The 13-Week Cash Flow Forecast: How to Build the Model and Analyse Variances Every Week

Why 13 weeks, and what does a weekly model add?

UK government guidance on preparing funding applications asks for a cash flow forecast covering the next 12 months, with receipts, payments and a running balance at the end of each month. ^[1] That forecast is useful for planning, but it does not answer treasury’s everyday question: is there enough cash for next week’s payroll?

That is where the 13-week model comes in. It is a weekly, direct-method forecast, usually owned by treasury, and a staple for treasury teams, particularly before major events such as closing an acquisition. ^[3] The direct method means presenting the major classes of gross cash receipts and payments, unlike the indirect method, which starts from profit and adjusts it. ^[4]

Thirteen weeks is a practical choice that covers a quarter in weekly detail; it is not a requirement of any accounting standard. What matters is that the horizon is long enough to see the large commitments coming and that it is updated regularly.

Start from the right balance, not the ledger balance

The costliest model errors start on the first line. The opening balance must be the actual bank balance after reconciliation, not the cash account in the ledger, which may include uncleared cheques or transfers that have not yet arrived.

• Reconcile each account separately: one line per bank account, and one line per currency where conversion between them is not immediate.
• Separate cash that is not available: deposits held as security for a letter of guarantee, or balances earmarked for a specific purpose, do not count as cash you can pay from.
• Distinguish cash from cash equivalents: IAS 7 defines cash equivalents as short-term, highly liquid investments that are readily convertible to known amounts of cash and subject to an insignificant risk of changes in value. ^[4] Anything that does not meet that description stays out of the available balance.

How do you build the receipts lines?

The first rule: record a receipt in the week the customer is expected to pay, not on the invoice date, and match the figures to what will actually appear in the bank account. ^[1] A customer who usually pays after 75 days will not change their habits because the contract says 30.

AFP recommends classifying receipts and disbursements by predictability and timing, and basing the forecast on actual bank activity rather than indirect measures such as a percentage of sales. ^[2] In practice:

• Key customers: one line per large customer, with an estimated collection date drawn from their actual payment record.
• Other customers: one aggregated line, using a collection rate derived from previous weeks.
• Non-operating receipts: an asset sale or a tax refund, included only where there is evidence of the date.
• Funding: confirmed funding only, on its own line; probable funding in a separate scenario.

Document the source and date of every assumption, because management will ask about any large change. ^[2]

How do you build the payments lines?

Payments are easier to predict because the organisation controls many of them, but they are also where surprises hide. Rank them from most binding to most flexible:

1. Payroll and statutory dues: in the week they are actually paid.
2. Taxes and zakat: in the weeks they fall due, according to the tax authority’s calendar in your country.
3. Loan instalments and rent: according to repayment schedules and contracts.
4. Suppliers: according to agreed terms and approved invoices, not a general estimate.
5. Capital expenditure: signed commitments only.

This ranking pays off later: if a shortfall appears, you know at once which items can be renegotiated in timing within their terms and which cannot wait.

Hypothetical example: how one late collection changes the picture

A hypothetical Gulf organisation has a reconciled bank balance of AED 80,000 at the start of week 1. That week it expects to collect AED 50,000 from a key customer and to pay AED 100,000 in payroll and supplier dues.

One collection moving to week 2 turned a positive balance of 30,000 into a shortfall of 20,000. Because the model showed this weeks in advance, the team has options: chase the customer early, reschedule supplier payments within their terms, or line up an existing facility before the due date.

So that every positive figure is not read as safety, agree with management a minimum closing balance, calculated for example as a number of weeks of fixed payments. Every week whose forecast balance falls below that floor needs a recorded decision: who follows up, what action, and when it is reviewed.

The weekly cycle: how does the model stay alive?

A model that is not updated loses its value within weeks. AFP describes a practice of pulling actual transactions from bank data every week and categorising them, analysing variances and following up the large ones with the business, alongside quarterly reviews that measure accuracy and diagnose misses. ^[3]

At each close, replace the week just ended with actuals, roll the actual closing balance forward as the next week’s opening balance, then add week 14 at the end, so you always have 13 weeks in front of you.

Give every line an owner who knows where its figures come from: collections with the credit and collections team, payroll with HR, suppliers with accounts payable. Knowing where the numbers originate and how others build them is one of AFP’s recommendations; without it the model becomes one person’s guesswork. ^[2]

Variance analysis: timing or permanent?

AFP recommends validating the forecast through variance analysis, comparing actuals with the forecast and adjusting, and digging into variances and business trends the forecast has not yet captured. ^[2]

The decisive question for every large variance is whether it is a timing issue that will reverse in a later week or a permanent change in the amount. A timing variance moves to its new week; a permanent variance means an assumption is wrong and must be corrected in every future week. Agree in advance on the accuracy measure you will track, as AFP advises defining accuracy measures with stakeholders from the outset. ^[2]

Do not rely on the weekly model alone. One practice AFP describes is comparing the direct model periodically with the earnings-based three-statement model, since the first sees cash through collections and the second through earnings. ^[3] Aligning the short-term forecast with the longer-term plan is also among its ten recommendations. ^[2]

Gulf deadlines that must sit in the right weeks

Return and payment deadlines differ between Gulf countries, so do not apply one country’s date to another:

Put these amounts in the week the deadline falls, and check your tax authority’s calendar every year, because actual dates are announced for each period. ^[6] In other countries, refer to the competent tax authority there.

Conclusion

A 13-week model will not predict the future precisely, but it gives you weeks to act rather than days. Start from a reconciled bank balance, record every amount in the week it actually moves, and make weekly variance analysis a fixed habit that separates timing from permanent change.