Syntha Lab RU

Budgeting methods: rolling forecast, ZBB, what-if scenarios

Three ways to keep a budget alive: a rolling forecast that always looks a year ahead, zero-based budgeting for expenses and scenarios with preset triggers. They drive the decision on what to do when the exchange rate, demand or interest rate leaves the budget assumptions, and when to revise the plan.

Finance Brand and retail Verified: 2026-11-13

Open interactively →

Why this document

An annual budget is stale by spring: the exchange rate, demand and weather move faster than the plan. The finance director picks the method for the task: a rolling forecast keeps the horizon, zero-based budgeting exposes costs kept out of habit, and scenarios say in advance what to do if demand falls ten percent. That is why triggers should be set before a crisis, when decisions are made calmly. The owner gets not one number but a range and an action plan for each case.

At a glance

What is this document and why is it needed?
Three ways to keep a budget alive: a rolling forecast that always looks a year ahead, zero-based budgeting for expenses and scenarios with preset triggers. They drive the decision on what to do when the exchange rate, demand or interest rate leaves the budget assumptions, and when to revise the plan.
Who owns the document and how often is it rebuilt?
The finance director owns the method; line owners refresh drivers; the rolling forecast quarterly (or monthly for cash), ZBB once a year on a selected block of expenses, scenarios at budget approval and when a trigger fires.
What mistakes are the most common?
Scenarios without triggers → when it happens nobody knows what to do → write the metric, threshold and action. A scenario is built as a percentage of revenue → the effect on cash and covenants is invisible → recompute the whole model. The rolling forecast became fitting to the budget → useless → fix the budget as a target and show the gap.

For whom

Brand
Scenarios matter more for a brand because the production decision is irreversible: volume is fixed 4 to 6 months ahead. In Model B demand −10% cuts EBITDA from 35,048 to 26,017, FX +15% to 26,607, both together to 18,621. The season purchase is already paid and delivered, so stock does not shrink with sales: it grows by 9,022, year-end DIO goes from 240 to 307 days, and year-end cash falls from 14,351 to -373. The brand runs the rolling forecast by collection: every three months it recomputes sales of the future season from preorders and buy-through.
Retail
For a retailer the main risk is fixed store costs (rent, staff): with demand −10% EBITDA in Model R falls from 34,800 to 12,511 (by 64%), because costs do not shrink after revenue and the purchased stock stays on the shelves: it grows by 26,376, DIO goes from 150 to 207 days, year-end cash drops from 1,766 to -44,764. The retailer runs the rolling forecast by store and week; ZBB suits rent, marketing and administrative costs; scenarios link to opening and closing outlets.

What it consists of

Rolling forecast

  • The idea: the horizon is always the same (4 quarters or 13 months). When a quarter closes, a new one is added at the end and the nearest quarters are recomputed from actuals and fresh drivers. The annual budget stays the target, the forecast shows where the business is actually going.
  • Retailer: What to update: drivers (traffic, conversion, sell-through, FX, rate), not all lines. A detailed forecast for 1 to 2 quarters, the rest at a coarser level.
  • What not to do: do not fit the forecast to the budget. A gap between forecast and budget is information for decisions.

Zero-based budgeting (ZBB)

  • The idea: every line is justified anew, not “last year plus inflation”. Expenses are split into decision packages and ranked by return; the budget goes through while money lasts.
  • Suits: marketing, administrative expenses, rent (open or close a store), IT projects. Does not suit: materials and production costs, which follow volume.
  • Brand: Example for brand marketing (RUB 18.6 million): Mandatory: site, ad registry, platform contracts — 22% (4,092); Proven: channels with payback above target — 46% (8,556); Test: new channels and bloggers — 20% (3,720); Nice to have: image projects — 12% (2,232). If RUB 1.86 million less is needed, cut from the bottom: image projects, then tests. Ranking by return (effect ÷ cost; illustrative inputs of the example): proven 4.2; test 1.8 (expected return with a 40% success probability); image 0.6 (expert estimate); mandatory spend stays outside the competition. Packages are taken from the top by return until the cost sum exceeds the limit.

What-if scenarios and triggers

  • A scenario = a set of changed assumptions (demand, FX, rate, commission, rent) and a recomputation of the whole model: P&L, cash flow, balance sheet. Three to five scenarios: base, demand down, FX, cost of money, a combination.
  • A trigger = a measurable condition that starts a pre-prepared action (for example sell-through by week 6 below plan by 8 pp). Without a trigger a scenario stays a picture.
  • For every scenario compute three numbers: EBITDA, the minimum cash within the year and the closing cash; check whether covenants are breached.

