Expense Budget: Fixed or Variable in a Seasonal Operation?
Why the textbook split does not hold in a seasonal operation
Every expense budget starts the same way in every textbook: tag each line as fixed or variable. In a property that trades at the same tempo all year, the tagging holds up. In a property that opens in May and closes in October it does not, because when volume goes to zero some costs do not follow it down, and others come down only because you made a decision.
Take a concrete line. The wage of the caretaker who stays on site through the closed months goes into the fixed column by textbook rules. But that wage is paid because you decided not to leave the property unattended out of season. Change the decision and the cost changes. So it is not fixed — it is discretionary. Once it sits in the fixed column it drops out of the budget conversation and nobody questions it again.
Getting this wrong does not stay contained to one line. Your break-even point, your contribution per room night and your season-length decision all break at the same time. Which is why the expense step of the budget calendar is not a formality you queue up after the revenue step — it is arithmetic you run alongside it.
I will use one property throughout. The property and every figure below are hypothetical — I constructed them, they are not any real operator's data. The point is arithmetic you can follow, and no line here should be read as an industry figure. Amounts are in Turkish lira (TL).
| Parameter | Hypothetical value |
|---|---|
| Rooms | 200 |
| Season | 1 May – 31 October (184 days) |
| Sellable room nights | 200 × 184 = 36,800 |
| Budgeted occupancy | 70% |
| Room nights sold | 36,800 × 0.70 = 25,760 |
| Budgeted ADR | TL 3,000 |
| Rooms revenue | 25,760 × 3,000 = TL 77,280,000 |
| Non-rooms revenue (F&B extras, spa) | TL 11,700,000 |
| Total revenue | TL 88,980,000 |
| Total expense | TL 75,001,840 |
| Operating profit | TL 13,978,160 |
Is a fixed cost really fixed? The closed-season test
There is a practical way to find out whether a line is genuinely fixed: take a month when the property is shut and write down what you paid. A line that goes to zero in a closed month is variable. A line that does not go to zero but could, if you decided differently, is discretionary. A line that neither goes to zero nor responds to any decision you can make this year is genuinely fixed.
| Line | Textbook class | Paid in a closed month (hypothetical, TL) | Actual behaviour |
|---|---|---|---|
| Depreciation | Fixed | 700,000 | Genuinely fixed — outside this year's decisions |
| Insurance | Fixed | 90,000 | Fixed; the cover level is a separate decision |
| Core management payroll | Fixed | 450,000 | Follows the season-length decision |
| Caretaking and security | Fixed | 120,000 | Follows the closed-season decision |
| Energy, water and waste base load | Variable | 180,000 | Semi-variable — never reaches zero |
| Planned maintenance | Fixed | 350,000 | Deliberately scheduled into the closed months |
| Operating payroll | Variable | 0 | Genuinely variable |
| Cost of goods sold | Variable | 0 | Genuinely variable |
| Closed-month total | 1,890,000 |
In this hypothetical the six closed months carry 1,890,000 × 6 = TL 11,340,000. No rooms sold, no guests on site. The textbook split books all of that 11.3 million as one fixed lump. The closed-season test shows that TL 920,000 of the monthly figure — core payroll, caretaking, planned maintenance — is really the consequence of a season-length decision.
The practical difference: a line you call fixed never reaches the budget meeting. A line you call discretionary does.
What actually drives a variable cost: occupancy, covers or revenue
Tagging a cost "variable" is half the work. If you have not written down what it varies with, the budget line cannot be computed. A hotel P&L carries four different drivers side by side, and they are not interchangeable.
| Expense line | Real driver | If you pick the wrong driver |
|---|---|---|
| Housekeeping consumables | Room nights sold | Tied to revenue, an ADR move fakes a cost move |
| Laundry | Room nights sold | Tied to guest count, it inflates on double occupancy |
| F&B cost of goods sold | Guest nights | Tied to room nights, it drifts in a family-heavy season |
| OTA and card commission | Revenue | Tied to room nights, a rate rise hides the commission |
| Energy, water and waste | Occupancy + open days + weather | Tied to revenue, the closed-season base load disappears |
| Payroll overtime | Peak occupancy days | Tied to average occupancy, it flattens the peak cost |
Something you can apply: write the driver and the unit rate next to every variable line, not the annual total. "Laundry: TL 2,447,200" is not a budget line. "Laundry: TL 95 × room nights sold" is a budget line — it recomputes itself the moment the occupancy assumption moves.
