How to Build a 13-Week Cash Flow Forecast

All right, here is a thing I have noticed after two decades of sitting with founders and their numbers. Almost nobody gets surprised by their annual model. They get surprised by a Tuesday. The quarter looks fine, the P&L looks fine, and then payroll lands three days before the big customer pays, and a perfectly profitable company suddenly has a very real problem on its hands. A 13-week cash flow forecast is the one report that closes that gap, and it is not complicated to build. Let me show you.

Here is the reframe that makes the rest of this make sense. Your accounting tells you what happened. Your annual model tells you whether the business works. Neither of them tells you which week you run short, because neither of them is built in weeks. A 13-week forecast is a rolling, weekly projection of the actual cash moving through your actual bank account over one quarter. Opening cash, cash in, cash out, closing cash, thirteen times across. It is not an accounting deliverable. It is an operating one, and it is the only view that answers the question that keeps founders up.

The stakes are more ordinary than dramatic, which is exactly why this gets skipped. The JPMorgan Chase Institute studied 470 million transactions across 597,000 small businesses and found the median business holds 27 cash buffer days, with the bottom quarter sitting at 13 days or fewer.¹ Their framing: "half of all small businesses hold a cash buffer large enough to support 27 days of their typical outflows."² That research runs on 2015 data, so I would treat it as an order of magnitude rather than today's reading, but the order of magnitude is the whole point. Twenty-seven days is less than one 13-week forecast. If that is anywhere near you, this is not a finance exercise. It is the instrument panel.

So let me walk you through how the thing actually works, why the convention landed on thirteen weeks and why nobody can tell you where that came from, what belongs on each of the four lines and the one almost everyone gets wrong, the weekly hour that turns a spreadsheet into a system, and a couple of stories about what happens when nobody is watching the calendar.

How the forecast actually works

The important word is direct. Your cash flow statement is built the indirect way, starting from net income and adjusting for non-cash items and balance sheet movements. That is the correct method for reporting what already happened. It is the wrong method for seeing what is about to. A 13-week forecast uses the direct method instead, which as Financial Edge Training puts it, "distills all activities of a firm down to cash receipts and cash disbursements."³ Wall Street Prep describes the same mechanic as forecasting weekly cash receipts less cash disbursements, and notes the output usually carries one week of actual history alongside the projection.⁴

Cherry Bekaert lays the difference out about as cleanly as I have seen it done. The indirect model is retrospective, it is prepared monthly or quarterly, and it depends on a completed close, which means the insight arrives days or even weeks after the period it describes. The direct model is forward-looking, updated weekly, and gives you visibility without waiting on the close at all.⁵ Their line on it stuck with me: "While indirect models explain what has already occurred, the 13-week cash flow model equips organizations to anticipate what lies ahead and act accordingly."⁶ That is the entire case for building one, and notice it has nothing to do with being in trouble.

Structurally you are looking at a spreadsheet with thirteen columns, one per week, and four summary rows. Closing cash for week one becomes opening cash for week two, and so on across the quarter, which is what makes the thing a chain rather than thirteen separate guesses. Underneath those four summary rows sit the line items, and I will be straight with you, the line items are where all the work is. The four rows take an afternoon. Getting the detail underneath them honest is what separates a forecast you act on from a forecast you ignore.

Why thirteen weeks, and why nobody can tell you where it came from

I want to be upfront here, because this is the part where a lot of articles quietly invent a tidy history. I went looking for an authoritative source on why the convention settled on thirteen weeks. I could not find one. There is a story circulating that the big restructuring firms standardized the format in the 1980s and 1990s for Chapter 11 work, and it appears in a dozen software vendor blogs, and I could not trace it to any of those firms or to any primary source at all. So I am not going to repeat it as fact.

