The 13-Week Cash Flow Forecast for Restaurant Groups

An eight-unit group in northern New Jersey called on a Thursday in February. The trailing-twelve P&L showed roughly $1.9 million of EBITDA and a four-wall margin north of 19% at their best location. The operating account was $61,000 short of Friday's payroll. Nothing had gone wrong operationally. Sales were up. February simply held three payrolls, the quarterly sales tax remittance, and the annual general liability premium that auto-drafts on the 14th. Three known events, none of them a surprise, all inside the same seven days.

A 13-week cash flow forecast is what a restaurant group builds so that call happens in November instead of February. It doesn't predict the future. It tells you which week the account gets thin, early enough that your remaining options are cheap ones. Here's how the model is structured, how to keep it honest, what order to pull levers in when week nine turns red, and the case for not building one.

What is a 13-week cash flow forecast for a restaurant group?

It's a weekly, direct-method projection of every dollar moving through your bank accounts over the next quarter, built from settlement and due dates rather than accounting periods. Thirteen weeks is one quarter, one covenant test cycle, one sales-tax cycle: long enough to catch the annual items that wreck operators, short enough to forecast individual lines rather than trends.

The word that matters is direct. An indirect statement starts with net income and adjusts for depreciation and working capital swings, which is right for a footnote in your annual financials and useless weekly: it says nothing about which Tuesday the distributor debits your account.

It's also a different instrument from your monthly package. Good multi-unit restaurant financial reporting tells you whether unit four earned its keep last month. This tells you whether the group funds the next 90 days.

Why restaurants specifically break the cash intuition

Restaurants collect revenue faster than almost any business in the economy, and that speed is what causes operators to misread their position. Card batches settle in one to two business days, there's essentially no receivable, and money lands every morning. After a decade of that, most operators develop a gut sense that cash health tracks sales health.

The inflows are smooth. The outflows are not. What lands lumpy:

Biweekly payroll, which produces 26 pay periods a year, so two months every year carry three payrolls

Payroll tax deposits, which for semiweekly depositors follow each payday within days (IRS deposit rules)

Sales tax remittance, monthly in most states, often the largest non-payroll payment a group makes

Insurance premiums, plus the workers' comp audit true-up that arrives after a growth year

Percentage rent reconciliations, billing you for a strong year months after it ends

Franchise royalties, ad fund contributions, debt amortization, and equipment leases

Capex that refuses to wait: a walk-in compressor, a hood system, an HVAC unit in July

Owner tax distributions on K-1 income the owner never held as cash

Receipts have traps too: delivery platforms remit on their own calendar, gift cards redeem without new cash, and the processor keeps 2.5% to 3.5% of every card sale.

The National Restaurant Association's 2026 State of the Restaurant Industry report found that 42% of operators said their restaurant was not profitable, with more than nine in ten citing food, labor, insurance, and swipe fee pressure. Timing risk decides whether you survive a soft quarter or buy expensive money.

The anatomy of a model built the right way

Four rules make it accurate enough to act on inside two months.

Receipts go in at settlement date, net of what the processor keeps

Forecast card receipts by the date money hits the bank, not the date the guest ate, net of blended processing cost and chargebacks. Delivery remittances get their own line because they arrive on a different cadence.

Disbursements go in at due date, not invoice date

Your broadline distributor invoices daily and drafts weekly. Your linen company is net 30 and nobody has ever paid them on day 30. Build the AP line from how your top vendors actually get paid, not from the invoice register. Payroll goes in on pay date, payroll taxes on deposit date.

Periodic items get their own dedicated line

The highest-value row holds what happens four times a year or once a year: sales tax, quarterly estimates, insurance premiums, percentage rent true-ups, franchise fees, workers' comp audits, personal property tax. Load all thirteen weeks of them on day one. That row is where the February call comes from.

Weekly buckets, thirteen columns wide

Not monthly, which hides the intra-month trough where groups get hurt. Not daily, which nobody maintains past week three. Below: four weeks of an eight-unit group running roughly $19 million a year. The live model runs thirteen columns wide on identical rows.

Look at week four. Closing cash of $286,000 is not a small number — a group watching only the bank balance feels fine right up until it doesn't. Headroom is the number that matters, and it's negative. Set the threshold at one full payroll cycle, net pay plus the tax deposit, plus a few days of ordinary AP. This group set theirs at $350,000. Then manage headroom, not the balance.

The rolling discipline that keeps the model from becoming decoration

Re-forecast every Monday morning, without exception, and lock last week's actuals beside what you forecast for it. Drop week one, add week fourteen, and the model stays thirteen weeks deep forever. About 45 minutes of controller time once it's built.

The part almost everyone skips is variance tracking by line. Every Monday, record forecast versus actual for each row, in dollars and percent. Set tolerance bands of roughly 3% on receipts and 5% on disbursements, and require a written one-liner for anything outside them. Do that for six to eight weeks and weeks one through four tighten to about 2%, because you've found where your assumptions were wrong. Weeks nine through thirteen stay near 10%, which is what directional means. Skip the variance step and the model becomes a spreadsheet somebody updates until they stop.

