Cash Flow Forecasting for Agencies: How to Build One You'll Actually Use

Most agencies find out they have a cash problem the week they can't cover payroll. That's backwards. Cash flow forecasting for agencies exists to move that discovery three or four weeks earlier, back to the point where you still have options — delay a hire, push a vendor payment, call a slow-paying client — instead of scrambling for a line of credit on a Thursday.

The reason agencies get surprised more than other service businesses is timing mismatch. You bill retainers on the 1st, net 30. You pay contractors on delivery, sometimes weekly. You run payroll every two weeks no matter what's in the bank. A P&L that shows a healthy 18% margin says nothing about whether Friday's payroll clears.

What a real forecast needs to show

A cash flow forecast has three components, and most spreadsheets people call "forecasts" are missing at least one of them.

Starting cash. Your actual bank balance today, reconciled — not what QuickBooks says before you've checked it against the bank feed.

Expected inflows, by week, by client. Not "we'll collect about $90,000 this month." Retainer client A pays $8,000 on the 1st, has for fourteen months straight. Project client B invoiced $22,000 on the 15th, net 30, average actual collection 48 days based on their last four payments. Different clients get different assumptions, because they behave differently.

Committed outflows, by week. Payroll dates, contractor invoices due, software renewals, the loan payment, the estimated tax transfer. These are usually more predictable than inflows, which is exactly why most of the forecasting effort should go toward getting the inflow side right.

The 13-week model

Thirteen weeks — one quarter — is the standard window for a reason. It's long enough to see a slow month coming while you can still act on it, and short enough that your assumptions about individual clients still hold up. A 12-month forecast is useful for planning a hire; it's useless for deciding whether to run payroll on the 15th.

Build it as a rolling model: 13 columns, one per week, starting cash balance carried forward from the prior week's ending balance. Each week: starting cash + inflows − outflows = ending cash. Ending cash becomes next week's starting cash. Update it every week with actuals for the week that just closed, and push the forecast window out one more week so you're always looking 13 weeks ahead.

For accuracy, weeks 1 through 4 should be within about 10% of what actually happens — you know who's invoiced, you know payroll dates, there's not much left to guess. Weeks 9 through 13 will be looser, built on averages and pipeline instead of confirmed invoices. That's fine. The point isn't a precise number thirteen weeks out — it's spotting the week where the line dips below your minimum cash threshold while there's still time to do something about it.

The agency-specific problems that break simple forecasts

Collection lag varies by client type, not by average. If your retainer clients pay in 5 days and your project clients average 48, blending them into one "days sales outstanding" number hides the real risk. Forecast retainer and project revenue on separate lines.

Ad spend pass-through inflates both sides. If you bill $30,000 in media spend through your agency and pass it straight to Google or Meta, that's $30,000 of inflow and $30,000 of outflow that tells you nothing about your actual margin or risk. Strip pass-through spend into its own line so it doesn't drown out the numbers that matter.

Late payers are a pattern, not a surprise. Pull your last six months of accounts receivable. If 20% of invoices consistently land 15+ days past terms, build that lag into the forecast for those specific clients instead of assuming everyone pays on day 30.

Tax transfers get forgotten. If you're moving 25-30% of profit to a tax holding account weekly (you should be), that outflow needs its own line in the forecast — not a surprise that shows up as a smaller-than-expected balance in April.

Building yours in four steps

1. Pull 90 days of actual bank activity. Not the P&L — the bank feed. You need to see when money actually moved, not when it was invoiced or booked.

2. Group clients into payment-behavior buckets. Retainers that pay like clockwork, project clients with a known average lag, and anyone who's chronically late. Three or four buckets is usually enough.

3. Lay out 13 weeks of committed outflows first. Payroll, contractors, recurring software, debt payments, tax transfers. These are known, so get them locked in before you touch the inflow assumptions.

4. Layer in inflows by bucket and reconcile weekly. Apply your payment-behavior assumptions to what's actually invoiced and in the pipeline. Then every week, replace that week's forecast with the real number and roll the model forward. The forecast that never gets checked against reality is worse than no forecast — it just gives you false confidence.

How often to update it

Weekly, on the same day, ideally right after you reconcile the bank account. Monthly is too slow to catch a payroll gap with enough runway to fix it — by the time a monthly forecast flags a problem, you're often inside the two-week window where your only options are a credit line draw or a hard conversation with a vendor. A 15-minute weekly update is what keeps the forecast useful instead of decorative.

This is one of the recurring deliverables built into our fractional CFO work — a live 13-week forecast tied to your actual Digits transaction data, updated weekly, not rebuilt from scratch every quarter. Pair it with a clean read on your payroll-to-revenue ratio and you've got the two numbers that catch most agency cash problems before they become payroll problems.