You’re in an unfamiliar shower. The water comes out freezing, so you crank the hot tap. Nothing happens — still cold — so you crank it further. Then, a few seconds later, a wall of scalding water hits you. You yelp and crank the cold tap hard. Cold again? No, still scalding, so you crank it more — and a moment later you’re shivering under an icy blast. You have just performed, with your own hands, the single most important phenomenon in this entire course: a balancing loop with a delay, oscillating.
In the last lesson, a balancing loop was a model of competence: it sensed the gap to its goal and closed it, gliding smoothly to a target and holding there. That picture had one hidden assumption — that the loop can see the result of its last action immediately. Break that assumption, and the well-behaved thermostat turns into the maniac at the shower tap. The villain is the delay, and it is the reason real systems so rarely settle calmly. They overshoot. They swing. They ring like a struck bell.
What a delay is
A delay is a lag between an action and the moment its effect can be sensed or felt. The hot water is on its way; you just can’t feel it yet, because it’s still travelling down the pipe. The delay isn’t a mistake in the loop — it’s a gap in time between cause and perceived effect. And almost every real loop has one: it takes time for an order to be delivered, for a hire to become productive, for a policy to bite, for a population to grow, for a body to metabolize a drug, for a planted forest to mature.
Recall from lesson 1 that stocks create delay — a bathtub doesn’t empty the instant you pull the plug. A delay is what happens when the thing your loop is trying to control sits behind one of those slow-to-change stocks. The loop acts, but the stock responds late, so the loop is always, in effect, reacting to old news.
The one-sentence version
A delay is the lag between acting and seeing the result. A balancing loop that can’t see its last move’s effect yet keeps acting on stale information — so it overshoots its goal, swings back, overshoots again, and oscillates.
Before you read — take a guess
You adjust a shower's temperature, but there's a 3-second pipe delay before you feel each change. You keep cranking until it feels right. What's the most likely result?
Why a delay turns a stabiliser into an oscillator
Here’s the mechanism, step by step. A balancing loop’s job is to close the gap between where the system is and where you want it. To do that, it needs to know where the system is right now. The delay robs it of exactly that: it can only see where the system was a few moments ago. So it sizes its correction to stale information — and by the time the correction lands, the situation has moved on, the correction is now too big (or pointed the wrong way), and the loop has overshot.
Then it does the same thing in reverse. Sees the overshoot (late), corrects hard (too hard, because it can’t yet see that the first correction is already working), overshoots the other way. Each round it’s fighting a ghost — the system as it used to be. That back-and-forth is oscillation: a repeating swing around the goal instead of a clean approach to it.
Don’t take it on faith — make it happen. Switch the simulator below to Balancing, then drag the Delay slider up from zero and watch the smooth glide of lesson 3 buckle into a wave:
Run the loop
Add a delay and watch it ring
Pick a loop type, set its strength, and — for a balancing loop — add a delay. Watch how a reinforcing loop runs away, a balancing loop glides to its goal, and a delay makes that same loop overshoot and oscillate.
A reinforcing loop feeds on itself: 20 compounds to about 3673 — roughly 183.7× the start — and just keeps climbing. Nothing here pulls it back; the output is its own input.
Notice the two knobs that control how bad it gets: the size of the delay and the strength of the correction. A longer delay means more time to over-adjust before reality catches up. A more aggressive correction means each over-adjustment is bigger. Push both far enough and the oscillation stops shrinking and starts growing — a system tearing itself apart, which engineers call instability.
The pitfall: ignoring the delay
Linear intuition says “I made a change, so the effect should show up now.” In a delayed system that intuition is poison: the effect of your last move hasn’t arrived yet, so if you keep pushing until you see a result, you will always push too far. The fix is counterintuitive — act, then wait for the delay to play out before acting again.
