Profitable companies fail. Not often, but often enough that every owner should understand the mechanism: revenue is recognized on delivery, cash arrives on collection, and the gap between those two events is where businesses die. A thirteen-week rolling cash flow forecast is the instrument that measures that gap.
It is not a new idea. Turnaround practitioners have used the thirteen-week format for decades, which is the point. It survived because it works. What has changed is how quickly you can build one and how little manual effort it now takes to maintain.
Why thirteen weeks
One quarter, expressed weekly. The horizon is deliberate on both ends.
Long enough to see a problem while you can still act on it. A cash shortfall visible nine weeks out is a manageable situation. You can accelerate collections, defer discretionary spend, draw on a facility, or have an early conversation with your lender. The same shortfall discovered nine days out is a crisis with three bad options.
Short enough to be honest. Weekly granularity forces specificity. You are not forecasting “November collections” as a single number; you are forecasting which invoices land in which week, which is a claim you can actually check. Monthly forecasts hide timing risk inside averages. Weekly forecasts expose it.
And it is a rolling forecast, not a static one. Every week you drop the week that closed and add a new week thirteen out. The horizon never shortens, which means you never arrive at quarter-end with no visibility.
What the model actually contains
The structure is simple, and simplicity is a feature. Anyone should be able to read it without a walkthrough.
Opening cash. Actual bank balance at the start of the week. Not book cash, bank cash, reconciled. This is the anchor, and it is the one line that must never be an estimate.
Collections. Built from your actual receivables aging, invoice by invoice, placed in the week you expect payment based on how that customer has behaved historically rather than on their stated terms. A customer whose terms are net-30 and whose behavior is net-52 should be modeled at net-52.
Other inflows. Facility draws, tax refunds, deposits, asset sales. Anything real and dated.
Payroll. Its own line, always. It is the largest and least deferrable outflow most companies have, and it needs to be visible in the week it clears, not blended into operating costs.
Payables. From your actual aging, scheduled by when you intend to pay rather than by invoice date. This is where the model becomes a decision tool: you can see the effect of moving a payment run before you make it.
Fixed and committed. Rent, debt service, insurance, taxes. Known amounts on known dates.
Closing cash, and the covenant line. Closing balance, and directly beneath it, your minimum required balance, whether that is a covenant, a facility floor, or your own comfort threshold. The distance between those two lines is the number that matters, and it should be the first thing the eye lands on.
The discipline that makes it work
Building the model is a day of work. Making it useful is a weekly habit, and the habit is where most implementations fail.
Compare forecast to actual, every week, on the same page. This is the step people skip and it is the entire source of the model’s value. Last week you forecast $412,000 in collections. You collected $327,000. Why? Answering that question every week is what turns a spreadsheet into an early warning system, because the pattern of your misses tells you more than any single forecast.
Hold the same meeting at the same time. Thirty minutes, once a week, same attendees. Ours run Monday mornings. Consistency is what makes the trend visible.
Track your accuracy. Forecast versus actual, as a percentage, by week. Most companies start at sixty to seventy percent on collections and reach ninety after a couple of months. That improvement is not the model getting better. It is your understanding of your own business getting better.
Model scenarios, not just a base case. Base, downside, and severe. The downside case is the one that earns its keep: what happens if your two largest customers each pay two weeks late in the same month? Knowing that answer in advance is the difference between a decision and a scramble.
Excel, or something smarter
A fair question, and the honest answer is that it depends on volume and on how much manual work you are willing to absorb.
Excel remains completely defensible. It is transparent, every formula is inspectable, your lender and your board can read it without training, and there is no implementation project. For a company with a manageable number of customers and vendors, a well-built workbook maintained weekly is sufficient. Most of the models we build start here.
The cost is time. Pulling the aging reports, refreshing the bank balance, rolling the weeks forward, realistically two to four hours a week of skilled attention, indefinitely. And it is fragile in the ordinary way spreadsheets are fragile: one broken reference, one person on holiday.
Automation changes the maintenance cost, not the thinking. This is where we now spend real effort for clients. Direct feeds from the bank and the ledger, so opening cash and both agings refresh without anyone touching them. Collection timing predicted from each customer’s actual payment history rather than from their terms. Variance commentary drafted automatically, so the weekly meeting starts from a summary rather than from a blank page. Alerts when the projected balance approaches the covenant line, rather than waiting for someone to notice.
What automation does not do is remove judgment. The model still requires someone who knows that a particular customer always pays late in December, that a specific vendor will accept a deferred payment without damaging the relationship, that a receivable is technically current and practically doubtful. That knowledge is not in your ledger. It is in your controller’s head, and the model is only as good as how much of it gets encoded.
So the sequence we recommend: build it in Excel first and run it manually for a quarter. You will learn your own business. Then automate the data collection, keep the judgment, and reinvest the recovered hours somewhere they compound.
What it gives you beyond survival
The obvious benefit is not running out of money. The less obvious ones are where the real return sits.
You negotiate from a position of knowledge. When a vendor asks for shorter terms, you know exactly what that costs you in week seven. When a customer asks for longer terms, you can price the concession.
Your lender conversations change character. Bringing a thirteen-week forecast with a tracked accuracy record to a credit discussion is a materially different meeting from bringing a trailing P&L. It signals control, and control is what a credit committee is actually assessing.
Investment decisions get faster. Whether you can fund a hire, a piece of equipment, or an inventory build becomes a question with an answer rather than a question with an argument.
And in a transaction, it is table stakes. Any institutional buyer or lender will ask for it. Having thirteen-week forecasts with historical accuracy going back a year says more about management quality than any narrative in a deck.
Start this week
You need four things: your reconciled bank balance, your receivables aging, your payables aging, and your payroll calendar. Everything else is arithmetic and judgment.
Build thirteen columns. Populate the first four with real detail and the remaining nine with your best estimate. Put your covenant floor on the sheet. Then next Monday, come back and mark it against what actually happened.
Do that for a quarter and you will know your business better than any report has ever told you.