The 13-Week Cash Flow Forecast for Small Businesses: How to Build It, Roll It, and Where AI Fills In

In a 2016 study of Chase business banking accounts, the JPMorgan Chase Institute found the median small business held 27 days of cash buffer, meaning it could cover 27 days of outflows if every inflow stopped. A quarter of the firms held fewer than 13 days. Restaurants had the thinnest median, at 16. A 2020 follow-up across 25 metro areas found medians between 11 and 18 days. Whatever your number is, it is probably measured in weeks, and a monthly P&L will not warn you in time.

That is the case for the 13-week cash flow forecast. Stampli describes it as a "rolling, weekly-grain, direct-method forecast of cash receipts and disbursements covering one quarter." Each part of that sentence matters, so here is what each means in practice.

Why weekly, why direct, why 13

Weekly because cash problems happen on specific days. Payroll lands on a Friday. Rent is due on the 1st. A big customer pays on the 20th. A monthly view nets those out and hides the Tuesday when your balance dips below zero.

Direct because you are forecasting actual money moving in and out of the bank, line by line, rather than starting from net income and adjusting. Accrual accounting is the right way to measure profit. It is the wrong tool for knowing whether you can make payroll, because revenue gets booked weeks before the cash arrives.

Thirteen weeks because it is one quarter: near enough that the first few weeks can be forecast with real precision, far enough out that you see a shortfall in time to do something about it (draw on a line, delay a purchase, call a customer, talk to your bank).

The layout

One row per cash category, one column per week. Thirteen forecast columns, plus a column for last week's actuals. The structure:

Section	Lines	Where the numbers come from
Beginning cash	Bank balance at the start of the week	Prior week's ending cash (week 1: actual bank balance)
Receipts	Collections on existing AR; collections on new sales; card and processor deposits; other (refunds, asset sales, loan draws)	AR aging by expected pay date; sales pipeline; processor payout schedule
Disbursements	Payroll and payroll taxes; rent; inventory and materials; vendors on AP; insurance; loan payments; taxes; owner draws; capex	Payroll calendar; AP aging by due date; lease and loan schedules; tax calendar
Net cash flow	Receipts minus disbursements	Formula
Ending cash	Beginning cash plus net cash flow	Formula
Minimum cash	The floor you will not go below	Your call: often a payroll plus a cushion
Alert	Flags any week where ending cash is under the minimum	Formula

Receipts: forecast by customer, at least for the big ones

The most common mistake is to forecast receipts as "sales this month divided by four." Do it from the AR aging instead. For each open invoice over a set size, write down the week you expect the cash, based on how that customer actually pays. The terms printed on the invoice are a hope. A customer on net-30 who pays at day 45 goes in week 7 for an invoice dated this week. Small invoices can be grouped and spread using your average collection pattern.

Disbursements: start with the ones you cannot move

Payroll, payroll taxes, rent, loan payments, and sales tax remittances come first. They are fixed in amount and date, and missing them carries penalties. Then add AP by due date from your aging. Then the flexible items: inventory buys, owner draws, capex. When the forecast shows a shortfall, the flexible items are where you act.

Rolling it every week

A 13-week forecast you build once is a spreadsheet. One you roll every week becomes a management tool. The weekly routine, which should take under an hour once the model is set up:

Load last week's actuals. Pull the actual receipts and disbursements from the bank and put them next to what you forecast.
Check variances. Stampli's advice is to "record actuals weekly and track forecast-vs-actual by line" and to "investigate variances over a set threshold." Pick a threshold that fits your size.
Drop the old week, add week 13. As Stampli puts it, "roll the forecast forward every week so week 14 becomes week 13."
Update receipts from the current AR aging. Move any invoice that slipped.
Update disbursements from the current AP aging and any new commitments.
Look at the lowest week. If it is under your minimum, decide what you will do about it this week, while you still have room to move.

The variance review is where the forecast gets better. If you always overestimate receipts from one customer, you will see it within a month, and you can fix the pattern you assume for them.

Where AI fills in actuals, and where it should stay out

The heavy lifting in a weekly roll is loading and categorizing actuals. That is a good job for automation:

Bank categorization. Pulling last week's bank lines and sorting them into your forecast categories (payroll, rent, vendors, customer receipts) is pattern work. Ledger tools already do most of it. A script or an LLM can map those categories to your forecast rows.
Matching receipts to invoices. Telling which deposit paid which invoice updates your AR aging, which updates next week's receipts forecast.
Drafting variance notes. Given forecast and actual by line, a model can write "Receipts $18,000 under forecast: Customer A's payment slipped to next week" if it can see the AR data. A person confirms.
Payment-timing patterns. A model can look at each customer's history and suggest the week a new invoice will likely pay. Treat that as a starting point you can override.

Some ledgers ship a forecast already. Xero's US plans include a cash flow forecast of 30 days on Early, 60 days on Growing, and 180 days on Established. Puzzle lists cash, burn and runway insights on its Core plan. Intuit's Finance Agent, aimed at mid-market firms, includes scenario planning and forecasting. These are useful for a glance. A 13-week model you own is different, because you put in the facts the software cannot see: the customer who told you on the phone they will pay late, the equipment purchase you are planning, the tax payment your CPA estimated.

Where AI should stay out: setting the minimum cash level, deciding which bills to delay, and deciding when to draw on credit. Those are calls with consequences for vendors, staff and lenders, and they belong to the owner.

A note on who needs this most

In the Fed's 2024 Small Business Credit Survey, 51% of employer firms named uneven cash flow as a financial challenge and 56% named paying operating expenses. If your revenue is seasonal, lumpy, or tied to a few large customers, you are in that group, and the weekly forecast earns its hour every Monday. If you have months of cash and steady card revenue, a monthly view may be enough. Be honest about which one you are.

Get the workbook

The 13-Week Cash Flow Forecast ($49) is a working Excel workbook with weekly receipts and disbursements, a rolling actual-versus-forecast view, and a minimum-cash alert that flags any week that dips under your floor.

If you would rather have this built than build it, Agor AI sets up AR, AP and close automations for small finance teams. See Agor AI. Start with the $250 AI Automation Audit, credited toward a build.

This is part of a weekly series on AI for finance and back-office work. Get it by email.

This article is educational and is not accounting, legal or tax advice. Talk to your CPA before changing how you close your books or file.

Sources

Ariel Agor was Director of Finance & Accounting, then VP of Operations, at THC Design, a Los Angeles cannabis cultivator (2018 to 2019), and holds an MBA in Finance. He builds back-office AI agents at Agor AI. This piece was drafted with AI assistance and edited by Ariel. Get the weekly newsletter.