When to use it
Reach for the delay-and-oscillation lens whenever a system swings, hunts, or boom-and-busts instead of settling. The instinct is to blame the swings on some outside shock. Usually there’s no shock at all — just a balancing loop trying to do its job through a lag. Ask: what is this loop trying to stabilise, and how late does it see the result?
A worked oscillation, traced step by step
Let’s put numbers on the shower. Your goal temperature is 100 (comfortable units), the water starts at 60 (cold), and there’s a delay before each adjustment is felt. A patient operator who eases off — adjusting gently and giving the delay time — gets a damped oscillation that shrinks toward the goal:
| Step | What you feel | Your reaction | Overshoot vs. goal (100) |
|---|---|---|---|
| 1 | 60 (freezing) | Crank hot | −40 |
| 2 | 140 (scalding) | Back off to cold | +40 |
| 3 | 80 (chilly) | Nudge hot | −20 |
| 4 | 110 (a bit hot) | Nudge cold | +10 |
| 5 | 95 (almost) | Tiny nudge | −5 |
| 6 | 102 (basically fine) | Done | +2 |
The swings halve each round: 40 → 20 → 10 → 5 → 2. That’s a damped oscillation — it rings, but the ringing dies out and the system settles. Now watch what an impatient operator does — someone who panics and over-corrects each time, refusing to wait for the delay:
| Step | What you feel | Your reaction | Overshoot vs. goal (100) |
|---|---|---|---|
| 1 | 60 | Crank hot, hard | −40 |
| 2 | 150 | Slam cold, harder | +50 |
| 3 | 40 | Slam hot, harder still | −60 |
| 4 | 170 | Panic | +70 |
Same delay, same goal — but because the corrections grow instead of easing, the swings grow: 40 → 50 → 60 → 70. This is a growing (unstable) oscillation, and it’s the worse failure: trying harder makes it worse. The cruel joke of delayed systems is that the aggressive, “do something now!” response is exactly the one that blows them up.
Two people use the same delayed shower. One settles to comfortable in six adjustments; the other ends up wildly bouncing between scalding and freezing. What did the second person do wrong?
The bullwhip: oscillation that travels and grows
The shower is a single person fighting one delay. Now chain several delayed balancing loops together and you get one of the most expensive phenomena in business: the bullwhip effect (often taught through the “beer game,” a supply-chain simulation).
Picture a chain: customers → retailer → wholesaler → factory. Each link reorders to keep its shelves stocked (a balancing loop, with a goal of “enough inventory”), and each order takes time to arrive (a delay). Now a small, one-time bump in customer demand ripples upstream:
| Link | Sees… | Orders… | Why it overshoots |
|---|---|---|---|
| Retailer | Demand up a little, shelves dipping | A bit extra, plus a safety buffer | Can’t restock instantly (delay), so over-orders |
| Wholesaler | The retailer’s inflated order | Even more, plus its own buffer | Reads the bump as bigger than it is |
| Factory | The wholesaler’s doubly inflated order | Ramps production hugely | Sees a tidal wave that was a ripple |
By the time all that product finally arrives, the original demand bump is long gone — so now everyone is drowning in inventory and slams orders to zero, which starves the chain, which triggers panic re-ordering, and the whole system swings between glut and shortage. A tiny, smooth change in real demand becomes a violent oscillation that gets bigger the further upstream it travels — the crack of a bullwhip. The cause isn’t stupidity or greed; it’s structural: stacked delays plus each link reacting to stale, amplified signals.
Each is a system that swings or settles. Sort it by what the delay does.
Place each item in the right group.
