पाठशाला Pathshala · संचालन Sanchālan, Operations · Lesson 22 · Build

The 13-week cash forecast

Runway tells you how many months the money lasts on average. The thirteen-week forecast tells you which Friday it runs out. Built from the bank, rolled every Monday, owned by one person.

Pathshala, The Founder Library · 11 October 2026 · 7 min read

Fishing boats resting on the mudflats of Granville harbour at low tide. — Photograph: SlimMars 13 · Pexels

Companies rarely run out of money on average. They run out on a particular Friday, when salaries are due and a customer who always paid on the fortieth day has not paid by the sixtieth. Runway measured in months cannot see that Friday. A forecast measured in weeks can, and usually two months before it arrives.

This lesson builds the thirteen-week cash forecast: what it is and why thirteen weeks, where each line comes from, the Indian dates that make some weeks heavier than others, how to roll it every Monday, and what to do when it shows a hole. It assumes the company already computes [runway](/library/runway-how-many-months-you-really-have) and closes its books monthly; this is the instrument that sits underneath both.

Why thirteen weeks, and why weekly

Thirteen weeks is a quarter, long enough to see the next three payrolls and short enough that most of what is in it is already known. The invoices that will be collected in the next thirteen weeks have mostly been raised; the salaries are fixed; the rent, the statutory dues and the large vendor payments are on contracts. A forecast this short is mostly fact with a little judgement, which is why it can be trusted in a way an annual plan cannot.

Weekly matters because cash is lumpy inside a month. A company that collects ₹45 lakh in a month and pays ₹45 lakh in salaries on the last working day can be comfortably positive on a monthly view and short by ₹20 lakh on the twenty-eighth, because most of the collections arrive in the second half and the salaries go out before them. The monthly MIS will show that month as break-even. The weekly forecast shows the hole.

It is also a different question from the one Paul Graham asks in Default Alive or Default Dead?, where a startup is default alive if, holding expenses constant and growing revenue at its recent rate, it reaches profitability on the money it has left. That is a question about the shape of the company over years. The thirteen-week forecast asks whether this company can meet its obligations on each of the next thirteen Fridays. A company can be default alive and still miss a payroll in week nine.

Building it from the bank, not the P&L

The forecast starts from one number, today’s reconciled bank balance across every account, and only cash moves it. Revenue recognised, accruals and depreciation have no place here. Build it in one sheet with thirteen weekly columns, Monday to Sunday, and four blocks of rows.

Brown envelopes stacked in neat rows. — Every open invoice is a row in the forecast, placed in the week its customer actually pays rather than the week it falls due. Photograph: serdar barış · Pexels

Receipts. Not billing divided by four. List every open invoice from the receivables ledger by customer and place it in the week that customer actually pays, judged by their last three payments, not by the due date on the invoice. A customer on thirty-day terms who has paid on day fifty-two, fifty-five and forty-eight goes in at day fifty-two. Add invoices still to be raised in the window, again placed by the payer’s habit. Large customers get their own rows; the long tail can be one row at its historical average.

Payroll and statutory dues. Net salaries in the week they are paid. Tax deducted at source on salaries and on vendor payments in the week it is deposited: for a non-government deductor, within seven days of the end of the month and by 30 April for March, under Rule 218 of the Income-tax Rules 2026. Provident fund, ESI, GST and professional tax each on their own monthly date, taken from the [compliance calendar](/library/compliance-calendar-private-limited-company). These are the outflows that cannot slip, and they cluster in the first three weeks of each month.

Operating payments. Rent, cloud, software and the payables ledger, each in the week it will be paid. Payments to micro and small vendors go in no later than forty-five days from acceptance, because paying them later costs the deduction under section 37(2)(g) of the Income-tax Act 2025, the provision that was section 43B(h) of the 1961 Act, as the Finance Ministry restated in a Rajya Sabha reply of 21 July 2026. Lumpy items last: advance tax instalments, annual insurance, a security deposit, a hardware order, loan repayments.

Close each column with the closing balance and a buffer, the minimum the company will not go below. Half a month’s payroll is a sensible floor for a company of thirty to a hundred people: enough to absorb one late customer without a phone call to an investor.

Reading the forecast

At its defaults, with a fifth of collections slipping by a month, the illustrative company never touches zero, but its lowest week is the last one, a payroll week, about ₹10 lakh above the buffer. Now slip the largest invoice by nine weeks, out of the window. Week thirteen goes below zero: a payroll that cannot be paid, visible from week one. Let the slipped collections rise to thirty per cent and week ten falls under the buffer as well. Nothing about the company’s P&L changed in either case. Only the timing did.

Three things are worth reading every week. The lowest balance and its week, which is the number the founders need. The weeks under the buffer, which is the number that sets the urgency. And the largest single receipt in the window, because a forecast that survives only if one customer pays on time is a forecast that rests on that customer, and someone should be talking to them this week.

A shortfall seen ten weeks out is a collections task. Seen one week out it is a crisis.

Rolling it every Monday

A forecast built once is a document; one rolled weekly is a control. Every Monday morning the owner, the finance head or the founder until there is one, does four things in under an hour. Drop the week that ended and add a new week thirteen. Compare last week’s forecast with the bank: forecast receipts against actual, forecast payments against actual, with every gap above a set amount, say ₹2 lakh, explained in one line. Update the receipts from the receivables ledger and any new invoices. Restate the lowest balance and its week, and send that one line to the founders with the sheet.

The comparison is the part most companies skip and the part that matters most. After eight weeks the forecaster knows which customers pay when they say and which do not, and whether the team is systematically optimistic about receipts. Measure the error: the sum of absolute differences between forecast and actual receipts, as a share of actual receipts, each week. Under ten per cent is a forecast that can be trusted for decisions; above twenty-five per cent it is a wish, and the receipt rows need rebuilding from payment history.

What to do when it shows a hole

The forecast exists to buy time, and the levers it buys time for are slow. In rough order of speed and cost. Collections: call the customers whose invoices are overdue or about to be, starting with the largest, and ask for a date; offer a small discount for payment this week only when the gap is real. Billing: raise invoices on the day work is accepted, not at month end, and move annual customers to upfront billing at renewal. Payments: reschedule large non-statutory payments, never salaries, statutory dues or micro and small vendors. Spend: pause hiring and discretionary spend that has not yet been committed. Capital: a working-capital line or an investor bridge, which takes weeks to arrange and is why the forecast must see the hole early.

Each lever has a lead time, and the forecast tells you which ones are still open. With ten weeks of warning all five are available. With four weeks, collections and payments. With one week, only the phone call nobody wants to make. The forecast does not create cash. It creates the weeks in which the slow levers can still be pulled.

The Monday ritual

Every Monday by eleven: the forecast rolled, last week’s variance explained, the lowest balance and its week sent to the founders in one line. Every week the lowest balance is under the buffer: a named owner for each of the three largest receipts in the window, with a call made before Wednesday. On the fifth working day of each month, when the books close, the forecast is reconciled to the closed bank balance and the [MIS](/library/monthly-close-and-mis-report) cash page is checked against it. Once a quarter, the error rate for the quarter is reviewed and the receipt rows rebuilt from the latest payment history. The whole ritual costs about an hour a week. The payroll it protects costs a great deal more.

Nothing here is legal, tax or investment advice. The company in the figure is illustrative; statutory due dates checked on 10 October 2026 and should be confirmed against your own registrations.