Brand and retailer: what is computed differently

  • Brand: the rolling forecast runs by collection: every three months the sales of the future season are recomputed from preorders and buy-through; purchasing of materials before orders are confirmed is limited to the share of core items.
  • Brand: the order of stop-costs: first discretionary marketing and hiring, then capsules and image projects, last the samples of core items; the freed cash is counted in RUB thousand for each step.
  • Retailer: the rolling forecast runs by store and week; ZBB is applied to rent, marketing and administrative costs, and triggers are tied to category sell-through in weeks 4 to 8 of the season.
  • Retailer: the order of stop-costs: openings and refurbishments first, then marketing, then re-supplies; store rent and staff are cut only by a decision to close or change the format.

Brand: how the document differs and what to add

  • Brand: the brand rolling forecast is built by collection season and factory orders: the horizon is no shorter than the time from order to arrival in the warehouse.
  • Brand: the “demand −10%” scenario hits the brand harder than the retailer because of the long cycle: the purchase is already paid, DIO rises from 240 to 307 days.
  • Brand: the FX scenario (+15%) matters at an import share of 60% in cost: it moves the price of the next batch while current stock stays at the old rate.
  • Brand: a platform commission rise of 3 pp is a channel scenario: for a brand with a marketplace share of about 16% of revenue the effect is moderate, for a brand with a large online share it is critical.
  • Brand: brand triggers are tied to partner sell-through and preorders: low sell-through by week six of the season stops the second purchase wave.
  • Brand: in the brand zero-based budget platform contracts, the site and product marking are mandatory; test channels and image projects are ranked lower and switched off first.
  • Brand: the gap between preorder and actual partner take-off is shown in the forecast as a separate line: part of wholesale is not confirmed and returns later.
  • Brand: a factory delivery delay scenario of 3 to 4 weeks is needed on its own: it shifts wholesale receipts and the entry into the season window.

Retailer: how the document differs and what to add

  • Retailer: the retailer rolling forecast is built by week and store: the next 13 weeks in detail, then by quarter.
  • Retailer: the “demand −10%” scenario hits retailer cash hardest: the purchase is fixed, DIO rises from 150 to 207 days, closing cash moves from 1,766 to -44,764.
  • Retailer: a rent indexation scenario of +10% is a fixed cost: it does not depend on sales and immediately changes the breakeven of every store.
  • Retailer: a weather and traffic scenario (a warm autumn, mall flows) matters for seasonal categories: it is built on traffic and conversion, not on a general revenue fall.
  • Retailer: retailer triggers are tied to weekly sell-through and stock age: 8 pp below plan by week six freezes re-supplies.
  • Retailer: in the retailer zero-based budget rent, minimum-shift payroll and utilities are mandatory; marketing, refurbishments and new formats are ranked by payback.
  • Retailer: a weak-store closure scenario is computed with exit payments and lease obligations: a simple closure rarely saves money in the first year.
  • Retailer: a three-week delay of the seasonal purchase matters for the retailer: shelves are empty at the start of the season while rent and payroll continue unchanged.

How to read the result and which decisions follow

  • A scenario takes the quarterly cash minimum below the fixed base: write down which action starts and when (order freeze, deferral, credit line); without a written trigger the scenario protects nothing.
  • A scenario effect on EBITDA below 3% of the base (a guide): do not change the plan, keep it under monitoring; above 10%: revise purchasing and costs immediately.
  • The forecast is systematically above the actual: compute bias (BIAS) and accuracy (WAPE); a bias above a few percent means the team fits the forecast to the plan and an independent review is needed.

Control checks

  • The base scenario in the scenario model matches the approved budget to the rouble; otherwise deltas are computed from different bases.
  • The scenario switch changes only the declared assumptions: check that other cells are unchanged and the balance sheet agrees in every scenario.
  • Typical traps: constant DIO in a demand-fall scenario (a false gain), the purchase recomputed after demand retroactively, a trigger without an owner and a deadline.

Where the data comes from

Data sourceFields
Actuals for the latest quarterssales, margin, expenses, balances, weekly sell-through
Plan driverstraffic, conversion, basket, prices, FX, rate, commission, indexation
Preorders, buy-through, purchase ordersquantities, delivery dates, prepayments
Contracts and offers (loans, leases, platforms)rates, notice periods, penalties

How to build it step by step

  1. Step 1: Choose the horizon and frequency: 4 quarters, updated quarterly (cash monthly or weekly).
  2. Step 2: Define 8 to 12 drivers that move the result; assign an owner to each.
  3. Step 3: Close the quarter, enter actuals and find deviations from budget by drivers, not by lines.
  4. Step 4: Update the forecast of the nearest quarters; add a new quarter at the end.
  5. Step 5: For expenses that need revision run ZBB: decision packages, ranking by return, a cut-off by budget.
  6. Step 6: Build scenarios by recomputing the whole model and a table: scenario, EBITDA, minimum cash, closing cash, covenants.
  7. Step 7: For each scenario set a trigger (metric, threshold, deadline) and an action with an owner.
  8. Step 8: Define “stop costs”: what is frozen first, what second; how much cash it frees.
  9. Step 9: Agree boundaries with the owner: at which scenario additional funding is needed and dividends are canceled.
  10. Step 10: Quarterly check fired triggers and forecast quality (WAPE error and bias).

