13-week cash flow forecasting for tight runways
Key takeaways
- 13 weeks is one quarter: short enough to forecast specific invoices and payments, long enough to act on what you see.
- Forecast collections, not invoices. The single most common error is projecting revenue when the model needs cash arrival dates.
- Start from the bank balance, not the P&L. This is a cash model; profit is a different question.
- The weekly update ritual (drop the finished week, add week 14, correct the assumptions that missed) is what keeps the model alive. Skipping it twice kills it.
- The two outputs that matter: your runway in weeks, and the trough week, the lowest projected balance.
Why 13 weeks
Thirteen weeks is one quarter, and the number isn't arbitrary. It sits at the intersection of two constraints. Shorter horizons (four to six weeks) are accurate but useless: by the time you see a problem, there's no time to fix it. Longer horizons (six to twelve months) give you time but force you into monthly buckets and averaged assumptions, which is exactly the fog a cash crisis hides in. Thirteen weeks is the longest horizon at which you can still name individual invoices and payment dates, and the shortest at which you can still do something about what you find.
The format has a telling origin: it comes out of corporate restructuring and turnaround work, where lenders to distressed companies demand it weekly as a condition of continued support. When a company's survival is measured in weeks, this is the instrument the professionals reach for. That's worth absorbing: the tool built for companies in intensive care is a very good tool for companies that would simply prefer never to end up there.
What goes into the model
Cash receipts: collections, not invoices
The top section is cash coming in, and here's the distinction the whole model lives or dies on: you forecast when customers will pay, not when you'll invoice them. A $20,000 invoice sent this week on net-30 terms to a customer who habitually pays ten days late is a week-6 receipt, not a week-1 receipt. Build receipts from your accounts receivable aging (what's outstanding, and when each invoice will realistically land) plus expected collections from sales you haven't invoiced yet, lagged by your customers' actual payment behavior, not their contractual terms. Your accounting system already knows each customer's real behavior; a quick look at when their last five invoices actually paid gives you a per-customer lag that beats any blanket assumption.
Disbursements
The bottom section is cash going out, dated by when it leaves the account: payroll (with the exact pay dates), supplier payments, rent, loan payments with interest, and the lumpy items that sink naive forecasts, like quarterly tax payments and annual insurance premiums. Go through the last three months of bank statements once and you'll capture almost everything.
Opening and closing balance per week
Each week: opening balance + receipts − disbursements = closing balance, and that closing balance is next week's opening. The closing-balance row is the forecast; everything above it is working.
Sample structure
| Wk 1 | Wk 2 | Wk 3 | Wk 4 | .. | Wk 13 | |
|---|---|---|---|---|---|---|
| Opening balance | 48,000 | 41,500 | 52,300 | 33,100 | ||
| Customer collections | 22,000 | 31,000 | 9,500 | 28,000 | ||
| Other receipts | 0 | 0 | 0 | 0 | ||
| Total receipts | 22,000 | 31,000 | 9,500 | 28,000 | ||
| Payroll | (21,000) | 0 | (21,000) | 0 | ||
| Suppliers | (6,000) | (14,200) | (5,700) | (8,900) | ||
| Rent | (1,500) | 0 | 0 | 0 | ||
| Loan payment | 0 | (2,000) | 0 | 0 | ||
| Tax | 0 | (4,000) | (2,000) | 0 | ||
| Total disbursements | (28,500) | (20,200) | (28,700) | (8,900) | ||
| Net cash flow | (6,500) | 10,800 | (19,200) | 19,100 | ||
| Closing balance | 41,500 | 52,300 | 33,100 | 52,200 |
Notice the shape: weeks alternate between negative and positive because payroll hits fortnightly and collections are lumpy. That volatility is invisible in a monthly view, and it's precisely what this model exists to expose. Week 3's $33,100 is fine; the same pattern starting from a $15,000 balance would put week 3 within one late customer payment of a missed payroll.
Building it, step by step
- Start from today's actual bank balance. Not the P&L, not the balance sheet cash from last month's close. The P&L contains accruals, timing assumptions, and non-cash items; this model tracks the bank account, so it starts where the bank account is right now.
- List receipts by week. Work through your AR aging invoice by invoice, dating each by realistic payment behavior. Add expected collections from future sales, lagged appropriately.