Four cost families, four different behaviours
In a seasonal property most of the cost base lands in four families. All four deserve the "semi-variable" label, and none of them behaves like the others.
Payroll
Payroll is not one line, it is three contract types: the core team that works all year, the seasonal operating crew, and the preparation crew that opens and closes the property. Each behaves differently in the budget.
| Crew | Headcount | Monthly gross (TL) | Months | Annual (TL) | Class |
|---|---|---|---|---|---|
| Core (year-round) | 12 | 37,500 | 12 | 5,400,000 | Fixed |
| Seasonal operations | 85 | 30,000 | 6 | 15,300,000 | Committed |
| Preparation and close-down (April, November) | 20 | 30,000 | 2 | 1,200,000 | Committed |
| Total | 21,900,000 |
Still hypothetical: in this structure only 5,400,000 / 21,900,000 = 24.7% of payroll is fixed in any meaningful sense. The other 75.3% follows season-length and crew-size decisions — but those decisions get made in March and cannot be unwound in July. A line that looks variable in the budget behaves like a fixed cost the moment the season opens.
This family also has a side the budget never sees. Turnover runs high on a seasonal crew, and the cost lines that never reach the P&L do not show up in the wage row. The same goes for training cost for a seasonal team, which in most properties sits outside the payroll budget in a separate and incomplete line.
Energy, water and waste
This family does not go to zero out of season. Humidity control, frost protection, security lighting and pool plant maintenance all continue. The base load is the quietest fixed cost in the expense budget.
| Month group | Monthly (TL) | Months | Total (TL) | Class |
|---|---|---|---|---|
| Closed (November–April) | 180,000 | 6 | 1,080,000 | Fixed — base load |
| Open, shoulder (May, October) | 900,000 | 2 | 1,800,000 | Semi-variable |
| Open, peak (June–September) | 1,700,000 | 4 | 6,800,000 | Semi-variable |
| Total | 9,680,000 |
In this hypothetical spread the base load is 11.2% of the annual total. The TL 8,600,000 that falls in the open season needs splitting too: part of it is the price of having the doors open at all (public-area climate, pool heating, lighting), part of it moves with occupancy. For the rest of this piece I split that 8,600,000 evenly — 4,300,000 as an opening commitment, 4,300,000 as occupancy-driven. That split is my assumption as well, not a measurement.
On the chart-of-accounts side this family was redefined for 2026: the 12th Revised Edition of USALI retired the old Utilities Schedule and replaced it with an Energy, Water and Waste (EWW) Schedule, and the edition took effect on 1 January 2026.
Maintenance and renewal
A maintenance budget is two parts that behave in opposite directions. Planned maintenance gets pushed into the closed months: you choose the timing and the amount is largely predictable. Reactive maintenance surfaces mid-season: you do not choose the timing and the amount climbs with occupancy and intensity of use.
In the hypothetical property, planned maintenance is 350,000 × 6 = TL 2,100,000, in-season reactive maintenance is TL 850,000, and expensed renewal work is TL 1,400,000 — TL 4,350,000 in total. Of that, 3,500,000 was deliberately placed in the closed months, which means you set the calendar. Holding a maintenance budget on one line puts the part you control in the same bucket as the part you do not.
Cost of goods sold
Cost of goods is the cleanest variable family in the budget, because nobody argues about the driver. Provided you pick the right one.
The hypothetical arithmetic, assuming 1.9 guests per room: 25,760 room nights × 1.9 = 48,944 guest nights. At TL 260 of F&B cost per guest night → 48,944 × 260 = TL 12,725,440. Housekeeping and laundry consumables at TL 95 per room night → 25,760 × 95 = TL 2,447,200. Revenue-driven OTA and card commission at 14% of rooms revenue → 77,280,000 × 0.14 = TL 10,819,200. All three unit rates are hypothetical.