Formulas

MeasureFormulasNote
Driver deviationDeviation, % = (Actual − Plan) ÷ Plan × 100Compute by driver (conversion, basket), not by total revenue: the cause becomes visible.
Scenario effectΔ = Result(scenario) − Result(base)The result comes from recomputing the whole model, not from multiplying revenue by a percentage.
Forecast error and biasWAPE = Σ|Actual − Forecast| ÷ ΣActual; MAPE = mean over points of |Actual − Forecast| ÷ Actual; Bias = Σ(Forecast − Actual) ÷ ΣActualThe formula with a sum in the denominator is WAPE (weighted error); MAPE is computed per point and averaged and is inflated on small sales. The package (S&OP) uses WAPE, so the terms are consistent. Bias shows systematic skew; compute error at the level where decisions are made.
ZBB cut-offPackages are ranked by Return = Effect ÷ Cost and taken while Σcost ≤ LimitMandatory costs go first outside the competition.

Formulas in notation

Rolling horizon
F_(t..t+H) = A_(t−1) + Fc_(t..t+H)
A is the actual of the closed period, Fc is the forecast for the next H periods; after a period closes the window moves one step.
Stuck inventory
Inv_1 = P + Inv_0 − COGS; DIO = Inv_1 / COGS · 365
P is the purchase fixed before demand fell, Inv_0 is opening stock; when COGS falls stock does not shrink and DIO rises on its own.
Zero-based budget
B = Σ_(i ≤ k) c_i, Σ_(i ≤ k) c_i ≤ E
Expense packages c_i are ranked by priority; the first k enter the budget while the sum stays within the limit E.
Review trigger
θ: |Fact − Plan| / Plan ≥ θ
The threshold θ is set in advance for each metric; exceeding it starts a forecast recomputation and the decisions of the trigger map.

Worked example

What-if scenarios on the end-to-end models: EBITDA, in-year minimum cash and closing cash, RUB thousand

ScenarioBrand: EBITDABrand: min cashBrand: closing cashRetailer: EBITDARetailer: min cashRetailer: closing cashTrigger and action
Base case35,048−3,50014,35134,800−15,8981,766Plan approved
Demand −10% in all channels26,017−9,982−37312,511−47,697−44,764Sell-through by week 6 of the season more than 8 pp below plan → freeze re-supplies and the second purchase wave
FX rate +15% (import share of cost 60% and 50%)26,607−10,1732,26414,368−32,245−20,031The rate above plan by 8% for two weeks in a row → recompute prices of the new batch and shift purchasing to rouble contracts
Loan rate +5 pp35,048−4,04613,25434,800−17,575−468The cost of money 3 pp above plan → postpone dividends and accelerate repayment of the costliest line
Brand: platform commission +3 pp; retailer: rent +10%34,071−3,91613,49230,566−18,685−1,953A notice of commission growth or rent indexation → recompute channel and store contribution; decide on a price rise or renegotiation
Demand −10% and FX +15% together18,621−14,158−7,903−5,501−62,837−62,837Both triggers fired → launch the crisis plan: freeze hiring and discretionary marketing, negotiate deferrals, a second funding line

Takeaway: in the “demand −10%” scenario purchases and stock are held at the base level (the goods are already bought), so cash gets stuck in inventory: the brand ends the year at -373 against 14,351 in the base, the retailer at -44,764 against 1,766; a 5 pp rate rise hurts less. Shrinking stock together with COGS (constant DIO) would make brand cash even rise, which is a modeling error, not a gain. The retailer loses 64% of EBITDA on a 10% revenue loss: this is the operating leverage of fixed costs. The retailer’s minimum cash in the combined scenario is -62,837: an agreed funding line is needed before the season. All scenarios are recomputed by the whole model, not by percentages.

