to the wrong place, so the valve now corrects on stale information and starts hunting too. The loop shape is the same. The parameter is different. The symptom is the same symptom you had with the tap.
"Everything that repeats is a loop." It is not. A pattern that repeats is a pattern. To call it a loop you need a mechanism: some output that comes back as an input, and a plausible path for it to do so. A symptom that comes back weekly because the calendar says so is a chain that happens to end in the same place each time. Calling that a loop tells you nothing, and it can hide a simple cause behind a fashionable word.
The other misconception is that understanding a loop means you can control it. It does not. Knowing that a balancing loop will settle does not tell you where it will settle. Knowing that a reinforcing loop will run away does not tell you how fast, and the speed is usually what matters. Loop thinking tells you the shape and the direction. It does not do the arithmetic for you.
Limits
Not everything is a loop. A repeated pattern is not proof of a loop, and a plausible story about a circle is not proof of a mechanism. The discipline is to insist on a mechanism: name the output, name the input it returns as, and say how it gets back there. If you cannot fill in that sentence, you do not have a loop yet.
Second limit: loops are hard to see from inside. The person in the loop is usually the one who cannot see it, because the loop is made of the ordinary actions that feel like individual decisions. Each action looks local and reasonable. That is the signature of a loop, and it is also the reason loops are so hard to diagnose by asking the people inside them.
Third limit: a model of a loop is still a model. The curve you draw smooths the noise out, and the noise is sometimes the story. Real loops run at different speeds on different days, and the variation can be more informative than the average. A clean diagram of a loop is a hypothesis, not a measurement.
Stuck
You are stuck in loop thinking when every problem becomes a cycle, usually a cycle of the same two or three factors, and nothing ever changes. The tell is the vocabulary: everything feeds back into everything, and the system is complex, and the only conclusion is that you need to intervene more.
The way out is to find the delay and the target, because those are the two handles. Ask: how long does it take between the action and the effect? And: what is this loop actually aiming at, and who chose that? A loop with no target is a reinforcing loop, which means it is running away, which means the real question is not how to stabilise it but whether it is running in the direction you want. And a loop whose target keeps moving upward is a different problem again, and the fix is not better performance. The fix is a target that stops moving.
Sources
- The stock and flow vocabulary, the idea of reinforcing and balancing loops, and the role of delay come from the systems thinking tradition, and from the study of control systems.
- The thermostat and the oscillating shower are the standard worked examples used to introduce feedback and delay.
- No specific studies, names, or dates are cited here. All examples are ordinary situations rather than documented cases.