Mastering Cash Flow: A Step-by-Step Guide to Creating a 13-Week Rolling Forecast

The 13-week cash flow forecast is the most useful document many finance teams do not produce. It answers one question the monthly management accounts cannot: will we have enough money in the bank on any given Friday for the next three months? Profitable businesses fail on that question, and the ones that get through tight periods are usually the ones that could see the pinch coming six weeks out.

This is a practical how-to. It sets out how to structure the forecast, how to run the weekly routine, how to compare it with actuals, how lenders use it, the mistakes that make forecasts useless, and the tools worth considering. If you are an employer who needs someone to build and own the model for you, our page on the interim Financial Controller for cash flow visibility issues covers that brief, including rates and timescales.

Why thirteen weeks

Thirteen weeks is a quarter, expressed in the unit that matters for cash: weeks, not months. The horizon is long enough to see a problem while you can still do something about it (chase a debtor, delay a purchase, arrange facility headroom, talk to a lender before you need to) and short enough that the numbers can be built from actual knowledge rather than assumption. Beyond thirteen weeks, weekly forecasting becomes guesswork dressed as precision.

The rolling part matters equally. Each week you drop the week just completed and add a new week thirteen, so the horizon never shortens and the forecast never lapses into an annual exercise nobody trusts by March. For the longer-range equivalent, see our guide to building a rolling forecast.

Cash is not profit — the reason this exists

Profit is an accounting measure of performance over a period; cash is money in the bank on a date. A business can invoice £200,000 in a strong month, book the profit, and be unable to pay wages because the invoices are on sixty-day terms and the payroll goes out on the 28th. Timing is the whole game, which is why a cash forecast is built from when money actually moves rather than when revenue is recognised. If your management accounts show a healthy month and your bank balance disagrees, the gap between them is working capital, and the 13-week forecast is how you manage it deliberately rather than reactively.

Direct method: receipts and payments, not profit adjustments

There are two ways to forecast cash. The indirect method starts from forecast profit and adjusts for non-cash items and movements in working capital. It suits longer-range, three-statement models where the P&L and balance sheet are being forecast anyway. The direct method lists the actual cash receipts and payments expected in each period: customer receipts, payroll, suppliers, VAT, loan repayments.

A 13-week forecast should use the direct method. It is built from the ledgers and the diary, can be checked line by line against the bank each week, and shows the timing of large items that an indirect model smooths away. The rest of this guide assumes a direct-method build.

Step 1: Start with the bank, not the ledger

Open the forecast with the actual cleared bank balance at the start of week one, across every account. Not the ledger balance, and not the balance including uncleared items: the money genuinely available. Starting from a ledger figure that includes uncleared items introduces an error that carries through the whole forecast. Where the business has a facility, note the headroom separately. Available cash and available funding are different things, and the forecast should show both.

Step 2: Build receipts from the debtor ledger, not from sales

This is where most forecasts go wrong. Cash receipts are not next month’s sales. They are the collection of invoices already raised, plus a realistic estimate of new invoices raised and collected within the window. Take the aged debtor listing and, for each significant invoice, put the cash in the week you genuinely expect it, not the week the terms say it is due. For most customers those are different weeks, and the difference is the largest single source of forecast error.

Two disciplines help. Track each major customer’s actual payment behaviour: a customer who has paid at day 52 for two years will not pay at day 30 because the invoice says so. And split the ledger, modelling large customers individually and estimating the long tail in aggregate from historical collection patterns.

Step 3: Build payments from commitments

Payments divide into three kinds, and their reliability differs sharply.

• Fixed and known: payroll, PAYE and National Insurance, VAT, rent, loan repayments, insurance. These are diarised and near-certain. Enter them on the actual payment dates, including the awkward ones: a quarterly VAT payment landing in week seven changes the picture entirely.
• Committed but variable: supplier payments from the aged creditor listing, entered by expected payment week.
• Discretionary: capital expenditure, recruitment, marketing. These are the payments you can move, which is exactly why they should be shown separately: they are the levers you will pull if a week looks tight.

