The 13-week cash forecast lenders now expect

The arithmetic is simple

Cash at the end of a week is cash at the start, plus what comes in, minus what goes out. Coming in is mostly receivables collected. Going out is payables paid, payroll, debt service, taxes and rent. Add the items you know are coming that don’t fit the normal pattern — an annual insurance premium, a tax payment, a large equipment deposit, a customer prepayment — and you have the forecast.

None of that is hard to understand. It’s hard to keep right.

Why lenders ask for it now

A 13-week forecast used to be something a lender asked for when a borrower was in trouble. Increasingly it’s a standing expectation for asset-based lines, covenant reporting and renewals, especially when rates are higher and margins are thinner. Thirteen weeks is a quarter: long enough to see a crunch coming, short enough to be built from real invoices and real bills rather than assumptions.

Lenders also compare it. Last month’s forecast against what actually happened is the fastest way for a credit officer to decide whether management knows its own business. A forecast that’s always optimistic is worse than none.

A cash forecast is only as credible as last quarter’s forecast turned out to be.

Why the spreadsheet stopped working

The spreadsheet worked when the business was smaller and the lender was patient. Someone exported open AR and open AP on Monday morning, pasted them into tabs, adjusted the collection dates by feel, typed in payroll and the loan payments, and rolled it forward. It took half a day and it was out of date by Tuesday.

It breaks down in predictable ways:

• Collections don’t follow due dates. Customers pay late, some reliably, some unpredictably. A forecast built on due dates is wrong in a way that always flatters you.
• Most outflows never become a bill. Loan payments, card settlements, bank fees, sweeps and many recurring vendors are paid without an open payable in the system. When we compared one company’s actual weekly outflows from the bank against what open AP predicted, the bank showed far more going out than the payables ever predicted. A forecast built only on open payables is wrong by that gap every single week.
• Payroll is a separate system. The biggest predictable outflow lives in the payroll provider, not the ERP, and gets typed in by hand.
• Opening cash is a guess. The GL balance lags the bank. The real starting point is what the bank says this morning.
• It depends on one person. The spreadsheet lives in somebody’s head as much as in the file. When they’re out, the forecast doesn’t happen.

What a forecast built from the ERP does differently

We replaced the spreadsheet with a forecast that lives inside the ERP and rebuilds itself from live data:

• Opening cash from the bank feed, not the general ledger, with a check that warns when the feed stops arriving or starts looking wrong.
• Receivables by expected collection, using each customer’s actual payment history to move an invoice from its due date to when that customer really pays.
• Payables by due date, plus open purchase orders and sales orders for what’s committed but not yet billed.
• Payroll from the payroll integration, historical runs rolled forward, not retyped.
• A run-rate for everything that never becomes a bill, calibrated against what actually left the bank account, so the forecast stops missing the same money every week.
• Manual entries for the known exceptions — the insurance renewal, the tax payment, the large deposit — dated and labelled, so the thing you know is coming is in the number instead of in someone’s memory.
• Recurring debt service as scheduled entries, kept out of the run-rate so it isn’t counted twice.
• An exception briefing: an AI pass over the week that flags what looks unusual, such as a bill far larger than that vendor’s norm or a large customer suddenly paying slower.

It runs thirteen weeks by default and can stretch further. More importantly, it’s the same numbers every time, from the same sources, whoever opens it.

Accuracy comes from comparison

The feature that matters most is the least glamorous: compare forecast to actual, week by week, and adjust the assumptions that missed. Reconciling a week of bank activity against the forecast is what exposed how much non-bill spending the open-AP view had been missing. Without that loop, a programmatic forecast is just a faster way to be wrong.

Why this matters beyond the lender

The lender is the reason most companies start. The owner and the controller are why they keep going. A forecast you trust changes conversations: when to place the big buy, whether to take the early-pay discount, how hard to push a slow customer, whether the line will be tight in week nine. Those used to be judgment calls made with a half-day-old spreadsheet. Now they’re made looking at the same number the bank will see.

Building it surfaced several quiet NetSuite behaviors that would have produced a confident, wrong forecast without anyone noticing. Those are in the companion piece.

Related: the five bugs that would have made the forecast confidently wrong, and the order shipped, the invoice never came.