The end of the budget game

The game, named in 2001

Jensen’s indictment — published in the Harvard Business Review as the summary of a paper whose full title says everything, “Paying People to Lie” — identified corporate budgeting’s two-stroke engine of value destruction. In formulation, subordinates lowball what they know and superiors inflate what they demand, so the numbers that emerge coordinate nothing; the organization plans against a fiction both sides authored. In realization, behavior bends around the kinks of the bonus schedule: sprint when the target is reachable, sandbag next year’s baseline when it is exceeded, defer everything possible when it is out of reach — channel stuffing, spend flushes, the fourth-quarter repertoire every controller recognizes and no one books as a cost.

Why it survived every planning tool

Twenty-five years of planning platforms did not dent the game, because the game was never about the spreadsheet. It runs on information asymmetry: the people who know the business’s honest trajectory are the people paid against the number they disclose. Every implementation faithfully digitized the negotiation. The ratchet made it self-perpetuating — beat the target and next year’s rises, so the rational manager beats it by a little, forever. The budget process is not broken planning; it is working incentives, pointed at the wrong objective.

The machine baseline ends the negotiation

The self-steering plan (the previous paper in this section) changes the one variable the game depends on. When the expectation is produced by agents from live drivers — decomposable on demand, with forecast bias tracked per owner as a visible metric — the baseline stops being testimony and becomes measurement. There is nothing left to negotiate about what we expect; the conversation moves, correctly, to what we will attempt. Sandbagging does not become immoral; it becomes impossible, because the sandbagger’s private information is no longer private. The asymmetry the game ran on for a century is an artifact of who produced the forecast — and that just changed.

The objection arrives on cue: models can be gamed too — feed the machine pessimistic pipeline data and the sandbag survives in the inputs. It is the right worry aimed at the wrong architecture. In the self-steering state the drivers are not testimony either; they are drawn from source systems — bookings, consumption, orders — that already exist for other reasons and are reconciled by the same evidence machinery the rest of this publication describes. A manager who wants to bias the forecast must now bias operational data that the close, the auditors, and the revenue-assurance agents are all watching, which converts a safe negotiating tactic into a control incident. The game does not merely become harder; it changes category, from politics to misconduct — and organizations are considerably better at policing the second than refereeing the first.

What the game does to the operation

The budget game is usually filed under compensation; its costs land in operations. The fourth-quarter sprint is a working-capital event — receivables ballooning as revenue is pulled forward, discounts burning margin to move the kink, inventory shipped to distributors who did not ask for it. The January hangover is a close event: returns, rebates, and reversals that the reconciliation and revenue-assurance machinery of Revenue, Reconciled then spends a quarter unwinding. And the spend flush is a procurement event — December consuming budget so January can claim it back. Every one of these is volume the autonomous operation must process and evidence, generated not by the business but by the payout curve. Ending the game is, among everything else, a load-shedding program for the back office: the cheapest transactions to automate are the ones that stop existing.

The transition: a year of running both numbers

Taking the sandbag away is a political act, and pretending otherwise is how the redesign fails. The workable sequence runs a full year. First quarter: the machine baseline runs silently beside the negotiated budget, and finance privately calibrates bias by owner — no consequences, just measurement. Second: both numbers appear in the pack, labeled, and the gap between negotiated commitment and measured expectation becomes a discussable fact rather than an accusation. Third: targets for the coming year are set as explicit ambition above the machine baseline — with a one-time amnesty that resets every ratchet, because the first honest baseline will embarrass precisely the managers who played the old game best, and punishing them for it teaches the organization to fight the new system. Fourth: the payout curve goes linear with the new comp cycle. A year sounds slow until it is compared with the alternative — every faster route this author has seen attempted converts the finance function’s most useful reform into its most resented one.

Designing targets after the game

Target setting survives; it finally becomes honest. The design has three parts. Separate the two numbers formally — the forecast (the machine’s unbiased expectation, owned by finance) and the target (leadership’s chosen ambition above it, owned by the business) — and publish both, so stretch is a visible policy rather than a smuggled negotiation. Linearize the payout, as Jensen prescribed: reward performance itself across the range rather than performance relative to a kink, and the fourth-quarter repertoire loses its arithmetic. And measure truth-telling: when bias-by-owner is on a dashboard, the honest forecaster is finally the well-reviewed one. What disappears is not accountability but theater — and with it, the quiet lesson Jensen warned about, that an organization which pays for gamed numbers is training its people in exactly the integrity it will later wish it had.