The Annual Plan as an Open-Loop Controller

From WFM Labs

The annual plan as an open-loop controller is a critique of the conventional annual operating plan in service workforce management, read through control theory and the budgeting literature. An open-loop controller sets its output from a model of the world made at the start and does not use measurements of the world to correct it; a closed-loop controller does. The annual plan is open-loop by construction: it is built once from a forecast, decomposed into monthly, weekly and daily targets, and the sanctioned response to a measurement that disagrees with it is an explanation, not a change. The page sets out why that structure produces the behaviour operations blame on forecasting, what the budgeting literature found when it examined the same structure in finance, and what replaces it.

The page is one of a series; the architecture that replaces the cascade is on Workforce Planning as a Control System, and the loops are catalogued on Feedback Loops in Workforce Planning.

What the annual plan is for, and what it is used as

An annual plan does three jobs that an organization needs done. It states what the organization expects to happen, so that capacity can be bought ahead of a pipeline delay. It states what the organization wants to happen, so that performance can be judged. And it records what has been bought, so that spending can be governed. These are a forecast, a target and a commitment. They are different objects with different owners, different update rules and different consequences when wrong.

The conventional plan carries all three in one number. That is the structural fault. A forecast should move whenever the world moves; a target should move only when leadership decides; a commitment should move only when something is bought or released. One number cannot do all three, so in practice the commitment wins: the number is defended, the forecast is bent toward it, and the target is quietly re-based when it is missed.

The budgeting literature reached the same diagnosis for the financial budget. Tying targets, pay and resource allocation to one budget rewards managers for lying about the number, and the remedy is to sever the links so that forecasting, target-setting and resource decisions are separate acts.[1] The Beyond Budgeting movement called the annual budget a fixed-performance contract that raises cost, delays response and centralises decisions, and proposed relative targets, rolling forecasts and devolved authority in its place.[2] A balanced review found that most firms kept their budgets and added rolling forecasts alongside, and that the debate had moved toward improving budgets rather than abandoning them.[3] Workforce planners will recognise the pattern: a monthly reforecast that is mostly the annual plan is not a reforecast.

The cascade as a control structure

Read as a controller, the cascade has these properties.

  • One measurement path, sampled once a year. Actuals flow upward as variance explanations, aggregated by month and reconciled by quarter. The only place a measurement changes the plan is the next annual cycle, by which time the measurement is a year old.
  • A fixed set-point. The headcount envelope is set from the plan's peak and held for the fiscal year. A step change in demand after the plan is signed is answered inside the envelope: overtime, occupancy, deferral, a quieter squeeze on handle time.
  • An unobserved state. The supply line, the requisitions open, offers out and classes in training, is not part of the gap the monthly cycle closes. The gap is "plan minus people on the floor", so hiring continues while earlier hires are still in the pipeline.
  • Aggressive correction on a long delay. When the plan does move, the whole gap is hired at once, into a pipeline of two to four months. The combination of an aggressive correction and a long delay is the textbook condition for oscillation.[4]
  • A drifting goal. The handle-time standard, the occupancy target and the service goal are revised "to reflect performance" when they are missed for long enough. The goal erodes toward the result rather than the result rising toward the goal.[5]

None of these requires a forecasting error. They persist with a good forecast that is made once and then defended, because the world does not hold still for a year.

What it looks like in an operation

Commonly reported signatures include the following. Requisitions are opened in one quarter and frozen in the next. Attrition during a freeze is "unexpected" though it runs at the same rate every year. Overtime is paid for months while a hiring request waits for the next planning cycle, and the overtime is cheaper than the hire for long enough that the wait looks prudent. The forecast is challenged in every review because it disagrees with the plan, so the forecaster learns to produce a number that does not. Service recovers in the first quarter when the new plan lands, and the recovery is credited to the plan.

In the companion simulation at loops.wfmlabs.com, the cascade with its own pipeline delay rings after a single step in demand with a period of roughly three pipelines, and in that flat world the two policies tie on cost [estimated]. Across a seasonal year the cascade spends about three percent more per contact than a closed-loop policy facing identical demand, with service a statistical tie and the closed loop showing about a quarter less hiring churn [estimated]. A slow decline on its own widens the gap to about nine percent [estimated]. With the pipeline delay removed the two policies converge to within about two percent; roughly half of the seasonal gap is attributable to delay, the rest to forecast anchoring and churn smoothing.

