Doubting the Ruler

Argyris and Schön drew a distinction that has never been more useful and never been more ignored. Single-loop learning corrects action in order to hit a target: the result missed the goal, so we adjust what we do and try again. Double-loop learning goes up a level and questions the target itself—the assumptions, the governing variables, the very definition of success against which the action is being judged. A thermostat is a perfect single-loop learner: it works tirelessly to hold the room at the set temperature. It never once asks whether the room should be that temperature at all. Most organizations are thermostats.

This is not a criticism so much as an observation about where effort goes. The entire apparatus of modern management—dashboards, targets, key results, key performance indicators, continuous improvement—is machinery for optimizing within a fixed definition of success. It is genuinely valuable machinery, and the discipline it imposes is real. But it is also self-limiting in a way that is rarely acknowledged: an organization can become superbly, expensively excellent at the wrong thing, and every instrument on its dashboard will report that things are going well right up to the moment they are not.

Goodhart's law names the trap at the heart of single-loop excellence. When a measure becomes a target, it ceases to be a good measure. Any metric, optimized hard enough, is eventually gamed or hollowed out: the teacher teaches to the test, the support team closes tickets without resolving problems, the sales force hits its quarterly number by borrowing against the next quarter. The harder the organization drives toward the measure, the more thoroughly the measure detaches from the thing it was supposed to represent. Single-loop learning, pursued with enough vigor, corrupts its own instrument.

Given all this, one might expect double-loop learning to be common. It is rare, and the reason is not intellectual but political. Questioning the ruler is threatening in a way that improving against it is not. Metrics are never merely technical; they encode power, justify budgets, define whose work counts, and stand as monuments to past decisions. To ask "are we measuring the right thing?" is to imply that someone, once, chose the wrong thing—and that someone usually still has a title and a stake. The organizational immune system treats the question as an attack, and so the question is not asked, and the ruler goes undoubted precisely where doubting it matters most.

The present moment raises the price of that avoidance. When conditions shift quickly, the criteria of success age just as knowledge and experience do. A metric that faithfully captured what customers valued a few years ago can quietly stop tracking it, while every incentive in the building still points toward the old number. Single-loop optimization will then do exactly what it does best: drive the organization harder and harder toward a target that no longer means what it once meant, with a diligence that looks, from the inside, exactly like good management.

What is missing in most organizations is not the capacity to question their criteria—people do it constantly in private, in corridors and over drinks—but any legitimate, expected occasion to do it in the open. There is a time and a forum for reviewing performance against targets; there is almost never a time and a forum for reviewing the targets against reality. The first review is routine and safe. The second is exceptional, uncomfortable, and usually forced by a crisis that has already made the answer obvious and expensive.

When the questioning of criteria finally does happen, it is usually a crisis that forces it—a collapse in results, a market lost, a new entrant that competes on a dimension the incumbent never thought to measure. By then the answer is obvious and the cost of having missed it has already been paid in full. This is the most expensive possible way to learn that your ruler was wrong. The alternative is not heroism but scheduling. It means making the review of criteria a routine, expected, and psychologically safe occasion, deliberately decoupled from any individual's failure, so that the question "are these still the right measures?" is asked while it is still cheap to answer and while there is still time to act on the answer. An organization can put such an occasion on its calendar. Almost none does—because the calendar is already full of reviews of performance against the very measures that themselves go unexamined.

Every organization audits its performance against its measures. Almost none audits its measures against the world. And the second audit is exactly where the largest and most durable errors hide—not in failing to hit the target, but in hitting, year after year, a target that stopped being worth hitting. It is the last place most of us think to look, because looking there requires doubting the very instruments by which we have been reassuring ourselves that all is well.