The first thing I ask a new fractional CFO client is "show me your cash forecast." About half pull up an annual budget — a 12-column spreadsheet built in October, last touched in January, already wrong by April. The other half don't have anything at all; they have a bank balance and a vibe. Both groups are flying blind in different directions.
The artifact that actually works is the 13-week rolling cash forecast. It's the financial document I've never seen a competent CFO operate without, and the absence of one is the single most reliable predictor that a company will be surprised by cash in the next 12 months. It's also misunderstood as a heavy enterprise tool — it isn't. Properly built, it takes 20 minutes a week to maintain and pays for itself the first time you spot a shortfall four weeks out instead of four days out.
Why annual budgets fail and the 13-week wins
Annual budgets fail for two structural reasons. First, they're accrual — they tell you what revenue you're recognizing, not what cash is landing in the account. A SaaS company with a great P&L can run out of money because annual contracts are paid quarterly and the accrual hides the lumpiness. Second, they go stale within weeks. A budget built in Q4 makes 24 assumptions about Q3 of next year that are all going to be wrong. By the time you notice they're wrong, the cash event has already happened.
The 13-week forecast solves both. It's cash-basis (only money that actually moves) and it rolls — every Monday you drop the oldest week and add a new one at the far end. The window is always the next 13 weeks. You can always see one full quarter of cash flow ahead, which is enough lead time to fix almost any solvable problem.
What goes into the forecast (and what doesn't)
The mistake most first-time builders make is treating the 13-week forecast like a mini P&L. It isn't. It's a bank-account simulator. The only inputs are events that change the bank balance.
| Section | What it captures | Examples |
|---|---|---|
| Starting cash | Today's actual bank balance | The number from your bank, not from QuickBooks |
| Collections | Money coming in by week | Invoiced AR by due date, expected new bookings, refunds received |
| Payroll | Net pay + payroll tax + benefits | Hits on the 1st and 15th in the US; whatever your cycle is |
| Vendor AP | Money going out by week | Open POs, recurring vendor bills, software subscriptions |
| Taxes | Quarterly estimateds + sales tax | Q1, Q2, Q3, Q4 federal estimateds; monthly sales tax filings |
| Debt service | Loan principal + interest | Term-loan amortization, line-of-credit interest |
| One-offs | Anything else that moves cash | Lawsuit settlement, equipment purchase, fundraise net of fees |
What's deliberately NOT in there: depreciation, amortization, revenue recognition adjustments, stock-based compensation, deferred revenue movement. None of those move the bank balance, so none of them belong in a cash forecast. If you want a P&L, build a P&L — but don't conflate them.
The five line items everyone gets wrong
After building hundreds of these, the same five inputs are mismodeled in roughly 80% of first drafts.
1. Collections timing. People put the AR amount in the week the invoice was sent, not the week it's expected to land. A net-30 invoice sent on May 1 doesn't generate cash until June 1 at best, and for most B2B customers that's actually June 15–30 in practice. Build an aging-based collection curve from your last 12 months of actual payment behavior; don't trust the contractual terms.
2. Payroll fully loaded. Net pay is about 70% of fully-loaded payroll. The other 30% is employer taxes (FICA, FUTA, SUTA), benefits (medical, dental, 401k match), and the third-party payroll-processor fee. Forecasts that only model net pay underestimate cash burn by 25–40%.
3. Sales tax payable. US sales tax is a pass-through, but it sits in your bank account between when you collect it and when you remit it. Lots of cash forecasts double-count it: they include the gross customer payment as collection AND treat the remittance as an outflow. The right move is to forecast the net of the two (sales-tax payable is a liability, not an expense).
4. The big quarterly cliff. Q4 federal estimateds, year-end 401k true-up, holiday-payroll-cycle quirks, annual software renewals — all of these stack in December. A January-built forecast almost always under-models the size of the December cliff. Walk through last year's December bank statement line by line before forecasting this one.
5. The fundraise net. Founders often plug in "$5M raise — Q3" as a single $5M inflow. The actual cash event is $5M minus banker fees (typically 5–7% on a primary), minus legal ($75–150k), minus the bridge of any convertible notes being settled at closing. Net is typically 88–92% of headline. That 8–12% delta has been the difference between "we made payroll" and "we didn't" more than once.
How to run the weekly review (in 20 minutes)
Monday morning, before any other meeting:
- Roll the window. Drop last week's actuals into the "actuals" tab. Append the new week 13 at the far end.
- Update starting cash. Pull the current bank balance from the bank — not from accounting. Variance between the two is one of the most useful signals you'll generate; investigate anything over 5%.
- Compare last week's forecast to actuals. Where did you miss? Late collection? Surprise vendor payment? Update the forward weeks based on the pattern.
- Check the trough. Find the lowest cash point in the next 13 weeks. If it's below 6 weeks of operating expenses, you have a problem to solve — and you found it 4–13 weeks before it would have surprised you.
- Email the CEO a one-line summary. "Trough at week 8, $1.4M, driven by Q3 estimated tax + the Series A bridge maturity. Plan: accelerate two enterprise collections in weeks 5–6."
That's it. Twenty minutes, every Monday. The discipline isn't the model — it's the consistency.
What to do when the forecast shows a problem
The forecast finds the problem; the founder solves it. The toolkit is roughly the same every time:
- Accelerate AR. Offer a 2% discount for payment in 10 days on your largest open invoices. Small concession, big cash impact.
- Stretch AP. Move recurring vendor payments from net-30 to net-45. Most vendors won't notice; the ones who do will negotiate rather than lose you.
- Trim discretionary spend. Travel, conferences, agency retainers, contractor budgets. These are the levers that move fast without touching headcount.
- Bridge financing. A revolving line of credit is the emergency lever; arrange it before you need it. Banks lend to companies that don't need the money, not to companies that do.
- Layoff planning. The last lever, the most painful, the one that requires the longest lead time. If the trough is more than a 10% cut to break even, the discussion has to start now — severance, COBRA, legal review, and notice all consume weeks.
A 13-week forecast doesn't prevent cash problems. It just guarantees you see them with enough runway to solve them with options other than panic.
The founders I've watched survive the hardest downturns weren't the ones with the most sophisticated models. They were the ones who looked at their cash position every Monday morning without fail and asked one question: where's my trough, and what's my plan if it gets 20% worse? The 13-week forecast is just the artifact that makes that question answerable.
