Forecast a seasonal business by modelling when cash moves, not just how much. Take at least three years of your own history, strip out the trend to find the seasonal shape, put every committed outflow and fixed-date obligation on the calendar, then find the lowest projected cash point across the cycle and test it against a peak that arrives late or short.

Seasonality is a timing problem before it is a volume problem

Most published advice treats this as a volume exercise: identify the busy months, plan around the quiet ones. That is the easy half, and it is not where finance teams get caught.

The money you spend to serve a peak leaves before the money that peak generates arrives. Stock is committed and paid ahead of demand. Seasonal staff are on payroll through the busy period while the invoices raised in that period are still inside their payment terms. Tax and contractual obligations land on fixed dates regardless. So your cash trough and your revenue trough are usually different months, and the dangerous point often sits just before the peak, when everything has been paid for and nothing has come back.

It is also worth checking whether this page describes you at all, because seasonality is wider than the sectors that dominate the writing about it. Federal Reserve Bank of Chicago research measuring the amplitude of seasonal employment across US industries found construction to be by far the most seasonal, at more than three times the level of retail trade, with government close behind retail and wholesale, information and education among the least seasonal. That measures employment rather than cash, and it is US data, but it makes the point: education-linked suppliers, events, agriculture-adjacent processing, rental and hire, and professional services with cyclical demand can all carry a materially seasonal cash profile. The structural feature is not the sector. It is a large share of annual revenue concentrated in a narrow window against year-round fixed costs. If your cash gap has no seasonal explanation at all, that is a different diagnostic.

That means mapping dates, not months. For every material cost category, four dates matter and they are rarely the same: when you commit, when you pay, when the related revenue is invoiced, and when that invoice converts to cash. Resist importing lead times from someone else's article. Published examples put stock commitments three to six months ahead of a peak, but those come from retail and hospitality write-ups, and two of the three research runs behind this page found no defensible general lead time across sectors. The shape generalises; the numbers come from your own purchase orders.

Model collections as a lag from the invoice date rather than as a share of the same period's revenue. That same arithmetic produces a trap worth pre-empting before you present: debtor days can spike during a peak even when every customer pays on time, because period-end receivables hold a month of invoices not yet due. It reads as a collections failure and it is not one.

How to quantify your seasonal pattern from your own history

Published guidance almost always says two to three years of history. That convention has a real basis, and the basis is more demanding than the convention. US Census Bureau seasonal adjustment tooling requires three years and notes that more than four is often needed for an adequate result. The ONS says identifying a changed seasonal pattern typically needs at least three years plus a real-world reason, because the evidence accumulates slowly. IMF guidance treats five years as the general minimum, with three acceptable only where the seasonal movement is strong and stable.

You are solving an easier problem than a statistics agency, because you also have order books and supplier schedules they do not. But two years gives one year-on-year comparison, which is a placeholder rather than a pattern. Three is a working floor. Five is better.

Step 1: Pull at least three years of monthly cash history. Use receipts and payments rather than invoiced revenue, per revenue stream and per entity. A group total can hide two entities with opposite seasons.

Step 2: Strip the trend before you read the season. Calculate a 12-month moving average and express each month as a ratio to it. The average covers a full year, so it cancels the seasonal cycle and the ratio is de-trended. Skip this and a growing business reads its own growth as seasonality.

Step 3: Index the outflows separately, not just the income. Build a profile for stock, payroll, freight and marketing in their own right. One revenue-shaped index applied across the forecast reproduces the exact error this page exists to prevent.

Step 4: Decide what each unusual period actually was. A data error gets corrected. A true one-off is excluded and documented. A step change in the size of the business is rebased, not averaged away. A change in the timing or shape of the season means your older indices are stale.

Step 5: Chart the index by year, not just its average. If your peak month's index has drifted in one direction across three or four years, the season is moving and a long average will lag it.

