How to build a 13-week cash flow forecast from books

To build a 13-week cash flow forecast, start from the reconciled bank balance as of Monday, then walk thirteen weekly columns forward: cash in from open invoices by each customer's actual payment timing, cash out from open bills by due date plus payroll, rent, loans, taxes, and the owner's draw on their dates, and an ending balance each week against a minimum the owner picked. From closed books it takes under an hour the first time and about ten minutes a week to roll forward.

We run an accounting firm, and the forecast is the first CFO deliverable our clients renewed without being asked. We wrote about why in the 13-week cash flow forecast is the CFO service clients renew. This is the build guide: the inputs, the grid, the weekly update, and the price.

Why it's the first CFO deliverable clients renew

The owner already asks the question every Monday: can we make payroll on the 15th, and what if the big customer pays late? A deliverable that answers the exact question the client is already asking gets opened. A KPI dashboard answers questions the client hasn't asked yet.

The forecast also fails visibly, in a useful way. The moment it's printed it's slightly wrong, because a customer paid early or a bill arrived. Updating it is a fifteen-minute standing call, and that call is where the advice happens: "if Miller pays late again we skip the equipment purchase; if they pay on time we prepay the line." The firm's hour goes to that conversation, not to the spreadsheet.

Inputs from closed books

Everything the forecast needs already exists in books that close monthly and reconcile weekly. The build is assembly, not invention.

Opening balance. The reconciled book balance of the operating account as of Monday, not the online banking number. Outstanding checks and deposits in transit are already reflected. If there are several accounts, one line each, or one consolidated line with the sweep rules noted.

Cash in: accounts receivable. Every open invoice, placed in the week it will actually be paid. Not the due date. The customer who has paid at 47 days for the last eight invoices gets forecast at 47 days, and the AR aging plus payment history gives that number per customer. Recurring revenue that isn't invoiced yet (the monthly retainer, the subscription) goes in on its usual date. Nothing else. No hoped-for sales, no "the proposal should close."

Cash in: other. Loan draws already approved, a tax refund with a date, an owner contribution the owner has committed to. Each with a source and a date, or it doesn't go in.

Cash out: fixed. Payroll on its dates, gross plus employer taxes, from the payroll calendar. Rent on the first. Loan payments from the loan sub-ledger, on their dates. Insurance, software, the owner's draw. These come from the recurring entries the close already maintains.

Cash out: accounts payable. Every open bill in the week it will be paid, which for most small businesses is the due date or the next pay run after it.

Cash out: variable. Card spend, fuel, supplies, the things that don't arrive as bills. A run-rate line from the last eight weeks of actuals, by category if the client wants it, one line if not.

Cash out: known one-offs. This is where the firm earns the fee, because the owner has forgotten two of them. The fourth-quarter estimated tax payment (January 15, 2027 for individuals; December 15, 2026 for calendar-year C corporations). The annual insurance premium. The equipment deposit. The balloon payment. The sales tax remittance. The year-end bonus run. Each on its date.

Building the grid and the assumptions column

The grid is thirteen columns, one per week, starting with the current week. Rows in this order:

1. Opening balance
2. Cash in, by source (AR collections, recurring revenue, other)
3. Total cash in
4. Cash out, fixed (payroll, rent, loans, insurance, draw)
5. Cash out, AP
6. Cash out, variable run-rate
7. Cash out, one-offs
8. Total cash out
9. Net cash for the week
10. Ending balance
11. Minimum balance (the owner's number, usually one payroll)
12. Cushion: ending balance minus minimum

Any week where the cushion is negative gets highlighted. That highlight is the advisory conversation.

To the right of the grid, an assumptions column. Every number that isn't a fact gets a sentence: "Miller pays at 47 days per history." "Card run-rate is the eight-week average of $6,200." "Line of credit draw of $25,000 assumed in week 6 if cushion is negative." The assumptions column is what makes the forecast defensible when the owner asks why week 9 looks bad, and it's what the weekly call updates.

The first build, from closed books, takes under an hour for a single-entity client: twenty minutes pulling the AR and AP and the recurring items, twenty minutes placing the one-offs, twenty minutes with the owner confirming the minimum and the assumptions. Multi-entity clients take longer because intercompany transfers have to be placed on both sides.

The weekly ten-minute update

Every Monday, before the call:

• Replace the opening balance with the actual reconciled balance.
• Drop the week that just ended; add week 14 at the far right.
• Move any invoice that was paid to actual; re-date any that wasn't per the customer's pattern.
• Add new bills; drop paid ones.
• Update the one-offs that got dates or amounts.
• Re-check the cushion row.

Ten minutes when the books are current to last week and the feed is reconciled. Two hours when they aren't, which is why the forecast can't be sold to a client whose close is late. Fix the close first.

The Tuesday call is fifteen minutes. The agenda is the highlighted weeks and what changed. Over a quarter, the owner starts to see patterns: the customer who is always late, the month where the card spend spikes, the payroll that lands the same week as the estimate. Those patterns become the decisions the quarterly meeting is built around.

Pricing it as a tier on the close

The forecast sells as a named monthly tier on top of the close, never as an hourly project. Our range has held at roughly 1.5 to 2.5 times the client's bookkeeping fee for the forecast plus the weekly call, with the top of the range for clients with real complexity: multiple entities, seasonal revenue, a line of credit to manage. A client paying $900 a month for the close pays $1,400 to $2,200 with the forecast.

The setup, the first build with the owner in the room, is billed once, separately. The monthly fee then covers the Monday update, the Tuesday call, and the highlighted-week conversation. Clients don't push back on that price when the grid arrives every Monday. They push back on "advisory" with nothing to show.

Where the platform comes in

Woven is our product, and the forecast is one of the report packages it produces off live books. The reconciled opening balance comes from statement-proven reconciliation, receivables are aged by each customer's actual payment history, recurring outflows post on their dates from the same recurring entries the close maintains, loan payments come from the loan sub-ledger, and the minimum-balance flag is a setting per client. The firm's hour goes to the Tuesday call rather than to assembling the grid. To see the forecast running as a scheduled package, book a demo.