What I can tell you is where the convention actually lives, and honestly that turned out to be more useful than an origin story. It lives in loan documents. Pull the negotiated language out of real credit agreements and debtor-in-possession facilities and the same defined term keeps surfacing with the same three requirements: thirteen weeks, rolling, and projected on a line-item basis, certified by an officer as prepared in good faith. One agreement defines it as "a thirteen-week rolling operating budget and cash flow forecast, which shall reflect the Parent's good faith projection of all weekly cash receipts and disbursements."⁷

Those agreements also carry something more interesting than a definition. They carry a tolerance. One facility tests operating cash disbursements every four weeks against the most recent 13-week budget, with a permitted four-week variance of 15 percent.⁸ I find that genuinely useful, because it is a real contractual answer to the question every founder asks me about ten minutes into this conversation, which is how accurate does this actually need to be. Fifteen percent over four weeks is looser than most founders assume and tighter than most founders achieve on their first few passes. Somewhere in that gap is a reasonable target.

Worth saying plainly that this is not a distress tool. Financial Edge notes the same model gets used by companies nowhere near a bankruptcy court, specifically to reassure lenders worried about liquidity and broken covenants.⁹ Eric Gelb and Joe Marchese at PKF O'Connor Davies write that "the 13-week forecast is essential for informed decision-making and often forms the financial foundation for restructuring plans."¹⁰ Both things are true, and the second one is precisely the argument for building it early. You want this instrument on the panel long before anyone uses the word restructuring in a sentence about your company.

My own reasoning on the number, offered as reasoning and not as a citation, is that thirteen weeks is one quarter. It reaches far enough out that you see a funding need coming while you still have room to do something about it, and it stays close enough in that you can name actual invoices and actual payment dates instead of guessing at averages. Push it to twenty-six weeks and the back half quietly becomes the annual model again. Pull it back to eight and you find out too late to act.

The four lines, and the one everybody gets wrong

Start with opening cash, and please do not skip past this, because it is the step almost everyone treats as trivial. Add up checking, savings, and money market. Then do the harder part that PKF O'Connor Davies recommends and ask how quickly each balance actually converts to good funds, and what it costs you to get at it, including things like early withdrawal penalties on a certificate of deposit.¹¹ A CD with a penalty attached is not week-one cash. Neither is a balance sitting in an account you cannot move money out of on a Friday afternoon. Opening cash means cash you can actually reach, and the number of forecasts I have seen that quietly overstate this by a hundred thousand dollars is not small.

Cash in should be driven by the date you expect money to land, not the date you sent the invoice. Start from the AR aging rather than the sales forecast. Wall Street Prep's approach is the right one here: forecast collections off DSO for the long tail, and use invoice-level assumptions for your larger customers, because one big account paying two weeks late is usually the entire variance in the whole model.¹² Once the existing receivables are laid in by expected date, layer new sales on top by expected collection date and discount them for probability, then add financing draws, tax refunds, and anything else genuinely expected.

Cash out starts with known obligations, in order of how immovable they are. Payroll and payroll taxes on the real run dates, and treat the taxes as their own line rather than folding them into payroll, because their timing is different and founders forget them constantly. Then rent, debt service, sales and payroll tax remittances, insurance, and capital expenditure. Only then trade payables, timed off DPO and your genuine vendor-by-vendor intentions with critical suppliers identified first.¹³ Financial Edge would have you keep capital expenditure in the operating block when it is operating in nature, and hold financing items and professional fees separately as non-operating.¹⁴

On the weight of that payroll line, Wall Street Prep makes a point worth internalizing before you need it: accrued wages and benefits are frequently the single largest disbursement once a company is under any kind of stress.¹⁵ Every other obligation in your business can be renegotiated over an afternoon phone call. That one cannot, and we will come back to it.

Closing cash is the fourth line, and it rolls into next week's opening. The thing I would add, which most templates leave out, is a minimum cash floor shown right on the face of the model with the gap to it calculated. That single addition changes the question you ask the spreadsheet. Instead of how much do we have, which is a number, you start asking which week do we breach and what are the three weeks before it worth doing, which is a decision.

One framework that helps enormously with all of this comes from the Association for Financial Professionals, which sorts inputs by how knowable they actually are: items that are certain, like a loan repayment or a tax deadline, items that are predictable, like payroll, and items that simply are not, like a repair you have not had yet.¹⁶ You forecast those three groups differently and you stop pretending the third group is knowable. AFP's related observation is one I would tattoo on the inside of every template: "Short-term forecasts are always likely to be the most accurate."¹⁷ That is the whole reason this horizon works.