Step 6: Hold the pattern as a written assumption. Record the index, the years behind it, what you excluded and who owns it. A seasonal factor buried in a formula cannot be reviewed.

The dated obligations that deepen the trough

Periodic tax and contractual payments are the most common reason a forecast that looked survivable turns out not to be. They get modelled as smoothed monthly accruals rather than the lumps they are.

In the UK, standard VAT is invoice-based and quarterly, so output VAT on peak invoices falls due whether or not those customers have paid. Most UK seasonal advice recommends the VAT Cash Accounting Scheme as the fix; at this size it does not apply, because the entry threshold is £1.35m of VAT taxable turnover and businesses must leave above £1.6m. Corporation tax is due nine months and one day after the period end for most companies, but taxable profits above £1.5m mean quarterly instalments during the accounting period instead, and that threshold is divided by the number of associated companies, so groups reach it sooner than expected.

In the United States, sales tax filing frequency is assigned by each state on liability volume rather than revenue, so a multi-state seller is often monthly in its larger states and quarterly elsewhere; on federal estimated tax, the IRS instructions to Form 2220 let corporations with seasonal income use the adjusted seasonal instalment method where the eligibility test is met. In Australia, quarterly BAS is the normal regime at this size, so GST collected at peak is remitted in the following quarter. In New Zealand, two-monthly GST filing is the norm, due on the 28th of the following month. Annual insurance renewals, rent quarters and facility fees belong on the same calendar, on their real dates.

Building it: two horizons, one set of numbers

Run a rolling 12-month monthly forecast and a rolling 13-week weekly forecast in parallel. The 12-month view carries a full cycle and is what sizes a facility and supports the board conversation. The 13-week forecast is a direct, dated view of receipts and payments, and tells you which week is tightest and which invoices are carrying it. Refresh the monthly view after each close and roll the 13-week view weekly, replacing the completed week with actuals. Running both in one spreadsheet is where seasonal models usually break, because the two horizons drift apart the moment either is edited by hand; there are alternatives worth weighing before you commit to maintaining two of them.

Then read off the number the exercise exists to produce. Take opening available cash, run cumulative net cash across the cycle, and find the lowest projected point. The gap between that point and the minimum operating balance you are willing to hold is your peak funding requirement. Calculate it from the cash forecast rather than an annual working capital ratio, and cross-check it against expected inventory, receivables and payables at the peak.

Stress-testing: what if the peak is late, or short

This is the section most seasonal guidance skips, and it carries the actual risk. A trough you can see coming is a planning problem. A peak that underdelivers after you have paid for it is a liquidity problem.

Start by discarding the round numbers. The downside percentages in circulation, usually 10%, 15% or 20%, are asserted without stated basis across vendor and advisory content. The professional bodies do not fill the gap either: sensitivity and scenario guidance treats the percentage as an assumption the analyst supplies. You are not missing a standard here. There isn't one.

Size the downside from your own record instead. Reconstruct what you were forecasting at comparable points before previous peaks, then measure two errors separately: how wrong you were on amount, and how wrong you were on timing. A peak arriving two weeks late is a different event from a peak arriving 10% smaller, and most models only represent the second.

Five scenarios are worth building.

• Volume miss. The peak arrives on schedule and comes in smaller.
• Timing slip. The volume arrives, later. Move the order, invoice and collection dates rather than cutting the annual total.
• Collection slip. Sales land as planned and customers pay later than assumed.
• Committed-cost squeeze. Stock, deposits and seasonal payroll happen as planned while the income side weakens.
• Combined. A smaller, later peak with the pre-peak costs already spent.

Then run the case that turns this into a decision. Rather than arguing about whether 15% is severe enough, start from your floor, whether that is a covenant, a facility limit or a minimum operating balance, and solve backwards for the combination of shortfall and delay that reaches it. That gives a board something it can act on: how much disappointment this season can absorb, and the date by which management has to move.

