13-Week Cash Flow Forecast Builder

Why thirteen weeks

A quarter, in weeks. Long enough to see a problem while you can still do something about it, which for cash means roughly a quarter: time to accelerate collections, delay a payment run, draw on a line, or go and ask for money. Short enough that weekly detail is credible, because you can reasonably say what lands in week 6 and you cannot say what lands in month 9.

It is the standard instrument in restructuring, in lender reporting and in any company that has become cash-aware, which is why a lender, a sponsor or an incoming CFO will ask for one by name.

Direct, not indirect

Two methods, and the distinction matters because the one in your accounting system is the wrong one for this job.

Indirect starts at net income and adjusts: add back depreciation, move working capital, arrive at cash from operations. It reconciles to the P&L and it is what appears in a cash flow statement. It cannot tell you what happens in week 6, because it was never built by week.

Direct lists what actually lands in and leaves the bank account, dated. Receipts from customers, payroll out, vendors out, debt service out. No accruals, no non-cash items, no depreciation, because depreciation never leaves a bank account.

That is the whole model. Everything else is the care taken over the timing of each line.

The three lines that are wrong in most forecasts

Collections. Companies forecast collections from revenue, which assumes customers pay. They pay late, in a pattern that is specific to your business and visible in your own aging. Start from the aging and a collection curve per bucket, then correct the curve from variance every week. After six weeks your curve is better than any benchmark.

Payroll. A monthly payroll cost divided by 4.33 produces a smooth weekly number that is wrong in every week. Biweekly payroll means twenty-six runs a year: two months carry three runs, and a thirteen-week window holds six or seven, never the six and a half a monthly budget implies. A forecast that misses one is wrong by a full payroll at exactly the point it mattered. Place payroll on its real dates, with employer tax and benefit remittances on theirs.

The annual and quarterly items. Insurance renewal, the software contract, the quarterly estimated tax payment, the audit fee, the property tax instalment. None of them is in a typical week and all of them are in some week. The ones that break a forecast are the ones nobody thought about because they happen once.

Liquidity is not cash

A company with $400,000 of cash and $3 million of undrawn revolver is in a different position from one with $400,000 and no line, and the closing cash row cannot tell them apart. Forecast availability alongside cash, and where the line is on a borrowing base, forecast the base too, because availability falls exactly when receivables fall, which is exactly when you need it.

The variance loop

The forecast is not the deliverable. The loop is.

Explaining the variance is the step people skip and it is the step that does the work. A collection that came in late because the customer disputes an invoice is a different problem from one that came in late because your terms are Net 60 and you forecast Net 30. The first is a conversation with a customer. The second is a wrong assumption that will keep being wrong until someone changes it.

What good looks like

Not accuracy in week 13. Nobody is accurate in week 13. A thirteen-week forecast is doing its job when weeks 1 to 4 are close, the direction in weeks 5 to 13 is right, and the low point does not move much between one week's version and the next. If the low point moves materially every week, the model is not yet describing the business, and the variance view is where you find out why.

The model the tool loads, and the fixture its tests assert every figure in this section against. Invented.

Calder Industrial Services, a 180-person field services company. Revenue is lumpy, customers are large and pay slowly, payroll is biweekly, and there is a $3 million revolver with a $500,000 minimum cash covenant. The CFO sends this to the bank every Monday. Weeks run Monday to Sunday.

Opening position, week 1 beginning Monday October 12, 2026:

October 12 is Columbus Day, a bank holiday, so nothing settles on the Monday of week 1.

Receipts. A customer-level aging of $5.07 million, pulled on Friday October 9, so it collects from the Monday after. Collection curves per bucket, from 95% of the current bucket over weeks 2 to 7 down to 15% of the over-90 bucket. Two customers are forecast by name instead, because between them they are $1.8 million of the book and the CFO knows when their AP runs are: Castellan Foods for $900,000 on November 24, and Halverson Machine Works for $900,000 on December 17. Both are found by name in the aging and taken out of it first, so neither is counted twice. New billing of $860,000 a week collects seven weeks later at 98%.

Disbursements. Biweekly payroll of $360,000 net plus $145,000 of taxes per run, Fridays from October 16. That is seven runs in this window, not six and a half: October 16 and 30, November 13 and 27, December 11, December 24, and January 8. October 30 is October's third payroll, and the grid marks it. December 25 is a bank holiday, so that run goes out on Thursday the 24th. Vendor payments from the AP aging go out on a Thursday payment run, and new purchases of $250,000 a week are paid five weeks later. Then the items that are not AP: rent, a fleet lease, sales tax, the term loan's principal and interest, an insurance premium, an annual software renewal, the December estimated tax payment, and $540,000 of service trucks already committed for Thanksgiving week.

What the grid showed.

Week 7 closes at $406,592.80, under the $500,000 covenant. It carries a payroll, the term loan, and the $540,000 of trucks in Thanksgiving week, against the thinnest week of aging collections. It is the only breach in the quarter, and by week 13 closing cash is $2,142,600.00 as new billing starts to collect.

What the slip test said. Week 7 only closes positive because Castellan's $900,000 lands in it. If Castellan pays a week late, week 7 closes at -$493,407.20: not a covenant breach, an overdraft.

What they did with five weeks of notice. Asked the truck dealer to deliver and invoice in week 8, which is a decision rather than a crisis. It lifts week 7 back over the minimum on its own, and no other week breaches: the new low point is $636,883.32 in week 9, because cash from week 8 on is unchanged by when within it the trucks are paid. Called Castellan's AP team to confirm the 24th. Left the revolver alone, because a forecast that fixed the breach by drawing on the line would have hidden it from the person who has to decide.

What the variance view is for, from week 2. Once actuals go in, any line that misses the same way three weeks running by 10% or more is named. If the 31 to 60 bucket collects well under its curve week after week, the tool says the curve is wrong, not unlucky, and the weeks ahead change when the curve does.

This builds a forecast from assumptions you supply. It does not connect to your bank, your accounting system or your billing system, it cannot see whether a customer actually intends to pay, and it does not know about anything you have not entered. It models one currency and does not convert. It is not a borrowing base calculation, a covenant compliance certificate, or a liquidity forecast in the regulatory sense. It produces no accounting entries and reconciles to nothing. A forecast is a set of assumptions arranged in columns, and it is exactly as good as those assumptions.