Weekly cash flow: before vs after with Masterestaurant

When the income statement stops telling you enough?

You close August at a 7% net margin and on September 5 there is no money for payroll: that is the figure exposing the limit of any income statement.

Nobody miscounted. Two clocks run at once, because the P&L is built on accrual, which books revenue the moment it is invoiced, while the bank moves on cash, and cash lands when it lands. In the gap between them sit the 45 days of credit your meat supplier granted, the rent deposit, the sales tax you collected three weeks ago and already rolled into purchases, and a payroll that will not wait one day. The U.S. Bureau of Labor Statistics puts independent restaurants that never see a fifth year at 45%; in most of those closures, the last quarterly P&L was printing black ink.

What the thirteen-week board actually is?

Thirteen weeks is the horizon where something can still be moved: renegotiate a term, push an oven purchase back a month, run a promotion on alcohol, a category 46% of operators surveyed by Technomic rank among the highest-margin on the menu.

The board carries no double entry and replaces no accountant. Two columns, that is all: money that truly LANDED in the register and the gateway, and money with a committed date of departure. Every Monday you close the previous seven days, already collected, and out comes one figure per week: 4,200 USD of slack for the 14th, 900 for the 21st, negative on the 28th if the corporate event goes uncollected. Ignore that negative and watch it compound: overdraft on the 28th, credit pulled in October, cash-and-carry prices in November.

Option 1: your own spreadsheet, for the single-location owner

One location, monthly sales under 60,000 USD and the patience to sit down for forty minutes each Monday: on that profile the homemade spreadsheet still wins. Zero in licenses, roughly four hours to build it, ninety minutes a week of discipline, and that is the entire switching cost. You understand every cell because you wrote it, and that understanding outweighs any handsome dashboard the day you must decide whether to postpone a protein order. The downside is real: nothing gets validated, pasting the bank statement introduces silent errors, and the file dies the month you fall ill. I recommend it as almost everyone's FIRST board, under one hard condition: inflows come from the bank statement and the gateway settlement, never from the POS sales report, which shows billing rather than spendable money.

Option 2: the treasury module inside your accounting software

Cross two locations or 150,000 USD in monthly sales and the arithmetic changes: the treasury module inside your accounting software (QuickBooks, Xero, Siigo, Alegra) buys two things a sheet cannot give you, automatic bank reconciliation and a history that survives turnover in the back office. Thirty to 90 USD a month, depending on plan. The real cost hides in the eight to twelve hours of mapping expense categories and linking accounts, never in the invoice. Then comes the deeper catch: these modules project by extrapolating the past, so a house where December and February live on different planets gets forecasts that look scientific and are not. Use them to record the past without a crack. Write the thirteen-week projection yourself, over events only you know: the town festival, the double payroll week and that dead maintenance stretch.

Option 3: AI cash forecasting wired into the POS

The model forecasts SALES, not collection. Keep that sentence handy before signing anything. The newest route wires your POS into an engine that learns from your history and projects daily revenue, and when it works it works well; TimeForge reports accuracy above 90% and labor cost cuts of 8 to 12% in AI-assisted scheduling, the same family of models. It fits you if you run three locations or more, hold eighteen months of clean data and employ someone who can read it. Switching is expensive: integration, catalog migration, ninety days before the model stops missing badly. And when aggregators move 35% of your revenue, in a market where 37% of adults order delivery weekly per UpMenu, fourteen days and a commission of 22 to 30% stand between forecast sales and spendable cash.

Sequence matters more than the tool

Buying software and expecting a miracle is the most expensive mistake on this list, because no board invents the money your structure is eating. Once prime cost, food cost plus loaded labor, crosses 65% of sales, the weekly flow gives you the DATE the air runs out and nothing else; valuable information, not a cure. Diego F. Reverse that order and you have bought a thermometer to cure a fever.

Input inflation changed the math on your buffer

Two weeks of operating expense as a cushion was a prudent rule through 2019. Today it falls short, and the cost data explains why. The producer price index for all food in the United States closed May 2026 some 35% above its February 2020 level, per USDA ERS with BLS data; the producer index for services rose 3.2% in 2025 against 2.5% for goods, so maintenance, insurance and software push from their own side. In Colombia, ACODRÉS reported menu prices up 9.8% since February 2025 to sustain 98,000 jobs. When inputs climb faster than you can reprice the menu without scaring off traffic, slack stops being a conservative owner's luxury: it is what lets you buy well instead of buying cornered.

When NOT to switch?

Stay where you are if the board you already keep, however crude it looks, warns you about the gaps three weeks out and you act on the warning.

That is the whole job of a cash flow. An ugly sheet filled in every Monday beats an elegant platform nobody has fed since March. Migration costs: eight to twenty hours of administrative work, two or three months with data living in two places, and a very concrete risk of losing the habit during the move. Do not switch under a vendor's pressure in high season, nor the month you open the second location, when your attention belongs on the floor. The legitimate signal is different: projection error above 15% two months running, or somebody other than you filling in the board.

Four differences that move the result

UNIT OF TIME, to begin with. A month is an accounting convention that matches no real cycle in this business: payroll lands biweekly, the protein supplier bills at 30 days, the gateway settles at 3, delivery at 14. Measure in weeks and the overlap stops hiding inside an average and shows up as concrete peaks, and a peak with a date on it can be moved. Owners who project thirteen weeks usually discover their problem was never profitability. It was the calendar, and calendars get negotiated. The NATURE OF THE DATA shifts too. Your P&L books revenue the day the invoice goes out; cash flow books it the day the money is spendable. In a house running 35% of sales through aggregators, that distance measures two weeks plus a 22 to 30% commission that never lands at all.

Four differences that move the result — in practice

Model real net by channel, rather than the gross figure on the POS, and you sidestep the classic error: feeling rich on a packed Thursday. Then there is CONTROL GRANULARITY. Comparing theoretical against actual consumption every seven days leaves a window narrow enough to fix things: a mis-portioned recipe, waste in cold prep, a kilo that went up without notice. Stretch that comparison to thirty days and the variance is no longer variance. It is a loss, and losses only get documented. The fourth one nobody discusses: OWNER PSYCHOLOGY. Holding a projected headroom figure, you negotiate differently with the landlord, with the bank and with the partner pushing for a second location, because you stop asking and start proposing with dates attached. That change of posture, measured in overdrafts never requested and terms actually won, is worth more than the two EBITDA points the exercise usually gives back.