The weekly hour that makes it real

A 13-week forecast built once is a document. Built every week, it becomes a system, and the gap between those two things is enormous. The rhythm that works, in my experience, is a standing thirty to forty-five minute meeting on the same day each week with three items on it, and no more than three.

First, actual versus forecast for the week that just closed. Not to assign blame, and I would say that out loud in the room, because the moment this becomes a blame exercise people start smoothing their numbers and the forecast dies. You are looking for which line you are systematically wrong about. AFP's guidance is to run variance analysis on a timely, consistent basis but to focus on materiality to cash flow rather than chasing every small difference.¹⁸ Greg Lattanzi, a senior treasury analyst at IGS Energy, put the underlying discipline simply: "Practitioners need to stay in touch and abreast of business changes."¹⁹

Second, roll the window. The completed week comes off the front and a new thirteenth week goes on the back. This sounds mechanical and it is the step that quietly protects you, because a forecast that never rolls slowly shortens into a two-week panic without anyone noticing it happened.

Third, confirm who owns each line. This is the part Alvarez & Marsal get right in how they build these. Christopher Duggan, a managing director there, identifies the core problem in most liquidity surprises as "the lack of reliable and transparent cash flow forecasting," and describes their method as bottoms-up specifically so that key assumptions are visible and accountability sits with the management team rather than with a spreadsheet.²⁰ In practice that means collections belong to whoever actually talks to customers, payables belong to whoever actually schedules payments, and payroll belongs to whoever runs it. A forecast where finance quietly guesses on everyone else's behalf is a forecast that is wrong in ways nobody in the room can explain.

One honest warning about the weekly cadence. Every update is also an opportunity to break the spreadsheet, and Wall Street Prep flags exactly this, that weekly updating means every update carries model error risk.²¹ Keep it on one page. Keep the formulas boring. Tie opening cash back to the actual bank balance every single week without exception. Ticks and ties. It is not glamorous work and it is the only reason anyone in the building keeps trusting the output six months in.

And if this feels harder than it should, you are in reasonable company. AFP's 2025 benchmarking work found that "Over 60% of treasury professionals cite cash or liquidity forecasting as the most challenging task they face."²² Those are people who do this full time, with tooling, at companies with actual finance departments. Meanwhile "the perception of cash forecasting as easy has been cut in half over the last seven years," with the share calling it easy falling from 28 percent in 2018 to 14 percent in 2025.²³ The expectations went up while the job got harder. That is not a reason to skip it. It is a reason to keep it simple enough that you will actually do it.

The statistic I went looking for and could not find

If you read anything at all about small business cash flow you will run into this line: 82 percent of small businesses fail because of cash flow problems. It is everywhere. I went looking for the study behind it and I do not believe there is one.

It appears on SCORE's site as a single standalone sentence with no citation, no study name, no sample size, and no link, written by an author whose own bio identifies him as a marketing director rather than a researcher.²⁴ The organization most often credited with it does not have it on the relevant page at all. Every trail I followed ran back to marketing blogs citing one another. I am telling you this partly because I would rather hand you a smaller number that is real, and partly because it is a fair summary of the state of advice in this corner of the internet, which is itself an argument for building your own forecast instead of trusting folklore.

So here are numbers that hold up. Roughly a quarter of new business establishments do not survive their first twelve months, and that is in an ordinary year, not a crisis one.²⁵ The SBA's read on the longer arc is that seven out of ten new employer firms last at least two years and about half survive five.²⁶ In the Federal Reserve's 2025 Small Business Credit Survey, 94 percent of employer firms reported at least one financial challenge in the prior twelve months and only 47 percent were operating at a profit.²⁷ And of the firms that went looking for financing, the Fed found "The most common reasons firms sought financing were to meet operating expenses (56%) or to pursue an expansion or new opportunity (46%)."²⁸ Read that last one twice. Most small business borrowing is a timing problem wearing the costume of a growth decision.

I will note, because they note it themselves, that the Fed survey is a convenience sample rather than a random one and should be read with that in mind.²⁹ I would still rather give you a caveated real number than a confident invented one.