The thing to watch: those three lines do not move together. Hold occupancy flat and drop guests per room, and F&B cost falls while commission does not. Building the budget off a single blended "variable cost ratio" makes that difference invisible.
Semi-variable costs and step costs
A semi-variable cost is a base plus a slope — energy is the standard case. A step cost behaves differently: it jumps once when you cross a threshold, then does not move across a wide band.
In seasonal hospitality the classic step cost is opening a second and third à la carte restaurant. Cross the occupancy threshold and the kitchen opens, the crew gets hired, and the cost is identical whether occupancy lands at 66% or 79%.
| Occupancy band | À la carte outlets open | Additional seasonal crew | Additional seasonal cost (hypothetical, TL) |
|---|---|---|---|
| ≤ 65% | 1 | — | 0 |
| 66% – 80% | 2 | 9 people × 4 months | 1,080,000 |
| > 80% | 3 | 18 people × 4 months | 2,160,000 |
This hypothetical property budgeted 70% occupancy, so it sits in the second band: the TL 1,080,000 step is already inside the TL 15,300,000 seasonal payroll. Something you can apply: record step costs in the budget notes with their band attached. "Above 66% this adds TL 1,080,000" is worth more in a revision round than any single number.
Why costs do not fall as fast as revenue does
Cost accounting rests on an assumption that the cost-volume relationship is symmetric in both directions. Anderson, Banker and Janakiraman tested that assumption on 7,629 firms over 20 years and found it does not hold: selling, general and administrative costs rise on average 0.55% per 1% increase in sales, but fall only 0.35% per 1% decrease. The authors attribute this to managers deliberately adjusting the resources committed to activities — costs are not sticky on their own, decisions are (Anderson, M.C., Banker, R.D. & Janakiraman, S.N. (2003) Are Selling, General, and Administrative Costs "Sticky"?, Journal of Accounting Research 41(1):47-63, doi.org/10.1111/1475-679X.00095).
One caution. Those coefficients come from cross-industry SG&A data, not from hospitality. You cannot use 0.35% as your own property's cost elasticity. What transfers is the mechanism: committed resources do not come back as fast as they went out.
In a seasonal property the mechanism bites harder, because most of the commitments are made before the season opens. The table below shows what happens in the hypothetical property if revenue comes in 10% below budget. The two right-hand columns are my elasticity assumptions.
| Line | Budget (TL) | Proportional expectation (−10%) | Assumed actual (TL) | Saving (TL) |
|---|---|---|---|---|
| Fixed costs | 23,060,000 | 23,060,000 | 23,060,000 | 0 |
| Seasonal crew commitment | 16,500,000 | 14,850,000 | 16,500,000 | 0 |
| Energy opening commitment | 4,300,000 | 3,870,000 | 4,300,000 | 0 |
| OTA and card commission | 10,819,200 | 9,737,280 | 9,737,280 | 1,081,920 |
| F&B cost of goods sold | 12,725,440 | 11,452,896 | 11,452,896 | 1,272,544 |
| Housekeeping consumables | 2,447,200 | 2,202,480 | 2,202,480 | 244,720 |
| Energy, occupancy-driven portion | 4,300,000 | 3,870,000 | 4,156,667 | 143,333 |
| Reactive maintenance | 850,000 | 765,000 | 821,667 | 28,333 |
| Total | 75,001,840 | 69,807,656 | 72,230,990 | 2,770,850 |
Hypothetical result: revenue falls from 88,980,000 to TL 80,082,000 (−10%), expense falls from 75,001,840 to TL 72,230,990 (−3.7%), operating profit falls from 13,978,160 to TL 7,851,010 (−43.8%). A ten-point loss on the top line turned into a forty-three-point loss on the bottom one.
That −3.7 / −10 ratio landing close to the 0.35 coefficient in Anderson et al. is a coincidence and validates nothing: I chose the elasticity assumptions, I did not measure them. The only thing verified here is the arithmetic itself.