Pair each scenario with the indicators that would tell you it is happening and the last date the related decision is still reversible. Bookings against the same week in prior years, pipeline conversion, deposit take-up and supplier lead-time slippage all move before cash does. The stock commitment, the hiring decision and the campaign spend each have a final reversible date, and the forecast's job is to attach a trigger to each while the money is still yours. Scenario planning is what stops this being an annual spreadsheet exercise.

A worked example

The figures below are illustrative rather than drawn from a customer, and are here to show the shape.

A UK supplier of equipment and consumables to schools, 34 staff, £6.4m revenue. Orders concentrate around the September term: August and September each run well above an average month, February at roughly half. Stock is committed with overseas suppliers in March and April on 60-day terms, so it is paid for in May and June. Seasonal warehouse and installation staff join in June. Schools pay on 45 to 60 days, so August and September invoices convert in October and November. The VAT quarter ending 30 September becomes payable on 7 November, largely on invoices not yet collected.

Sales say the trough is February. Cash says otherwise. The lowest projected point is mid-June, after the stock has been paid for and before a single term invoice has converted. The revenue trough is a comfortable month; the cash trough is four months earlier, and it is the one that needs funding.

Now apply the downside. If the peak comes in 12% short and two weeks late, the June low barely moves, because those costs were committed in April. What moves is the recovery: collections slide from October into November, the November VAT payment lands into a thinner balance, and a second low appears in December that was never the binding constraint in the base case. That second low is the number the facility has to cover.

Getting through the trough

Fund the gap in cost order, cheapest first, and be honest that the last option is the one most of the internet is written to sell you.

A reserve built from peak cash is cheapest and hardest to hold to. Phasing discretionary spend away from the trough comes next, with the caveat that cutting marketing or maintenance can damage the peak you are protecting. Customer terms are third: deposits on large peak orders pull cash forward at some cost to competitiveness. Supplier terms are fourth and often free, though they spend goodwill and need agreeing rather than assuming.

External finance is a legitimate answer, not a failure. An overdraft flexes with the pattern and charges on what you draw, but it is reviewable. A revolving facility gives more certainty across a multi-month gap. Invoice finance suits a trough that is genuinely cash tied up in B2B receivables, and does nothing for a pre-season build where no invoices exist yet. In the US, the SBA Seasonal CAPLine is worth knowing for its restriction: it funds seasonal increases in receivables, inventory and associated labour, so it finances the build-up to a season rather than survival through the quiet months, which is the opposite of what most seasonal-finance marketing implies.

Two things are worth saying plainly. The material you find on this topic is published overwhelmingly by banks, brokers, invoice financiers and lenders; the mechanics are usually accurate, but the framing tends to resolve every seasonal trough into a borrowing decision. And arrange whatever you need before the build-up begins, when your statements look their best. Waiting until revenue is already falling is when terms are worst.

If you are presenting to a lender, show the pattern rather than hiding it: monthly detail so the dip is visible, a 12-month forecast covering the full cycle, the downside case alongside the base, an assumptions log, and how the seasonal borrowing repays from converted receivables. After each peak, compare what you forecast against what happened and record both the amount error and the timing error, so next year's downside case is evidence rather than a guess. That is also when an early warning routine becomes useful, because you know which indicators moved first.

How Float fits

Float is cash flow forecasting software for finance teams running this kind of cycle. It pulls bank balances, invoices and bills from Xero or QuickBooks Online and keeps the forecast current with your accounting data, syncing every 24 hours with a manual sync available. That is current-on-sync accounting data rather than intraday bank feeds, which is the right basis for a model built on invoices and commitments.

Monthly and 13-week views run on the same data, so the seasonal shape and the tightest week come from one set of numbers. Expected payment dates sit on individual invoices and bills, which is what lets you model collection lag instead of assuming same-period conversion. Scenarios layer on top of the base forecast, so a smaller peak, a later peak and a slower collection cycle can be compared side by side without rebuilding anything. Float surfaces the lowest projected cash balance across the forecast window and the date you reach it, which is the peak funding number this page has been working towards. Multi-entity consolidation covers groups whose entities have different seasons, and forecasts export to CSV or PDF for a board or lender pack.