On the venture side, CB Insights looked at 431 companies that publicly shut down since 2023 and found that running out of capital tops the list at 70 percent, but their own framing of that number is the honest part: "Ran out of capital tops the list at 70%, but it's almost always the final cause of death, not the root problem."³⁰ That is exactly the right way to hold this. A 13-week forecast will not fix your product market fit. What it will tell you is how many weeks you have left to fix it, and CB Insights found the median company had 22 months between its last raise and its death. Twenty-two months is a lot of runway to work with, if you can actually see it.

The cautionary tale I would rather you hear from me

The cleanest illustration I know is Carillion, and it is clean precisely because Parliament documented it rather than a vendor blog.

Carillion reported profits and raised its dividend every single year. The House of Commons Library found that between January 2012 and June 2017 the company paid out 333 million pounds more in dividends than it generated in cash from operations, and that "In the eight years from 2009 to 2016, Carillion paid out 554 million in dividends, three quarters of the cash it made from operations."³¹ The July 2017 profit warning was 845 million pounds. By September the hit had reached 1.2 billion, enough to erase the profits of the previous eight years combined. It went down owing roughly 2 billion pounds to about 30,000 suppliers, most of whom recovered very little.

How it funded that gap is the part that should get the attention of anyone who has suppliers. It stretched them. Two parliamentary committees found Carillion imposing standard payment terms of 120 days while sitting as a signatory to the Prompt Payment Code, and Frank Field, who chaired the Work and Pensions Committee, said "The company used its suppliers as a line of credit to shore up its fragile balance sheet."³² To be precise, because precision matters when you are repeating a story about a failed company, Carillion's last finance director told Parliament that average 2017 payment was 43 days with about 5 percent of suppliers at 120.³³ The 120 days was the term imposed, not the blended average. The point stands either way. A profitable-looking company, a rising dividend, and cash walking steadily out the door the entire time. No annual model catches that. A weekly one does.

Then there is the opposite failure mode, which is not about solvency at all. In March 2023, thousands of funded, entirely solvent startups came within days of missing payroll because their money was briefly unreachable. Axios captured the distinction perfectly at the time, noting that "The companies' funds aren't gone, the issue Friday was access to those funds."³⁴ Fortune's account documents the mechanism, including that the payroll company Rippling was running about two billion dollars a month through the affected bank, so when those accounts froze its clients simply could not pay people.³⁵ Eynat Guez, the CEO of Papaya Global, said of payroll: "This is the most important liability and the heart of the relationship."³⁶ She is right, and it is why payroll gets its own line, with its own real dates, in every model I build.

And sometimes the failure is simply not knowing. Fisker temporarily lost track of millions of dollars in customer payments as deliveries scaled, ran an internal audit that stretched across months, and could not state accurately how much revenue it had actually collected. In some cases vehicles were delivered with no payment taken at all, and a former employee told TechCrunch that "Checks were not cashed in a timely manner or just lost altogether."³⁷ The company disclosed multiple material weaknesses in its SEC filings and filed for bankruptcy in June 2024.

I would add one note on the environment you are forecasting into, because two things are getting harder at once. Your customers are deliberately paying slower. The Hackett Group found days sales outstanding degrading for a second consecutive year among large US public companies as customer bargaining power drove extended payment terms, with accounts receivable now the largest single component of excess working capital at 600 billion dollars.³⁸ That is large-cap data and I will not pretend it is a small business benchmark, but the direction of travel reaches you, because those companies are often your customers. Vince Griffin at Hackett noted that "Finance leaders ranked working capital optimization as their top priority for the year in our 2025 Finance Key Issues Study."³⁹ Their optimization is your collection delay.

So here is where I would leave it. Open a blank sheet this week. Thirteen columns, four rows. Put in Friday's real bank balance, the invoices you genuinely expect to collect and roughly when, and every payment you know you owe with its actual date attached. It will be ugly and it will be wrong in places. Do it again next Friday and fix what was wrong. By about week four you will have something that tells you the truth about your calendar, and I have watched enough founders go through this to tell you what they say afterward. The relief is not in the number. It is in not being surprised anymore.