For the tax lines, take dates from HMRC rather than memory. Electronic PAYE payments are generally due by the 22nd of the month; VAT return payments are usually due one calendar month and seven days after the end of the VAT period; and Corporation Tax timing depends on the company’s size. Remember that employer’s National Insurance rose to 15% in April 2025, so older payroll assumptions may understate the PAYE line.

Step 4: Lay out the thirteen weeks

The structure is simple and should stay simple: opening balance, receipts by category, payments by category, net movement, closing balance, carried forward as the next week’s opening balance across thirteen columns. Add two rows underneath that do most of the useful work: facility headroom (closing balance plus available facility) and lowest point in the week if intra-week timing matters. A business that closes Friday at £40,000 but dips to £5,000 on Wednesday when payroll clears has a Wednesday problem the weekly closing balance conceals.

Keep the categories few enough to maintain. Six receipt lines and a dozen payment lines is plenty for most businesses; a forecast with sixty rows will be abandoned within a month. Show VAT gross within receipts and payments, with the VAT settlement as its own line, because the forecast tracks money moving.

A worked structure

Here is the shape of a simple thirteen-week model: the categories most businesses need and no more. Each column is a week, and the closing balance carries into the next week’s opening.

Line	Source	Notes
Opening bank balance	Cleared bank	Week 1 actual; thereafter prior week closing
Receipts — existing debtors	Aged debtor listing	By expected week, not due date
Receipts — new invoicing	Sales pipeline	Only where raised and collected inside 13 weeks
Receipts — other	Grants, refunds, asset sales	Include VAT refunds and R&D credits
Payments — payroll and PAYE	Payroll calendar	Actual payment dates, including PAYE by the 22nd
Payments — suppliers	Aged creditor listing	By expected payment week
Payments — VAT and tax	Filing calendar	The lumpy items that sink good weeks
Payments — loans and finance	Facility agreements	Capital and interest separately
Payments — discretionary	Management decision	Capex, recruitment, marketing — the levers
Net movement	Calculated	Receipts less payments
Closing bank balance	Calculated	Carries to next week
Facility headroom	Closing + available facility	The number the board should watch

For a more technical walk-through of the model mechanics, including formulas and checks, see our companion guide to building a 13-week cash flow forecast.

Step 5: Make it roll — the weekly routine

The forecast earns its value in the weekly update, not the initial build. Each week, ideally on the same day:

• Reconcile the closing balance for the week just gone to the bank.
• Enter the actual receipts and payments for that week.
• Compare them with the forecast and note the variances.
• Update the following twelve weeks with what you now know from the refreshed debtor and creditor ledgers.
• Add a new week thirteen and circulate the summary to whoever can act on it.

The whole routine should take a competent finance person an hour or two once the model is built. If it takes a day, the model is too complicated and will be quietly dropped.

Variance to actuals: where the learning sits

The variance step is what turns a spreadsheet into a discipline. A forecast that is never compared with actuals never improves. Keep a simple variance table each week: forecast, actual and difference for each line, with a one-line reason for anything material. Then split the reasons into two types. Timing variances (a receipt that arrived a week late) reverse in a later week and mainly tell you about customer behaviour. Permanent variances (a receipt that will not arrive, or a cost that was missed) change the cash position for good and need to flow through the rest of the forecast.

Over a few weeks the pattern of misses becomes the most useful output: a customer always a fortnight late, a category consistently understated. If week two is regularly wrong by a wide margin, the inputs need attention rather than the format.

Step 6: Add scenarios — but only two

Scenario analysis in cash forecasting tends toward elaborate uselessness. Two variants are enough for most businesses. A downside case: your largest customer pays thirty days late, or a major receipt slips a month. Model the specific, plausible thing that would hurt, not a generic percentage haircut. And a stress case: what combination of events would take the business below zero, and in which week? The value of the second is the answer to the follow-up question: what would we do, and how much notice would we need? Our guide to cash flow management in a downturn develops that argument in more depth.

Using the forecast with lenders and covenants

For a business with borrowing, the 13-week forecast is also a lender document. Facility agreements often include liquidity tests, such as minimum cash or minimum headroom, alongside financial covenants like leverage and interest cover tested on quarterly figures. The direct-method forecast answers the liquidity question directly. It does not test EBITDA-based covenants, so pair it with a three-statement forecast that projects the covenant calculations for the next test dates.

