Abrigo·Marrón Systems

Apuntes de Campo Field Note

Control Is Information, Not Authority

In one heavy-industrial capital portfolio we have studied closely, one project manager's gate dates landed. Not approximately — on the planned date, repeatedly, across a portfolio of six to nine concurrent projects, while colleagues carrying similar work slipped by weeks or quarters. The organization's explanation was that he was organized. Some of his peers thought he was lucky, or sandbagging his estimates. Neither was true, and the actual mechanism is more useful than either.

He was not exercising more control. He had more observability, and in a control system those are not the same thing — though every organization we have looked at confuses them, usually in the direction that costs money.

You cannot close a loop you cannot see

Control engineering separates two questions that organizations routinely collapse into one. Controllability asks whether the inputs you have can actually drive the system to the state you want. Observability asks whether the measurements you have are sufficient to determine what state the system is currently in. They are independent properties — duals of each other, in the formal treatment — and a system can fail either one while passing the other. A fully controllable system you cannot see is just as useless as a perfectly visible system you cannot move.

What is not independent is feedback. Closing a loop means acting on the difference between where the system is and where you wanted it, which requires an estimate of where it is. Lose observability and you do not lose the ability to push on the inputs; you lose the ability to push on them for a reason. You are steering a vehicle whose windows are painted over, using a photograph of the road taken last week. Organizations in that condition do not stop steering. They steer harder, and they call it accountability.

A project schedule is a state estimate. That is all it is: a claim about where the work currently stands and how fast it is moving. Most project schedules are published and then defended — which makes them open-loop. The plan was built from assumptions, the assumptions decayed on contact with reality, and the monthly status meeting became a ritual in which everyone restates the original dates with declining conviction until a slip becomes undeniable and is announced as a surprise. It was not a surprise. It was unobserved.

What the reliable manager did differently was unspectacular and almost entirely about measurement:

None of that is authority. All of it is instrumentation. He could forecast accurately because he could see accurately, and his gate dates held because a forecast built from observed state does not need to be defended — it needs to be updated, which is a much cheaper activity.

Escalation is what organizations do when observability fails

Here is the substitution to watch for, because it is nearly universal. When an organization cannot see what is happening, it reaches for authority instead. More frequent status reporting. Senior attendance at the meeting. A vice president on the call. An escalation path with names in it.

Consider what that actually accomplishes. Authority can compel a report. It cannot make the report true. In fact it reliably makes the report less true, because now the reporter has an additional incentive — to arrive with good news — layered on top of whatever incentive they had to be accurate. Pressure applied to a measurement channel does not improve the measurement. It biases it, and the bias is in the direction that hurts most, because the organization now has confident numbers that are wrong and has stopped looking for better ones.

Control that is not carried by information is not control. It is hope with an authority signature on it. El Enfoque — Information axiom

The counterparty who misrepresents

The same logic answers a problem every project manager eventually faces and few handle well: the vendor or contractor who tells you what you want to hear. The instinctive response is interpersonal — read the room better, catch the tell, confront the lie. For some people that works. For many competent engineers it does not, because real-time social inference is simply not their strength, and building a control strategy on your weakest channel is poor design regardless of how much character it displays.

The systems response is to stop relying on the channel you cannot trust. Specify in terms that are verifiable by inspection rather than by assertion. Define the acceptance test before the purchase order, not after the delivery. Require the fabrication milestone to be evidenced by something physical. Route the same question through an independent path — a colleague who knows that equipment, a subject matter expert in rotating machinery or fired heaters who will say nobody delivers that in fourteen weeks in about four seconds.

You have not won an argument. You have made the claim checkable, which means the argument no longer has to happen. The written specification is a far better battlefield than the conference room, and it has the additional virtue of still being there after everyone involved has been reassigned.

What to do with this on Monday

Take any commitment your organization is currently relying on. Ask three questions. What is the physical quantity that would tell me whether this is true? How long does it take that quantity to reach the person who can act on it? Who verified it, independently of the party who benefits from the answer?

If the first question has no answer, you are not managing that commitment — you are hoping about it. If the second answer is longer than the time available to respond, the loop is open regardless of how good the data is. And if the third is nobody, then what you have is not a status report. It is a position statement from an interested party, and it should be read that way.

The cheapest control system in any organization is a short, honest feedback path. It usually costs nothing but the willingness to find out something inconvenient slightly earlier than you would have preferred.

Observability and controllability in the control-engineering sense follow the standard treatment; Nise, Control Systems Engineering, is the usual reference. The portfolio described is de-identified and some details are generalized; no client or employer information appears here.

← Todos los apuntes