- Boom-and-bust hiring: mass-hire in the boom, mass-layoff in the bust, repeat
- A patient cook tasting a stew, adjusting salt a little, waiting, tasting again
- Cruise control gently easing the throttle as a hill approaches
- A supply chain where each link over-orders off the last inflated order
- A driver overcorrecting a skid, jerking the wheel each way harder
- A central bank that waits to see if its last rate change worked before the next
Boom and bust: the economy’s shower
The same machine runs the biggest cycles we live inside. Commodity prices are the classic case. Prices spike (say, oil is expensive), so producers invest in new wells and mines — but a new mine takes years to come online (a long delay). By the time all that new supply finally floods the market, prices have already been high enough to dampen demand, so the glut crashes the price. Low prices kill investment, so years later supply dries up, and the price spikes again. The cycle repeats roughly on the length of the delay. Hog farming, real estate, semiconductor fabs, and tech-worker hiring all do the same dance, for exactly the same reason: a balancing loop (supply chasing price) trying to hit a moving target through a long delay.
This is also where the course’s models stack. A boom is often a reinforcing loop (rising prices → optimism → more investment → more demand) that finally collides with a balancing loop (supply catching up), and the delay between them is what turns a smooth correction into a violent bust. You can now name all three parts of a boom-bust cycle by their machinery.
Match each term to its precise meaning.
Pick a term, then click its definition.
How to tame a delayed loop
If delays cause the trouble, the fixes follow directly — and none of them is “push harder.” There are three real levers:
- Shorten the delay. The fastest cure is to see results sooner. Real-time dashboards, faster feedback, shorter supply lines, smaller batch sizes. A shower with an instant-read thermometer at the head barely oscillates. (This is a leverage point, the subject of the next lesson.)
- Ease off the correction. If you can’t shorten the delay, react more gently and wait for the delay to play out before adjusting again. This is why good central banks move in small, spaced steps and why “let’s wait and see if the last change worked” is often the wisest thing in the room.
- Anticipate instead of react. Forecast where the system is heading, not where it was, and correct toward that. A skilled driver steers for where the car will be; a novice reacts to where it already is and fishtails.
Because “harder” increases the strength of the correction while doing nothing about the delay. A bigger correction applied to stale information is a bigger overshoot — which then demands an even bigger counter-correction. That’s the recipe for a growing oscillation. In a delayed system the heroic, decisive, “do more now” instinct is the trap; patience and smaller, spaced moves are the skill. The system doesn’t need more force. It needs you to stop reacting to its past.
Fill in the core principle.
Pick the right option for each blank, then check.
A balancing loop with a overshoots its target and , because it keeps correcting on information. The swings grow when the correction is too — so the fix is to shorten the delay or ease off, never to push harder.
When to use it
Use these fixes the moment you spot a system hunting back and forth. Resist the reflex to intervene harder or faster; ask instead, can I see results sooner, and can I afford to wait one delay before my next move? Most oscillation problems are solved by removing impatience, not by adding effort.
Recap
Big picture
Delays & Oscillation
- Delays & Oscillation
- What a delay is
- A lag between acting and sensing the effect
- Stocks create delays — the loop reacts to old news
- Why it causes oscillation
- Balancing loop acts on STALE info → overshoots the goal
- Corrects late → overshoots the other way → swings
- Damped (swings shrink) vs. growing (unstable)
- What makes it worse
- Longer delay = more time to over-adjust
- Stronger correction = bigger overshoot
- Trying HARDER amplifies the swing
- In the wild
- The scalding shower
- Bullwhip / beer game: swings grow up a supply chain
- Boom & bust: supply chasing price through a long delay
- How to tame it
- Shorten the delay (faster feedback)
- Ease off; wait one delay before correcting again
- Anticipate where it’s heading, not where it was
- What a delay is
Check yourself on delays
Why does a delay make a balancing loop oscillate instead of glide to its goal?
Check your answer to continue.
Where this goes next
You can now read all three behaviours a loop produces: a reinforcing loop runs away, a balancing loop settles, and a balancing loop with a delay oscillates. That’s the full grammar of systems. The last teaching lesson, Leverage Points, asks the practical question this whole course has been building toward: now that you can see a system’s loops and delays, where do you push to change what it does — and why is the obvious place almost always the wrong one?