What replaces it

The replacement is not a better forecast or a shorter cycle alone, but three structural changes.

  1. Three ledgers. The forecast is owned by the learning loops and is never signed. The target is set by leadership and measured against. The commitment is an envelope with authority bands, recorded as what has been bought and at what notice. Signature is reserved for breaching a band.
  2. A closed loop at every horizon. Each horizon samples at its own rate, commits only its first step, keeps a frozen zone shorter than the delay it protects, and pays a stated cost to break the freeze. The planning literature on rolling horizons found that re-planning too often has a cost of its own and that freezing the near horizon and charging for changes were among the most effective remedies compared, under certain and uncertain demand alike;[6][7] model predictive control is the same idea stated as an algorithm.[8] See Planning Nervousness and Frozen Zones.
  3. The supply line in every gap. The gap the capacity loop closes is plan minus people on the floor minus people on the way, weighted for the attrition they will suffer before they arrive. See The Supply Line in Capacity Planning.

The annual plan survives these changes. It becomes the commitment ledger and the envelope, recomputed continuously from a ledger of named assumptions rather than re-signed, which is the subject of Assumption Ledgers and the Continuous Annual Plan.

Objections

Finance needs one number. Finance needs a commitment it can govern and a forecast it can audit; it does not need them to be the same number, and the budgeting literature's advice to separate them came from finance.

A rolling plan is never stable. A rolling plan without frozen zones and change costs is not stable; one with them can be more stable than an annual plan that is defended until it is abandoned.

The pipeline is already counted. It is usually counted in the headcount report and not in the hiring decision. The test is whether a requisition opened last month reduces the gap computed this month.

Limitations

The critique is structural and does not claim that every annual plan behaves this way; an organization with a short pipeline, flat demand and a disciplined reforecast can run an annual plan without visible pathology, and the simulation reproduces that case. The budgeting evidence is largely case-based, and the control-theory reading is an analogy that borrows a lens rather than a proof. The simulation uses round, illustrative parameters and shows structure, not effect sizes.

Maturity Model Position

The cascade is the operating model at the lower levels of the WFM Labs Maturity Model™. The three-ledger split and the rolling envelope belong to Level 4: Planning in Distributions, where the plan is first expressed as a range with a version on every forecast; the closed loop at every horizon is the planning-level counterpart of the interaction-level loop in Level 5: Adaptive Orchestration.

See Also

References

  1. ↑ Jensen, M. C., "Corporate budgeting is broken—let's fix it", Harvard Business Review 79(10), 2001, pp. 94–101.
  2. ↑ Hope, J., Fraser, R., Beyond Budgeting: How Managers Can Break Free from the Annual Performance Trap, Harvard Business School Press, 2003.
  3. ↑ Hansen, S. C., Otley, D. T., Van der Stede, W. A., "Practice developments in budgeting: an overview and research perspective", Journal of Management Accounting Research 15, 2003, pp. 95–116.
  4. ↑ Sterman, J. D., Business Dynamics: Systems Thinking and Modeling for a Complex World, McGraw-Hill, 2000, chapter 17.
  5. ↑ Oliva, R., Sterman, J. D., "Cutting corners and working overtime: quality erosion in the service industry", Management Science 47(7), 2001, pp. 894–914.
  6. ↑ Blackburn, J. D., Kropp, D. H., Millen, R. A., "A comparison of strategies to dampen nervousness in MRP systems", Management Science 32(4), 1986, pp. 413–429.
  7. ↑ Sridharan, V., Berry, W. L., "Freezing the master production schedule under demand uncertainty", Decision Sciences 21(1), 1990, pp. 97–120.
  8. ↑ Rawlings, J. B., Mayne, D. Q., Diehl, M., Model Predictive Control: Theory, Computation, and Design, 2nd ed., Nob Hill Publishing, 2017.
Retrieved from "https://wiki.wfmlabs.org/w/index.php?title=The_Annual_Plan_as_an_Open-Loop_Controller&oldid=5651"