Building a 13-week cash-flow forecast

Why a 13 week cash flow forecast comes first

Owners of multiple businesses rarely get into trouble because they misjudged annual revenue. They get into trouble because a payroll run, a sales tax remittance and a slow-paying customer land in the same week.

The Federal Reserve Small Business Credit Survey 2024, covering 7,653 employer firms, found that 51% report uneven cash flows and 56% struggle to pay operating expenses. Neither is a story about profitability; both are about timing.

The JPMorgan Chase Institute, in "Cash is King", found the median small business holds 27 cash-buffer days and the bottom quartile holds 13. At thirteen days, a two-week delay on one large receipt is an existential event.

What the direct method is, and why it is the right one here

The direct method forecasts cash by listing the receipts and payments you expect, week by week, starting from the bank balance you have today.

The indirect method starts from projected net income and adjusts for non-cash items and working capital. That is right for a three-statement model over a year and wrong for the next quarter, because it hides which week the money is short.

Indirect tells you the group will generate 400,000 of operating cash next quarter. Direct tells you that in week seven one entity is 62,000 short because a progress payment slipped and rent, payroll and a remittance all clear on the same Friday.

Why thirteen weeks

Thirteen weeks is one quarter, the natural unit for the things that break cash.

It covers a full run of payroll dates, three rent cycles, at least one quarterly remittance, and a complete turn of a receivables book on 45 to 60 day terms. Shorter and you are reacting rather than planning. Longer and the weekly lines become invented, because nobody knows which Tuesday in week 34 a customer will pay. It is short enough to grade every Monday.

Opening cash, account by account

The forecast opens with the cash you actually have, per bank account, per entity, reconciled to the bank as of the same day.

This is where most forecasts go wrong. An opening balance taken from the accounting system rather than the bank includes cheques written but not cleared and excludes deposits in transit. If the opening number is wrong by 40,000, every week after it is wrong by 40,000, and the error is invisible because it never changes.

List each account separately and note restricted balances: customer deposits, pledged amounts, covenant minimums.

Receipts, by source rather than by revenue

Forecast receipts by how the money arrives, not by how revenue is recognised. Most groups have four or five real lines: collections on existing invoices, cash and card sales arriving on a settlement lag of one to three days, recurring contracted amounts such as rent or retainers, draws on a line of credit, and everything else.

The highest-leverage habit is forecasting collections from behaviour rather than terms. A customer on net 30 who has paid on day 52 for the last nine invoices pays on day 52. Putting them in week five because the invoice says net 30 is a data error you have chosen to keep. Since the aging report is the input, it helps to understand what DSO is actually telling you first.

Any receipt above 5% of a week's disbursements gets its own line and date.

Disbursements, by category

Disbursements split into the ones you control and the ones you do not.

Payroll is fixed in timing and nearly fixed in amount: actual pay dates, gross plus employer taxes plus benefit remittances, remembering that some months carry an extra pay date. Rent, leases and debt service are also fixed and known. Debt service is the line most often modelled wrongly, because owners enter the monthly payment and forget the balloon, the covenant-driven sweep, or that the date is the 15th rather than the 1st.

Tax remittances sink people. Payroll deposits, sales tax or GST/HST, instalments, annual filings. These are amounts you already collected on someone else's behalf, so they are not discretionary, and they arrive in lumps that a monthly view flattens into invisibility.

Accounts payable is where you have genuine control. If you pay on Thursdays, model a Thursday disbursement from the AP aging plus known upcoming bills. Moving a payables run from week six to week seven is the most common lever a 13 week forecast hands you, and a defined AP cadence is what makes the lever exist. Capital expenditure and distributions get their own lines, always.

This is general information rather than tax advice; remittance mechanics vary by jurisdiction.

The weekly roll

Every week, three things happen in order. Week one becomes actual, taken from the bank rather than from memory. A new week 13 is added at the far end. Then weeks two through twelve are re-forecast using what week one taught you.