Three practices make lender conversations go better. Show the headroom line and lowest point clearly. Raise a projected pinch early: if the model shows a drawdown or a breach risk in week nine, week two is the time to talk, and lenders respond very differently to a forecast presented early than to a request made late. And keep the variance history, because a lender’s confidence in your forecast rests on how accurate the last ones were. Under pressure, banks commonly ask for it weekly, with variance commentary. If HMRC liabilities are the problem, HMRC’s guidance on difficulties paying explains Time to Pay arrangements, and HMRC will expect to see a cash forecast before agreeing one. Our article on covenant and KPI tracking covers the monitoring side.

The mistakes that make forecasts useless

• Forecasting receipts from sales rather than collections: the error that produces optimistic forecasts and unpleasant Fridays.
• Using due dates instead of expected dates: the same error in a different costume.
• Omitting the lumpy items: VAT, Corporation Tax, annual insurance, the bonus payment. These are known months in advance and are exactly what sinks an otherwise sound week.
• Too much detail: a model with a row per customer and per supplier is unmaintainable, and unmaintained forecasts carry false authority.
• No variance review: producing the forecast without asking why last week was wrong guarantees it stays wrong.
• Treating it as a finance document: it should be discussed weekly with whoever can act on it, or it is arithmetic nobody reads.

Who should own it

This matters more than the tooling. In a small business the forecast is typically owned by the Finance Manager or a fractional finance lead. Where a founder is doing it in a spreadsheet at weekends, that is usually the clearest signal the business has outgrown its finance resource. In a mid-sized business it sits with the Financial Controller, who owns working capital alongside the close. Where the business is investor-backed, covenant-reporting or under real cash pressure, the forecast becomes a board-level document: the Finance Director owns its presentation while the FC owns its construction.

In every case, one named person should be accountable for the weekly update; forecasts owned by committee are updated by nobody. If nobody in the team has the time or experience, that is a hiring conversation rather than a software one. A fractional Financial Controller can run the routine one or two days a week, while an interim Financial Controller can build the model, establish the routine and hand over something the team can maintain. Our wider fractional finance page explains the options.

Spreadsheet or software?

For most businesses a well-built spreadsheet is entirely adequate: it is transparent, cheap, and everyone can interrogate it. Dedicated cash forecasting tools, including add-ons that connect to cloud accounting systems and bank feeds, earn their cost when the business has multiple entities or currencies, high transaction volumes that make manual entry impractical, or a need for actuals to populate automatically.

The honest test is how long the weekly update takes. If a competent person can run it in an hour, stay in the spreadsheet; if manual data-gathering has become the bottleneck, automate. Either way, build it so someone other than its author can pick it up. Our piece on AI for cash flow forecasting looks at where newer tools help and where they do not.

Using it: what the forecast is actually for

A forecast that is produced and filed has failed, however accurate it is. The output should drive three conversations. Collections: the debtor rows show which invoices matter this month and who to chase first. Payment timing: knowing which week is tight lets you sequence discretionary payments deliberately rather than paying whoever shouts loudest. Commercial decisions: the hire, the equipment purchase, the marketing push can each be tested against the forecast before commitment. The businesses that get most from a 13-week forecast treat it as a weekly management meeting item, not a finance deliverable.

Frequently asked questions

How is a 13-week forecast different from a three-way forecast?

The 13-week forecast is short-horizon and cash-only. A three-way (three-statement) forecast projects the P&L, balance sheet and cash together over a longer horizon and is a planning tool rather than a liquidity tool. Most businesses need both; our budgeting and reforecasting guide covers the longer view.

Should VAT be shown gross or net?

Gross. Cash forecasting tracks money moving, so receipts and payments include VAT, with the VAT settlement shown as its own payment line.

How accurate should it be?

Week one should be near-exact, weeks two to four close, and the back end directionally right.

Who should see it?

The management team weekly, the board monthly, and lenders or investors where covenants or facilities are in play.

How long does it take to build?

A competent finance professional can build a working model for a straightforward business in two to three days, with a week or two of iteration before it settles.