Feedback Loops: Why Your Best Intentions Keep Making Things Worse
You've been in this meeting before.
A problem emerges. Smart people propose a fix. The fix is implemented. Things improve briefly. Then the problem comes back, worse than before. More fixes are proposed. The cycle accelerates. Everyone is exhausted and nobody understands why.
You're not dealing with a bad team or a hard problem. You're dealing with a feedback loop you haven't identified yet.
The Two Types of Feedback Loops
The Two Loops That Run Everything
Every system in the universe is governed by two types of feedback loops:
Reinforcing Loops (R) — The Amplifiers
A reinforcing loop amplifies whatever is happening. Growth feeds growth. Decline feeds decline.
| Virtuous Cycle (positive) | Vicious Cycle (negative) |
|---|---|
| More users | Technical debt increases |
| → More data | → Slower feature development |
| → Better product | → Missed market opportunities |
| → More users | → Less revenue to invest in refactoring |
| (Google, Facebook, every network effect ever) | → More technical debt (every legacy codebase you've ever worked on) |
Reinforcing loops have no natural stopping point. They run until something external interrupts them.
Reinforcing loops are why exponential growth is possible — and why death spirals are so hard to escape. They have no internal brake.
Balancing Loops (B) — The Stabilizers
A balancing loop resists change. It seeks equilibrium. It pushes back against any deviation from a target.
| Thermostat (simple) | Market Pricing (complex) |
|---|---|
| Room temperature | Price rises above "fair value" |
| → Gap from target | → Buyers reduce purchases |
| → Heating/cooling action | → Inventory builds up |
| → Room temperature changes | → Sellers reduce price |
| (gap shrinks, action reduces, loop stabilizes) | → Price returns toward equilibrium |
Balancing loops always have a goal. Understanding the goal reveals the system's behavior.
Every balancing loop has an implicit goal — a target state it's trying to reach. When you push against a balancing loop without changing its goal, it simply pushes back harder.
Why Loops Go Wrong
Delays: The Source of Oscillation
Here's where it gets dangerous. Every feedback loop has delays — time gaps between actions and consequences.
Delays are the primary reason well-intentioned interventions make things worse.
| Month | Event |
|---|---|
| Month 1 | Engineers overwhelmed, velocity drops |
| Month 2 | Management decides to hire aggressively |
| Month 3 | Job postings go out |
| Month 4 | Interviews begin |
| Month 5 | Offers made |
| Month 6 | New hires start |
| Month 7 | Onboarding (velocity actually DROPS) |
| Month 8 | New hires productive |
| Month 9 | OVERSHOOT — now overstaffed |
If management keeps hiring during months 3–7 (because they don't see improvement yet), the overshoot becomes catastrophic. The longer the delay, the more likely you are to overshoot, undershoot, and oscillate wildly around the target you're trying to hit.
Policy Resistance: Why Fixes Backfire
The most frustrating pattern in systems thinking is policy resistance — when an intervention designed to fix a problem triggers a system response that neutralizes or reverses the fix.
Classic examples:
- Rent control (intended to keep housing affordable) reduces supply → prices rise for non-controlled units
- Adding highway lanes (intended to reduce congestion) induces demand → same congestion returns within years
- Cracking down on drug supply (intended to reduce use) raises prices → more crime to fund addiction
- Heroic engineers fixing production bugs (intended to improve reliability) removes pressure to fix root causes → same bugs recur
The system isn't broken. It's working exactly as designed. The fix failed because it addressed a symptom, not the structure that produces the symptom.
Working With Loops
Finding the Real Leverage
So what do you do?
Map the loops before you intervene. Draw out the reinforcing and balancing loops in your system. Identify the delays. Ask: what goal is this balancing loop trying to reach? Most policy resistance becomes obvious once you see the whole map.
Change the goal, not just the flow. A balancing loop with a bad goal will fight every fix you throw at it. Change what the system is optimizing for and the behavior changes automatically.
Work with the delays, not against them. If there's a 6-month delay between action and result, don't judge your intervention at month 2. Don't double down at month 3 because you don't see results. Patience calibrated to the delay is a superpower.
Strengthen virtuous cycles, weaken vicious ones. Find the reinforcing loops working in your favor and pour resources into them. Find the reinforcing loops working against you and interrupt them early — they're exponential, so early intervention is orders of magnitude cheaper than late intervention.
The goal isn't to control the system. It's to understand it well enough to work with it.
Next: how simple rules create astonishing complexity — and what emergence means for the systems you manage.