Capacity commitments: where they show up in the budget
There is nothing mysterious about the source of the stickiness — it is the contracts signed before the season opens. The single most useful output of an expense budget is a table of those commitments with their decision month attached, because the only moment a cost can still be unwound is before that month.
| Commitment | Decision month | Budget class | Amount (hypothetical, TL) |
|---|---|---|---|
| Season length (6 months open) | November | Fixed | 23,060,000 |
| Seasonal crew contracts | March | Committed | 16,500,000 |
| Energy opening commitment | December | Committed | 4,300,000 |
| Committed before the season opens | 43,860,000 |
Hypothetical ratio: 43,860,000 / 75,001,840 = 58.5%. More than half the cost base is committed before the first guest arrives. In working-capital terms that means most of this 43.86 million leaves as cash before the season starts or in its first weeks, while the revenue arrives between June and September. An expense budget is a cash-timing document as much as a profitability one.
Season length is a cost decision
The hypothetical property sells rooms for six months and pays payroll for eight: the TL 1,200,000 across April and November buys two months with zero rooms revenue. Add the TL 11,340,000 fixed load of the closed months and the property carries TL 12,540,000 in months where no room is sold — 16.7% of the cost base.
That figure turns the season-extension argument into arithmetic. Test opening April for sale. Additional sellable room nights: 200 × 30 = 6,000. Shoulder assumptions: 35% occupancy, TL 2,100 ADR.
- Additional rooms revenue: 6,000 × 0.35 × 2,100 = TL 4,410,000
- Additional variable cost: 2,100 room nights × TL 1,209 = TL 2,538,900
- Step cost of opening the month: energy 900,000 + net additional crew 1,950,000 = TL 2,850,000
- Hypothetical contribution: 4,410,000 − 2,538,900 − 2,850,000 = −TL 978,900
In this hypothetical, opening April loses roughly TL 979,000. Run the same arithmetic backwards and you get the decision threshold: contribution per room night is 2,100 − 1,209 = TL 891, required room nights are 2,850,000 / 891 = 3,199, required occupancy is 3,199 / 6,000 = 53.3%. April only earns its place if you expect to run above 53% — in this property, on these assumptions.
A hypothetical break-even for a seasonal property
Now put the whole distinction into one calculation. Every row in the table below belongs to the hypothetical property.
| Step | Calculation | Result |
|---|---|---|
| Committed cost (fixed + step) | 23,060,000 + 20,800,000 | TL 43,860,000 |
| Variable cost | total | TL 31,141,840 |
| Variable cost per room night | 31,141,840 / 25,760 | TL 1,209 |
| Total revenue per room night | 88,980,000 / 25,760 | TL 3,454 |
| Contribution per room night | 3,454 − 1,209 | TL 2,245 |
| Break-even room nights | 43,860,000 / 2,245 | 19,537 |
| Break-even occupancy | 19,537 / 36,800 | 53.1% |
| Margin of safety | (25,760 − 19,537) / 25,760 | 24.2% |
Hypothetical reading: this property breaks even at 53.1% occupancy, budgets at 70%, and holds a 24.2% margin of safety in between.
Run the same calculation on the textbook split — counting the TL 20,800,000 of step cost as variable — and the table changes: committed cost drops to TL 23,060,000, variable cost per room night rises to 51,941,840 / 25,760 = TL 2,016, contribution falls to 3,454 − 2,016 = TL 1,438, break-even room nights become 23,060,000 / 1,438 = 16,036, and break-even occupancy comes out at 16,036 / 36,800 = 43.6%. Nine and a half points of difference — and the property reads as safer than it is. That is precisely where the cost of a sloppy split shows up.
Aligning with the USALI expense schedules
You do not have to reinvent the chart of accounts to build your own fixed-variable split. The 12th Revised Edition of USALI took effect on 1 January 2026, and two of its expense-side changes speak directly to this piece: the Utilities Schedule was replaced by an Energy, Water and Waste Schedule, and new schedules were added for Payroll Full-Time Equivalents (FTE) and Annual Mandatory Brand and Operator Costs. The edition also opens a dedicated section for all-inclusive properties.
