Workforce planning as a control system
One operation, one world, two ways to plan it. The linear annual cascade against a closed loop at every horizon, on synthetic demand, in your browser.
In the planning cascade this page models, a plan is built once a year, decomposed downward into months, weeks and days, and defended until the next budget. Only the fastest loop, the interval, is closed: it is the only horizon where a measurement changes a decision inside the same cycle. The page runs that cascade (policy A) beside a planning system in which every horizon is a loop with its own sampling rate and frozen zone, the hiring pipeline is counted in every gap, and the forecast is never bent toward the budget (policy B). Both policies face the same demand, the same pipeline, the same operational levers, including the same annualised-hours band, and the same cost function. The method is on the wiki: Workforce Planning as a Control System and Feedback Loops in Workforce Planning.
Three terms. Supply line: people already decided on, requisitioned or in training but not yet on the floor. Plan of record: the headcount a policy is steering hiring toward. Frozen zone: the weeks ahead inside which a committed decision is not reversed.
The readout
How to read this
What is the same for both. The demand path and its seed. The pipeline: weeks to seat, weeks in training, the ramp. The operation's own shock absorbers: a free real-time absorb, an overtime and time-off band, an annualised-hours band in which hours banked in a trough are owed back at a peak, a deferrable share of work, a ceiling on how far handle time can be squeezed. The people dynamics: attrition that rises with sustained overtime. The cost function, including a charge on every change to the plan of record and on every cancelled requisition. The seasonal calendar, which both policies know.
What differs. Seven switches. Whether the supply line is counted in the gap, when hiring and when cancelling. Whether the forecast is updated weekly on actuals or monthly and blended toward the annual plan. Whether the finance envelope is fixed for the year or reset monthly from the forecast. How fast a gap to the plan is closed. Whether the whole gap becomes a hiring decision each month, or only the part that has persisted, weekly. Whether a frozen zone delays and protects each decision. Whether the quality aspiration is held or allowed to follow performance.
The ablation ladder turns A into B one switch at a time, so you can see which single change buys what, and which one hurts until the ones after it are thrown. The control preset removes the delays: with nothing to misperceive, the policies come out within about two percent, and what remains is the saving from fewer priced hiring moves, which a cascade could adopt without any other change. In the worlds with a pipeline, roughly half of the closed loop's advantage comes from counting the delay it faces; the rest comes from not anchoring the forecast to the annual plan and from making fewer, smaller hiring moves. The routing and frozen-zone switches are a trade-off: they lose in fast upward shocks and with long pipelines, where waiting for a gap to persist costs more than it saves.
What the numbers are not. Illustrative. The scale, the rates and the costs are round numbers chosen to be plausible for a service operation of a few hundred agents; they are not any company's. Index the results, do not quote them. Nothing here proves anything about a real operation; it shows what a structure does to itself.
The loops, named. Interval correction (minutes), the short-term hours band (days), pipeline-aware capacity (the hiring delay), structural elasticity (quarters), and the finance envelope (monthly). The pathologies A exhibits are hiring oscillation from the ignored supply line (Sterman 1989), quality erosion when the aspiration follows performance (Oliva and Sterman 2001, their Goal Erosion loop), and the overtime–attrition spiral. The catalogue is on the wiki page linked above.