Sometimes when you look at a control system, like a thermostat or advanced cruise control systems, it’s easy to forget how cool and omnipresent they are.
Think about how you stay upright on a bike. You know, you never are just perfectly balanced: you subconsciously lean a little, shift your weight, steer slightly, and so on. Those tiny adjustments are fundamental in keeping you upright. And their little dance is feedback control.
Probably the most well known and used feedback control system is the PID controller.
A PID controller works like this. You set some sort of objective, called the setpoint. Then the system measures how far you are from the setpoint. That measure is called process variable. The system now takes that and calculates the error: that’s how distant the process variable is from the setpoint. Next the system puts in a correction based on three control terms: a Proportional term, an Integral term, and a Derivative term.
The Proportional term is about making the correction well, proportional to the error. Big error gets a big correction, small error a small correction.
The Integral term is about taking into account all the past errors. This helps to make better corrections based on a broader context.
The Derivative term is about taking into account how fast the error is changing. Not just its size, but how it’s trending. This keeps things from oscillating wildly around the target.
So that’s a PID controller: proportional, integral, derivative.
For example, a climate control system measures say, temperature.
If it’s a lot hotter than the temperature you’ve set, it will start strong, if it’s close enough it will just be a breeze. That’s P.
If in spite of the air pumping the temperature doesn’t go down for a while, it will increase the power. That’s I.
And if the temperature is quickly getting close to your target temperature, it will reduce the power a little earlier so it doesn’t get freezing. That’s D.
Process Control and Humans
As it happens, I read something about this, and my brain went on a tangent wondering if humans and teams work the same way.
In my experience, when managing a process or a project, we tend to do some mix of the following:
- decide how to react in the moment
- apply to processes, previous experiences and established routines
- try to anticipate change and how things will evolve
We can consider acting quickly and adjusting on the fly as P, relying on past data and experiences as I, and looking ahead as D.
The thing is, we want to manage all kinds of different things, which means we might get the mix of P, I and D off in any specific case. Because, you know, biases, and human imperfection or something.
But, the mix of PID tells us something about how we approach the problem at hand, I think. I think the P tells us how much we are in the present, the I how much we are in the past, and the D how much in the future.
I like to imagine these as the corners of a triangle, that I called the PID Compass because why not.
Your handling of whatever sits inside that triangle.
Each edge represents a different posture:
- Responsiveness (P–I): how much you act versus how much you wait for more information.
- Foresight (I–D): how much you rely on what worked in the past versus what you sense is shifting.
- Agility (P–D): how well you move between responding and anticipating.
Note That there is no right or wrong posture in absolute terms. Don’t make the mistake of thinking that, for example, faster responses are always better: sometimes the right thing to do is pause and assess.
Note also that the edges are interconnected: for example if you tend to rely on what worked in the past a lot, you’ll probably end up less responsive, because you rely on clear retrospective information to act. So, being higher I impacts both Responsiveness and Foresight.
Using the Compass
Ok so how do I use this thing?
Here’s the compass. It’s pretty simple.
Here’s how to use it.
Think about the situation you’re assessing. Say it’s how you managed an incident.
First, you want to assess where you are, on the edges.
- Was the response swift or calculated? Plot it on the I-P edge.
- Was the response based on established processes or did you go on a tangent? Plot it on the I-D edge.
- Was the response reactive or proactive? Plot it on the P-D edge.
Done? Great.
Now, retrospectively, think about what could be better next time: should we, say, try to be less panicky? Maybe we need to bring more I in the picture. But be aware: increasing the I (for example by defining better procedures and runbooks) will make things more structured and organised, but slower and less open to creative approaches to fix the problem.
Maybe that’s what you want, or maybe you need to figure out the right tradeoff, and be more deliberate in the level of detail of the runbook or procedure, who knows.
Tuning, not centering
It’s tempting to think the goal is to reach the center of the triangle, the perfect center of gravity between P, I and D. It’s not.
In a physical control system, “balance” isn’t static. It’s tuning: small, continuous adjustments.
Healthy systems move. They swing toward P when reacting to change, then back toward I to consolidate, then into D to plan ahead. You don’t live in the middle; you orbit it.
Each edge and corner has its uses.
When things feel off, instead of asking what’s wrong?, try: “Am I being too reactive, too stable, or too predictive?” Then make a small adjustment in the opposite direction.
The PID compass is just a practical way to see how you’re regulating yourself or your team. That’s it.
The PID compass is also just something I came up with. I think it’s pretty useful, but I haven’t explored its uses deeply, or even thought much about how it could be better.
If you use it or find it interesting, or have ideas about how to make it more useful or intuitive, or whatever, I want you to tell me!