The FTE schedule earns its keep in a seasonal property, because it repairs the link between headcount and cost. In the hypothetical property: core 12 × 12 = 144, seasonal 85 × 6 = 510, preparation 20 × 2 = 40 person-months; 694 person-months / 12 = 57.8 annual FTE. On peak days there are 97 people on site, but the annual FTE is 57.8. Budget the crew off the peak headcount and you overstate cost; budget it off annual FTE and you lose sight of peak capacity. Both belong in the report, separately.
The practical payoff: once your cost families sit inside the USALI schedules, your fixed-variable split lives in the chart of accounts instead of being re-litigated every year.
Materiality: which expense line earns its own budget line
You can budget every cost on its own line, but it buys you nothing. The test is materiality. The IFRS Practice Statement 2: Making Materiality Judgements — issued in September 2017, and non-mandatory guidance rather than a requirement — puts the test plainly: would omitting or misstating the information change a user's decision? If it would, it is material.
Translated into an expense budget: a line earns its own row either on size or on decision relevance.
| Expense line | Annual (hypothetical, TL) | Share of cost base | Own line? |
|---|---|---|---|
| Payroll | 21,900,000 | 29.2% | Yes — broken out by contract type |
| F&B cost of goods sold | 12,725,440 | 17.0% | Yes |
| OTA and card commission | 10,819,200 | 14.4% | Yes — revenue-driven |
| Energy, water and waste | 9,680,000 | 12.9% | Yes — base load on its own row |
| Depreciation | 8,400,000 | 11.2% | Yes — outside this year's decisions |
| Maintenance and renewal | 4,350,000 | 5.8% | Yes — planned split from reactive |
| Other fixed (marketing, IT, licences, brand) | 3,600,000 | 4.8% | Yes |
| Housekeeping consumables | 2,447,200 | 3.3% | Yes |
| Insurance | 1,080,000 | 1.4% | No — inside other fixed |
Size is not the only test. In this hypothetical property 1% of the cost base is TL 750,018; a minimum-volume laundry contract sitting below that threshold still earns its own row, because it constrains the season-length decision. The reverse holds too: a line above the threshold that no decision hangs on does not need a breakdown.
The rule for revising an expense budget
An expense budget does not get revised every time an actual moves. Do that and the budget becomes a copy of the actuals, and loses the only thing it was for. Revise on three triggers:
- Season length changed. If the opening or closing date moved, recompute the fixed and committed base.
- A step threshold was crossed. If the occupancy expectation moved into a different band, add or remove the step cost. Movement inside a band is not a revision.
- A commitment was repriced. If the energy contract, a minimum-volume agreement or a collective labour agreement changed, update that row.
Anything else is not a revision, it is variance analysis — which is what the budget exists to support. Holding this discipline matters even more for the multi-property groups that want to build the expense budget property by property: if one property's definition of "fixed" is not the other's, group consolidation produces a meaningless total.
Something you can apply: write down which assumption changed next to every revision. If you cannot name the assumption, do not revise.
Conclusion
In a seasonal operation the quality of an expense budget comes down to how honestly the fixed-variable split was built. Three habits carry it: every variable line gets a driver and a unit rate; every step cost gets recorded with its band; every line that follows season length comes out of the fixed column and into the discretionary one.
Anderson, Banker and Janakiraman's finding explains why this matters. Committed resources do not come back at the speed they went out. In a seasonal property the asymmetry is sharper still, because most of those commitments are signed months before the first guest checks in. Which is why the real output of an expense budget is not a total, it is a decision calendar: a record of which cost you committed in which month.
For the coming budget round, one habit change is enough — write the driver next to every expense line. Next season's variance analysis will pay you back for it.
Splitting the cost base in a seasonal property is not a one-table job; it takes the chart of accounts, the driver definitions and the revision discipline held together. If you hit something awkward while building that split in your own property, or you would rather talk it through with the HRMIM team, write to us via the contact page — we are glad to share what we see on the finance side of a seasonal operation.