That third step is the one people skip, and rolling the window forward without revising the middle produces a forecast that is technically current and substantively stale. The cycle should take 60 to 90 minutes per entity; if it takes four hours, you are modelling office supplies rather than cash.

The variance review that makes the forecast improve

A cash forecast is only useful if you grade it. Each week, compare forecast to actual by line and record why each material variance happened.

The categories are limited: timing, amount, or omission. A receipt that arrived in week three instead of week two is timing. A payroll run 8,000 over is amount. A quarterly insurance premium nobody modelled is omission.

Omissions get fixed permanently and amounts get calibrated. Timing errors matter most, because repeated errors in the same direction mean an assumption about a customer is wrong, and correcting the assumption beats correcting the week.

A worked example: four entities, one roll-up

Consider a group of four. Entity A is a distribution business, Entity B a light manufacturer that sells most of its output through A, Entity C an equipment-leasing entity that leases machinery to B, and Entity D owns the brand and licenses it to both trading businesses. Each gets its own forecast, because each has its own bank account and lender. Here is week six.

The group flow columns are inflated. Of the 415,000 in receipts, 141,000 is intercompany: 84,000 that A pays B for finished goods, 26,000 of lease payments from B to C, and 31,000 of licence fees paid to D. The same 141,000 sits in disbursements, so external receipts are 274,000 and external disbursements 319,000. Closing cash rolls up cleanly because intercompany nets to zero, but the flow lines do not, and a group forecast that sums receipts without eliminating intercompany overstates inbound activity by more than half. That is multi-entity consolidation, one period forward.

The second observation is more urgent. The group closes week six with 187,000, which looks comfortable. Entity C closes with 12,000 and owes a 41,000 instalment on its equipment loan in week nine. The group is fine and one entity is not, and no consolidated view would have shown you that.

The answer is to move cash from D to C, three weeks early, which is enough time to do it properly: a documented intercompany loan with a note and a rate, or a distribution and a capital contribution, recorded in both sets of books. That notice is the difference between a clean entry and a transfer your accountant reclassifies nine months later. Cash is not fungible across legal entities without a transaction, which is why entity-level forecasting is not optional for a group.

Can QuickBooks or Xero do this?

Partially. QuickBooks Online and Xero are strong on the input side. They hold reconciled bank balances, an AR aging with customer-level payment history, an AP aging with due dates, and recurring bill schedules. Both offer a short-term cash view and basic budget-versus-actual reporting, which for a single entity is enough to start.

Neither does the forecast itself well. There is no weekly grid across a 13 week horizon, no versioning to compare this week's forecast against last week's, no variance log, no scenario switching, and for a group, no cross-entity roll-up with intercompany elimination.

So most groups end up in a spreadsheet, which is fine if you treat it as a system. Panko's synthesis of audits of 88 operational spreadsheets found 94% contained at least one error, with an average cell error rate of 5.2%. A forecast with a broken sum in week nine is worse than no forecast, because it is trusted. Lock the formula rows and have a second person tie opening balances to the bank weekly.

The deeper dependency is your close. A direct-method forecast starts from reconciled balances, so if your books close on business day 20 you are forecasting from stale inputs three weeks out of four. APQC's Open Standards Benchmarking across roughly 2,300 organisations puts the median monthly close at 6.4 calendar days, top quartile at 4.8, bottom quartile beyond 10. If your group sits in that bottom quartile, fixing why the close takes so long will do more for the forecast than any modelling change.

Planning is on cruisr's roadmap rather than something it ships today; what cruisr delivers now is current books, a fast close and group-level consolidation on top of QuickBooks Online or Xero, which is the input any forecast depends on.

Where to start

Build it for one entity, this week, in a spreadsheet. Thirteen columns, opening cash from the bank, four or five receipt lines, six or seven disbursement lines, nothing small or variable. Run it for four weeks and grade it every Monday. By week four you will know which lines you understand and which you were guessing at, which is when the rest of the group is worth adding.

If the weak link turns out to be the inputs rather than the model, that is the part cruisr works on. Get in touch.