Budget Management

The distinction almost everyone collapses

The budget is a commitment. It is approved — by a board, an owner, a funder — and it does not change. It is the basis on which performance is judged, and in many arrangements the basis on which people are paid. Its fixity is the entire point of it.

The forecast is an expectation. It updates monthly — three actuals plus nine forecast, then six plus six — and it is what staffing, purchasing and cash planning should actually run from. Its mutability is the entire point of IT.

Collapse the two and you lose both. Amend the budget mid-year and you no longer have a commitment to measure against; run the operation from a budget written before the year began and you are scheduling labour against a demand pattern that never arrived. About half of all organisations still do the latter, which is what the rolling-forecast adoption figure is really telling us.

And the cost of the confusion is asymmetric. A stale budget does not announce itself. It simply produces a series of misses that each look like an operational failure — right up until somebody notices they are all the same failure.

What makes a budget honest

• Build it bottom-up, against a demand calendar — not from last year plus a percentage. Revenue assembled from its drivers: volume and price, by segment, against a calendar of what will actually happen — events, seasonality, known contracts, the pipeline. Prior-year-plus-three-per-cent is not a budget. It is a hope with a spreadsheet around it, and it will be wrong in exactly the months that matter, because those are the months that differed from last year.

• Incremental margin is the cost-discipline metric, and it has a downside twin that nobody runs. How much of each additional unit of revenue reaches profit? In hospitality this is called flow-through and the rule of thumb is at least half — richer for room revenue, thinner for food and beverage. Every sector has its own version of the same test. But the more revealing one is “flex”: when revenue falls SHORT, do the costs actually shed? Most organisations measure the upside and never test the downside, which is how a business discovers, in a bad quarter, that its variable costs were not.

• Neither works without honest fixed/variable classification. And the classification is where the self-deception lives. Costs labelled variable that behave as fixed will destroy your flex the moment you need it, and they will do so quietly, because nobody re-tests the label until the revenue misses. If a cost has never once gone down, it is not variable, whatever the model says.

• Decompose rate from volume, or the monthly review will argue about the weather. A revenue shortfall driven by volume is a demand story: the market did not turn up, and that is largely not management’s doing. A shortfall driven by price, or a drift in cost-per-unit, is a management story. A static budget-versus-actual comparison blends the two into a single number that supports any narrative anyone wants to tell — which is precisely why it survives. Decomposition is what makes the conversation about controllables.

• The budget is a governance document, and the approval mechanics are not a formality. Board approval, owner approval, funder covenants — in hotel management agreements the operator must submit for owner approval 30 to 90 days before the year begins, and the budget then underpins performance tests and incentive fees. That is unusually explicit, but the principle is universal: somebody approves the number, and their approval means something. Which is why deadlock provisions exist — expert determination, or prior-year-plus-inflation for disputed lines — because budget disputes are routine and a governance process without a tiebreaker is not a process.

What happens when money expires

Before the transfers, the behaviour that makes them necessary — and here the evidence is unusually good, from an unusual place.

Liebman and Mahoney, publishing in the American Economic Review, examined US federal procurement data and found that in the final week of the fiscal year, when unspent budget simply vanishes, spending ran at about 4.9 times the weekly average. That much is predictable. What makes the study valuable is the second finding: they then measured the quality of what was bought. Across some $130 billion of IT projects, work procured in the year-end rush was between 2.2 and 5.6 times more likely to score below median. The money was not just spent faster. It was spent worse, and the difference is measurable.

And the distortion compounds, because a spree inflates the baseline. Next year’s budget is built on this year’s spending, which included the sprint — so the padding becomes permanent, and it is defended with the unanswerable observation that the money was, after all, spent.

The same research points at the fix, which is why it is worth citing rather than merely deploring. Where rollover was permitted — where unspent money carried forward instead of evaporating — the spike disappeared. The behaviour was never irrational. It was a correct response to a badly designed rule, and telling people to stop has never worked. Changing the rule does.