- List disbursements by week. Payroll on its actual dates, then suppliers, rent, debt service, and tax from your last three months of bank statements. Date the annual and quarterly items; they're the ones that ambush you.
- Compute the balance cascade. Opening + receipts − disbursements = closing, chained across all 13 weeks.
- Sanity-check week 1 against reality. If the model says week 1 ends at $41,500 and Friday's balance is $33,000, an assumption is wrong. Find it now, while the error is one week old.
An afternoon of work the first time. Twenty to thirty minutes weekly after that.
Keeping it rolling
This is where most attempts die, so treat the update as a ritual with a fixed slot: same day every week, ideally Monday morning against Friday's closing balance.
The ritual has three moves. Drop the completed week and record what actually happened. Add week 14 at the far end, so the horizon stays 13 weeks. Correct what missed: if the customer you forecast for week 2 didn't pay, don't just slide the receipt to week 3, ask whether your payment-lag assumption is wrong for that customer and fix the assumption, not just the cell. The model gets more accurate every week precisely because last week's misses tune this week's assumptions. A forecast updated fortnightly is stale; one updated "when things are busy" is dead within a month.
Reading the output
Three numbers come out of the closing-balance row:
Runway. How many weeks until the balance crosses zero (or your minimum operating buffer). The formula, when burn is steady: runway in weeks = current cash ÷ average weekly net burn. But the whole point of the 13-week model is that burn isn't steady, so trust the week-by-week cascade over the formula whenever they disagree.
The trough week. The lowest projected balance across the 13 weeks, and the date it occurs. You can be fine in week 13 and dead in week 7; the trough is the number to manage. If the trough is uncomfortably low, you have (by construction) several weeks of warning to pull the levers: accelerate collections, delay a discretionary payment, draw the credit line before you need it. Set yourself a floor (many operators use one payroll cycle of cash) and treat any forecast trough below it as an action item this week, not a note for later.
Covenant headroom. If a loan requires a minimum cash balance or coverage ratio, chart the projected position against the covenant line every week. Lenders respond much better to "we'll be tight in week 9 and here's our plan" than to a surprise breach.
Common mistakes
Forecasting revenue instead of collections. The model fills with invoice dates instead of payment dates and runs four to six weeks optimistic, which is the most dangerous direction to be wrong in.
Forgetting the lumpy items. Quarterly tax, annual insurance, the software contract that renews annually in one hit. Monthly-average thinking smooths them away; the bank account experiences them as cliffs.
Using monthly figures ÷ 4. Dividing monthly totals evenly across weeks erases exactly the payroll spikes and collection gaps the model exists to reveal. If the inputs are smoothed, the output is decoration.
Keep a rolling 13-week forecast fed by your live collections data.
When 13 weeks isn't enough
If the model shows you crossing zero inside the 13 weeks, its job changes from forecasting to triage, and the response changes with it. Move to daily granularity for the next four weeks. Rank every disbursement by what happens if it's late (payroll and tax never slip; some suppliers can). Get on the phone: to your largest debtors with a discount for immediate payment, to your bank before the crisis rather than during it, and to whoever might fund a bridge. The forecast's role in that conversation is to prove you know your position precisely, which is the difference between negotiating and pleading.
Frequently asked questions
What is a 13-week cash flow forecast?
A week-by-week projection of all cash receipts and payments over the next 13 weeks, rolled forward every week so it always covers one quarter ahead. It's the standard operational tool for managing runway.
How accurate should it be?
Weeks 1-4 should land within a few percent; weeks 10-13 are directional. The weekly update ritual is what closes the gap, since every miss tunes an assumption.
Who needs one?
Any business with under six months of runway, lumpy collections, or a lender to report to; honestly, any founder who's ever been surprised by their own bank balance. It costs 30 minutes a week.
How is it different from a cash flow statement?
The cash flow statement is historical and reports where cash went last period; the 13-week forecast is forward-looking and projects where cash goes next quarter. Same subject, opposite directions in time.
How do I calculate runway?
Steady-state formula: current cash ÷ average weekly (or monthly) net burn = runway in weeks (or months). When burn is lumpy, read runway off the week-by-week model instead.