The escalation ladder when week nine goes red

When the forecast shows a trough, work the levers in this order — cheapest and most reversible first.

Verify the trough is real. Re-check the three largest disbursement lines in the red week and the receipt timing feeding it. A surprising share are a double-counted debt payment or a premium in the wrong week.

Work purchasing terms. Moving a broadline account from net 7 to net 14 or 21 on a $19 million group frees roughly $100,000 to $200,000 of permanent working capital, and vendors grant it far more readily eight weeks out than on a Thursday.

Defer capex and discretionary projects. Anything not contracted, not permitted, and not a safety issue moves out of the window, with a written note of when it comes back.

Move owner distributions. Outside the tax portion, distributions are timing, not obligation. Shifting one three weeks is usually the largest lever available and costs nothing but a conversation.

Draw the revolver on purpose. A planned draw, sized to the trough, with a named repayment week, is a treasury decision. A last-minute draw is a signal to your lender.

Open the landlord conversation. Deferral, a temporary shift toward percentage rent, or blending a CAM true-up over months. Works where you're a strong tenant in a center that would struggle to replace you.

Call the lender before the covenant test, not after. Lenders respond well to a group arriving in week three with a model and a plan — very differently to one arriving after a fixed charge coverage ratio has been missed.

Owner capital or a subordinated note. Last: the most expensive flexibility you have, and using it first hides the problem underneath.

Four levers look attractive and are not. Do not stretch payroll tax deposits: withheld employee taxes are trust fund money, penalties are severe, and personal liability can attach to responsible individuals (IRS employment tax due dates). Do not delay sales tax either; many states treat what you collected as money held in trust, so confirm the rule in yours before you ever consider it. Do not take a merchant cash advance; the effective annualized cost frequently runs into triple digits. And do not quietly slow payroll, which costs you your best managers inside ten days.

How the same model answers the covenant question and the new-unit question

One model, three jobs. Because the window is exactly one quarter, add rows below headroom forecasting covenant math at the next test date: trailing-twelve fixed charge coverage, lease-adjusted leverage, minimum liquidity. That belongs in your board and lender package too. The question stops being "will we pass?" and becomes "what has to happen in weeks four through eight so we pass?"

The new-unit question works the same way. Your new-unit pro forma tells you whether unit seven earns an acceptable return over five years, not whether the group survives opening it. Pre-opening burn typically runs eight to fourteen weeks of management payroll, training labor, deposits, and inventory before the first guest pays for anything, then six to ten weeks of ramp. Drop that into the 13-week model and you see the trough the pro forma can't.

When a 13-week model is overkill

Sometimes a cash calendar is enough, and pretending otherwise is just work. If you run one or two units, carry no term debt, have no expansion inside 18 months, and manage nothing for third-party owners, a one-page calendar of known outflows does the job. Mark the three-payroll months, tax remittances, insurance renewals, and debt dates. Keep four weeks of disbursements in the account.

Two other honest limits. If your books close on the 25th of the following month, you can't build a reliable model yet: you don't know your AP position in time to forecast from it. Fix the close first. And if nobody will run it every Monday, don't build it: a forecast updated three times and abandoned is worse than none, because the group starts deciding off a stale one. That gap is one of the clearer signals a group needs real finance capacity.

Where we land

Restaurant groups don't run out of cash because they're unprofitable. They run out because daily settlement makes cash feel abundant while the outflows that threaten them arrive four times a year. A weekly, direct-method, thirteen-column model built on settlement and due dates fixes that mismatch, and the Monday variance review keeps it honest. Manage headroom against a stated minimum, work the levers in order, and never touch trust fund money. If your group carries debt or plans openings, it's the highest-return financial architecture you can build in a month.

Common questions

How is a 13-week cash flow forecast different from a budget?

A budget is annual, accrual-based, and organized by accounting period. A 13-week forecast is weekly, cash-based, and organized by the dates money actually moves. Your budget can be accurate for the year and still miss the week you can't fund payroll. Neither substitutes for the other.

How accurate should a 13-week cash flow forecast be?

After six to eight weeks of variance tracking, expect weeks one through four within about 2% of actual, weeks five through eight within 5%, and weeks nine through thirteen within 10%. The far weeks are directional by design. If near weeks miss by more than 5%, the cause is usually receipt timing or an AP line built from invoice dates.

Who should build and own the model in a restaurant group?

The controller or accounting manager builds and updates it; the CFO or owner reviews it Monday and makes the decisions it surfaces. About 45 minutes a week once the structure exists. If your controller is already underwater on the close, adding this without capacity means it quietly stops by week five.

Can my POS or accounting software just do this for me?

No. Your POS knows sales, not settlement timing net of processor fees. Your accounting system knows invoice dates, not when you actually pay each vendor. Both feed the model, neither replaces it. Judgment about vendor behavior and periodic items has to come from someone who knows your group.

What's a reasonable minimum cash balance for a restaurant group?

Set it at one full payroll cycle, net pay plus the tax deposit, plus a few days of ordinary accounts payable. On a group running roughly $19 million a year that lands near $350,000. Groups carrying term debt or an aggressive opening schedule should set it higher. The number matters less than managing headroom against it.