Float does not calculate your seasonal index for you. You derive the pattern from your own history using the method above; the software holds it, keeps it current as the data changes, and lets you stress-test it quickly enough that you actually do it. Xero and QuickBooks Online are live today; Sage Intacct is in development with a waitlist. Float for finance teams covers how teams of three to six set this up.

Frequently asked questions

How can I forecast cash flow for seasonal business variations?

Build the seasonal shape from at least three years of your own history, de-trended with a 12-month moving average so growth is not mistaken for seasonality. Then model cash timing separately from volume, because committed costs precede the peak and collections lag it. Run a rolling 12-month monthly forecast alongside a rolling 13-week weekly forecast, and read the lowest projected cash point across the cycle as your funding requirement.

How many years of history do I need to identify a seasonal pattern?

Three years is a realistic floor and five is better. The ONS works to a three-year minimum for identifying a changed seasonal pattern, US Census Bureau tooling requires three years and often more than four, and IMF guidance treats five as the general minimum. Two years gives a single year-on-year comparison, which is a placeholder rather than an established pattern.

Why is my cash trough in a different month from my revenue trough?

Because the costs of serving a peak are committed and paid before the peak's revenue converts to cash. Stock, deposits and seasonal payroll go out first, and invoices raised during the peak then sit inside their payment terms. In many seasonal businesses the lowest cash point falls shortly before the peak rather than during the quiet season.

How do I separate seasonality from growth in the same data?

Remove the trend before reading the seasonal shape. Calculate a 12-month moving average and express each month as a ratio to it, which cancels the annual cycle and leaves the trend. Comparing like month with like month year on year achieves the same separation differently. Applying a growth rate to individual months distorts the seasonal shape rather than preserving it.

What percentage should I use for a downside seasonal scenario?

There is no evidenced standard percentage, and the round numbers in circulation are asserted without stated basis. Size the downside from your own history instead: reconstruct what you forecast at the same point before previous peaks, and measure the amount error and the timing error separately. Then add a reverse case that solves for the shortfall and delay that would reach your minimum cash floor.

Do VAT and GST deadlines change for a seasonal business?

Not in any way that helps at this size. In the UK the standard regime is invoice-based quarterly VAT, so output VAT on peak invoices falls due before customers pay, and the Cash Accounting Scheme smaller businesses use is unavailable above £1.6m of VAT taxable turnover. Quarterly BAS in Australia and two-monthly GST in New Zealand are the normal regimes at this size. Model each as a dated cash event rather than a monthly accrual.

What do lenders want to see in a seasonal cash flow forecast?

Monthly detail rather than annual totals, so the dip is visible, across a 12-month forecast covering a full cycle. Add the downside and timing-delay scenarios alongside the base case, an assumptions log, and an explanation of how the seasonal borrowing repays from converted receivables. Showing the trough as modelled and funded reads better than a forecast with no dips in it.

How is our forecast data protected if we connect it to our accounting software?

Float retrieves data through your accounting platform rather than connecting directly to your bank, and the connection is one-way and read-only, so your accounting records cannot be changed. Role-based permissions control who can view, edit or manage the forecast, with Admin, Editor and Viewer roles, and two-factor authentication is mandatory for Xero users.

What does cash flow forecasting software cost for a finance team?

Pricing generally scales with the number of entities and users, which matters for seasonal groups running several trading entities. Float's current plans and what each includes are on the pricing page, and every plan starts with a free trial.

A seasonal business does not have a cash flow problem. It has a timing problem that becomes a cash flow problem when the forecast averages it away. Build one full cycle on your own numbers, find the low point, and test it against a peak that disappoints. Float's plans and a free trial are on the pricing page, so you can model your next season before committing to it.