When the budget itself has to move: transfers and virement

• A transfer is not a supplement, and conflating them has a name in every audit report. A transfer — virement, in the public-finance term — reallocates within the approved total: zero-sum, nothing new. A supplement adds money, and it always requires the authority that approved the original. Treat a supplement as a transfer and you have spent money nobody approved, using a process designed for money everybody did. It is usually an honest mistake. It does not stay one.

• Thresholds run off the delegation of authority — who may move what, between which lines, up to how much. Small intra-departmental moves at manager level; cross-departmental and larger moves escalating to the controller, and then to the owner or board. Personnel lines are classically ring-fenced, because a headcount budget quietly converted into something else is a decision nobody made. And capital-to-operating movement should be prohibited without top-level approval — that is the international virement pattern, and it exists because the two kinds of money answer to different people.

• Cumulative caps beat per-transfer caps, and this is the point most policies miss. A single transfer under the threshold is fine. Eleven of them, between the same two lines, over five months, are a structural overrun being laundered one slice at a time. Salami-slicing defeats a per-transfer cap completely, and it defeats it BY DESIGN — no individual approval ever looks unreasonable, because no individual one is. The counters are cumulative thresholds (US federal grant rules require agency approval past 10% cumulatively), transfer-velocity flags, and reporting variance against the ORIGINAL budget as well as the revised one. That last control is the one organisations resist, and its unpopularity is a fairly good indicator of how well it works.

• Contingency is a control account, not a slush fund. Drawdowns happen against a materialised, documented risk or an approved change. Never against scope creep — which is what contingency is most often spent on and least able to justify, because a materialised risk has a paper trail while a scope addition merely has an argument. Authorisation is tiered. And the burn curve is watched against progress: contingency consumed faster than percent-complete is the earliest reliable warning a project gives, and it arrives long before the cost report agrees.

• Some money cannot move at all, and the boundaries are not advisory. Restricted funds are restricted: grant categories, statutory funds, and the boundary between an owner’s reserve and an operator’s budget. A capital reserve cannot plug an operating hole without the owner’s consent — not because it would be difficult, but because the money is not yours to move. Fungibility has limits, and a system that does not encode them will eventually be asked to.

How budgets fail

Revenue is over-budgeted, so the expense budget is sized to income that never arrives. That is not one error; it is one error that automatically produces a second, larger one — and the second has already been spent by the time the first is visible.

No reforecast, so labour is scheduled to a demand pattern from last October and cost-per-unit drifts upward through the soft months. The result is a profit miss larger than the revenue miss, which is the specific signature of a budget nobody is updating.

And with no transfer policy you get one of two failures, which look nothing alike. Paralysis: every five-hundred-dollar move escalates, so nothing moves and the budget stops describing reality. Or leakage: budgets are quietly rewritten all year, and by December the approved plan and the operating plan have no relationship left. Transfers mask the overrun, so the board learns the real position at year end — after every option that might have been taken has expired. Contingency becomes a slush fund, drained by month eight, and then the risk it existed for actually materialises.

Meanwhile variance commentary is written without decomposition, so the monthly review becomes a debate about the market rather than about the controllables. Everyone leaves agreeing it was a difficult month. Nothing changes, because nothing was identified.

How Zepth runs budgets

Budgets and reforecasts by department and line item, with budget, committed and actual in a single view — because committed cost is the leg that budget-versus-actual leaves out, and it is the leg that tells you about the overrun before the invoice does.

Transfers route by threshold: amount, line types, and the cumulative position, which is the one most systems cannot see. Protected lines and reserve boundaries are hard rules rather than conventions, so the transfer that should not happen cannot. The audit trail assembles itself — requester, justification, source and destination, amount, approver chain, timestamp, triggering event. And contingency is governed as its own account, with drawdown logic.

Every report shows original against revised against actual. Which means the overrun-laundering pattern has nowhere left to hide, because the original budget never stops being visible.