How to build a 13-week cash flow forecast (and the three numbers to watch)

Most founders I speak with can tell me their revenue for the quarter but not the week they run out of money. That gap is where otherwise good businesses die. A thirteen-week cash flow forecast closes it, and the first version takes about two hours to build.

Why thirteen weeks, and not twelve months

A twelve-month forecast is a fundraising document. A thirteen-week cash flow forecast is an operating document. The difference matters because the decisions that save a company are weekly decisions: which invoice to chase this morning, which hire to delay by a month, whether payroll on the 25th actually clears.

Thirteen weeks is one quarter. It is long enough to see a problem while you can still act on it, and short enough that you are forecasting things you genuinely know: contracts signed, invoices raised, rent due, payroll dates, the VAT bill. Beyond a quarter you are guessing, and a guess dressed up as a forecast is more dangerous than no forecast, because you will make commitments against it.

The pressure is real and it is current. In the Federal Reserve Banks' 2025 Report on Employer Firms, 48% of employer firms said paying operating expenses was a financial challenge in the prior twelve months, and 36% reported uneven cash flow, a category that includes trouble collecting receivables.

Revenue tells you whether the business works. Cash tells you whether it survives long enough to find out.

The six rows that do all the work

Open a spreadsheet. Make thirteen columns, one per week, labelled by week commencing. Then build six rows and nothing else:

• Opening bank balance. The real figure from your bank this morning, every account combined, converted into one currency.
• Cash in, contracted. Invoices already raised, placed in the week you expect the money to land, not the week you sent them.
• Cash in, expected. Work verbally agreed but not yet invoiced. Keep this row separate and do not probability-weight it. Either you believe it lands that week or you move it out a week.
• Cash out, fixed. Payroll, rent, software, loan repayments, tax. These have known dates. Use the actual dates.
• Cash out, variable. Contractors, media spend, travel, supplier invoices, bank charges.
• Closing balance. Opening, plus in, minus out. It becomes the next week's opening.

That is the whole model. No depreciation, no accruals, no revenue recognition. A cash forecast asks one question thirteen times: does the money leave the account before it arrives?

Keeping contracted and expected apart

This single separation is what makes the forecast honest. When both rows are merged, optimism leaks into the closing balance and you stop trusting the file within three weeks. Keep them apart and you can read two futures off the same sheet: the one you can prove, and the one you are working towards. I run the fixed-cost decisions, hiring and commitments, off the contracted row only.

Build it from the bank balance, not the P&L

Your profit and loss statement is organised around when value was earned. Your bank account is organised around when money moved. Forecasting cash from the P&L imports a timing error you will not notice until it is a shortfall. Start at the bank instead:

1. Export the last ninety days of transactions from every account as a CSV.
2. Tag each line fixed or variable. Anything that recurs on a date at a predictable amount is fixed, everything else is variable.
3. Take the median weekly variable spend, not the mean. One unusual week of legal fees will distort an average and flatter your forecast.
4. Lay the fixed items into the thirteen weeks on their real calendar dates, including the quarterly and annual ones founders forget: insurance renewals, tax instalments, accountancy fees, domain and tooling renewals.
5. Add contracted inflows with a realistic landing week, based on how that specific customer has actually paid you before, not on the terms printed on the invoice.

That last point is the one most people get wrong. Payment terms describe an intention. Payment history describes a behaviour. Forecast the behaviour.

The inflow row is where forecasts lie

Nearly every cash crisis I have been asked to look at was a collections problem wearing a sales costume. The work was sold, the invoice went out, and then nothing happened for nine weeks because chasing felt awkward.

The structural picture in Europe does not help. Intrum's European Payment Report 2026 puts average B2B payment terms at 63 days, up from 60, with business-to-government terms sitting at 70 days, and the report describes payment discipline as weakening across the EU (summary via SMEunited). If you are selling to larger organisations or to the public sector, assume the gap and price for it rather than being surprised by it.

Build a collections ladder and run it the same way every time, so that chasing stops being a judgement call:

• Day 0. Invoice goes out the day the work is accepted, with a calendar due date on it, not "net 30". Send it to a named person in accounts payable and copy your commercial contact.
• Due date minus 3. A short, friendly note confirming the invoice is in their system and asking whether anything is missing, such as a purchase order number.
• Due date plus 1. A factual reminder. No apology, no explanation of why you need it.
• Due date plus 7. A phone call to accounts payable, then an email summarising what was agreed on that call.
• Due date plus 14. Your commercial contact hears from you directly, and you say plainly what happens next: new work pauses until the account is current.

Two structural fixes beat any amount of chasing. Take a deposit, 30% to 50% before work begins, and invoice on milestones rather than on completion. Both move cash to the left on your forecast without needing a single new customer. If the real problem is that not enough is being sold in the first place, that is a different repair, and I have written about the outbound and qualification side of it in Sales Growth.

Three numbers to read every Monday

Runway in weeks

Count forward to the first week your closing balance crosses zero. If it stays positive for all thirteen weeks, report that no shortfall appears in this forecast window; do not infer an exact runway beyond the period you have modelled. Weeks, not months, because weeks force action. The JPMorgan Chase Institute's study of 600,000 small businesses, Cash is King, found the median small business in that historical sample held 27 days of cash buffer. That was 2016 data, not a current estimate of your business. Use your own weekly forecast to find where your cash risk begins.

Collection lag

Average days between invoice date and cash received, measured over the last ninety days, customer by customer. Watch the trend rather than the absolute number. If your lag moves from 34 days to 45, you have just handed eleven days of working capital to your customers for free, and that shows up in the forecast as a hole six weeks out.

Forecast variance

Each Monday, note what you predicted last week's closing balance would be and what it actually was. Express the difference as a percentage. Under 10% and the forecast is a decision tool. Over 25% and you are reading fiction, so find the row causing it, which is almost always expected inflows.

Four failure modes, and what to do instead

• Building it once. A forecast built in January and opened in April is a historical document. Roll it forward every Monday so it always covers the next thirteen weeks.
• Forecasting revenue instead of cash. A signed contract is not money. Model the week the funds clear, including the payment processor's settlement delay.
• Omitting tax and currency. VAT, payroll taxes and corporation tax instalments sink more small businesses than bad products do. If you trade across borders, as I do between Europe, Africa and the US, hold the forecast in one currency and use a conservative rate rather than today's rate.
• Keeping it to yourself. Whoever can commit the company to spending, including a co-founder or an operations lead, should see the same closing balance you do. Decisions improve when the constraint is visible.

Stewardship is a word I take seriously, and in practice it is unglamorous. It means knowing what has been entrusted to the business, down to the week, before making a promise against it. A founder who cannot say what the balance will be in nine weeks is not being bold, they are being vague.

You cannot steward what you refuse to measure.

Your next two hours

Block two hours this week. Export ninety days of bank transactions, build the six rows across thirteen columns, and write today's real balance in the first cell. Then set a recurring thirty-minute appointment for Monday mornings to roll it forward and record your three numbers. By the fourth Monday you will be making different decisions, because you will be able to see the consequence of each one before you commit to it.

Keep the current version in your data room as well. Investors read a live thirteen-week forecast as evidence that a founder is in control of the business, and it sits alongside the other documents I would have ready before a first meeting, which I have set out in Investor readiness is a data room problem, not a pitch problem.