Control Systems

Closing the Loop

Controlling is the management function that makes the other three honest: measure results, compare against the plan, correct the gap. Without it, planning is fiction with a budget.

The control cycle has four beats:

1. Set standards — targets derived from the plan (units, costs, satisfaction scores).
2. Measure performance — actual results, gathered at useful frequency.
3. Compare — compute the variance: actual minus standard. Management by exception says ignore small noise; investigate variances beyond a set threshold, favorable or unfavorable — a cost 20% under budget can hide skipped maintenance.
4. Correct — fix the process, or, when reality has changed, fix the standard.

$$\text{variance} = \text{actual} - \text{standard} \qquad \text{variance \%} = \frac{\text{actual} - \text{standard}}{\text{standard}}$$

Three timings (the same trio as quality): feedforward control acts before the work (input inspection, hiring standards, preventive maintenance); concurrent control acts during (real-time dashboards, supervision); feedback control acts after (monthly closes, post-mortems). Feedback is the cheapest to build and the most expensive to rely on — the error has already shipped.

KPIs operationalize standards. Good ones are few, aligned to strategy, and hard to game — Goodhart's law warns that when a measure becomes a target, it stops measuring: reward call volume and calls get shorter, not better.

The balanced scorecard (Kaplan & Norton) fights single-metric myopia by tracking four perspectives at once: financial (how do shareholders see us?), customer (how do buyers see us?), internal process (what must we excel at?), and learning & growth (can we keep improving?). The four form a causal chain — trained people improve processes, which delight customers, which pay shareholders — so leading indicators up the chain warn before the money moves.

Control's design paradox: too little and drift goes unseen; too much and the measurement bureaucracy strangles the work it measures. The dose makes the medicine.

The four beats of control

Common pitfall: Investigating only unfavorable variances. A cost 20% under budget can mean skipped maintenance or unbuilt safety stock — management by exception cuts both ways.