Figures

Closing cash by scenario
Bar chart: Base case 14,351, Demand −10% in all channels −373, FX rate +15% (import share of cost 60% and 50%) 2,264, Loan rate +5 pp 13,254, Brand: platform commission +3 pp; retailer: rent +10% 13,492, Demand −10% and FX +15% together −7,903−80,000−60,000−40,000−20,000020,000Base caseBase caseBrand (Model B): 14,35114,351Retailer (Model R): 1,7661,766Demand −10% in all channelsDemand −10% in allchannelsBrand (Model B): −373−373Retailer (Model R): −44,764−44,764FX rate +15% (import share of cost 60% and 50%)FX rate +15% (importshare of cost 60%…Brand (Model B): 2,2642,264Retailer (Model R): −20,031−20,031Loan rate +5 ppLoan rate +5 ppBrand (Model B): 13,25413,254Retailer (Model R): −468−468Brand: platform commission +3 pp; retailer: rent +10%Brand: platformcommission +3 pp;…Brand (Model B): 13,49213,492Retailer (Model R): −1,953−1,953Demand −10% and FX +15% togetherDemand −10% and FX+15% togetherBrand (Model B): −7,903−7,903Retailer (Model R): −62,837−62,837RUB thousand
  • Brand (Model B)
  • Retailer (Model R)
Data as a table
MeasureBrand (Model B)Retailer (Model R)
Base case14,3511,766
Demand −10% in all channels−373−44,764
FX rate +15% (import share of cost 60% and 50%)2,264−20,031
Loan rate +5 pp13,254−468
Brand: platform commission +3 pp; retailer: rent +10%13,492−1,953
Demand −10% and FX +15% together−7,903−62,837

Closing cash by scenario. The same year under different assumptions: when demand falls with a fixed purchase, cash goes negative because stock stays on the shelves.

EBITDA deviation from the base by scenario
Bar chart: Demand −10% in all channels −9,031, FX rate +15% (import share of cost 60% and 50%) −8,441, Loan rate +5 pp 0, Brand: platform commission +3 pp; retailer: rent +10% −977, Demand −10% and FX +15% together −16,427−50,000−40,000−30,000−20,000−10,0000Demand −10% in all channelsDemand −10% in allchannelsBrand: −9,031−9,031Retailer: −22,289−22,289FX rate +15% (import share of cost 60% and 50%)FX rate +15% (importshare of cost 60%…Brand: −8,441−8,441Retailer: −20,432−20,432Loan rate +5 ppLoan rate +5 ppBrand: 00Retailer: 00Brand: platform commission +3 pp; retailer: rent +10%Brand: platformcommission +3 pp;…Brand: −977−977Retailer: −4,234−4,234Demand −10% and FX +15% togetherDemand −10% and FX+15% togetherBrand: −16,427−16,427Retailer: −40,301−40,301RUB thousand
  • Brand
  • Retailer
Data as a table
MeasureBrandRetailer
Demand −10% in all channels−9,031−22,289
FX rate +15% (import share of cost 60% and 50%)−8,441−20,432
Loan rate +5 pp00
Brand: platform commission +3 pp; retailer: rent +10%−977−4,234
Demand −10% and FX +15% together−16,427−40,301

EBITDA deviation from the base by scenario. Which factor hits profit hardest: sensitivity is read from largest to smallest and triggers are set against it.

Specifics by category

Apparel
Apparel: scenarios are built by season; the main trigger is sell-through at week 4 to 8 of the season; the action: a re-supply or a price reduction.
Footwear
Footwear: the size run makes demand error costlier; a trigger by size run completeness and the share of tail leftovers; re-supply is harder (a longer lead time).
Accessories
Accessories: re-supply is faster and cheaper, so scenarios are milder; FX sensitivity is higher for imported leather and hardware.
Kids
Kids: a trigger by the school calendar and weather; leftovers cannot be carried over without a loss in size, so the “lower demand” scenario is more dangerous.
Men
Menswear: steady sales, scenarios shift to FX and purchase prices; ZBB is effective in marketing.
Women
Womenswear: higher demand volatility; fast triggers (weekly sell-through) and splitting purchases into waves are needed.

Common mistakes

  • Scenarios without triggers → when it happens nobody knows what to do → write the metric, threshold and action.
  • A scenario is built as a percentage of revenue → the effect on cash and covenants is invisible → recompute the whole model.
  • The rolling forecast became fitting to the budget → useless → fix the budget as a target and show the gap.
  • ZBB applied to all expenses at once → the team burns out → take one block a year.
  • No stop costs → in a crisis cuts are random → rank in advance what to freeze.
  • The forecast is made at a level where no decisions are taken → error does not lead to action → compute where decisions are made (category, store, channel).

Russian specifics

  • Budgeting methods are not regulated by law; but scenario decisions (rate rise, FX) rely on current Bank of Russia data: take values at the date from the Bank of Russia, in the example they are set as assumptions.
  • Foreign currency payments pass currency control; in the FX scenario take into account the payment term, the bank fee and the contract conditions. Rule: Law 173-FZ of 10.12.2003: Federal Law of 10.12.2003 No. 173-FZ on currency regulation and control (rule check not completed).
  • Notice of a platform commission increase: the term and procedure are set by the platform economy law; the “commission rise” trigger must take this term into account (see the “Platform notice period” and “Commission increase” calculations). Rule: Law 289-FZ of 31.07.2025: Federal Law of 31.07.2025 No. 289-FZ on the platform economy.

Open interactively →