Cash flow forecast for freelancers: a 13-week method with three lines

A profit and loss report tells you what you earned last month. A cash flow forecast tells you whether you can pay rent in week 7. For freelancers the gap between the two is wide, because clients pay late, projects start when they start and taxes come quarterly. The SBA's financial planning resources make the same point: a business can be profitable and still run short of cash at the wrong moment. Thirteen weeks is long enough to see a gap coming and short enough that the numbers mean something.

The three lines of money in

Keep the lines cumulative: likely includes sure, and possible includes both. Then you can read three running balances and plan on the sure one. Leave out invoices more than 60 days overdue; list them separately and chase them, but don't plan rent around them.

Where each number comes from

• Client delay. For each client, average how many days after the due date their past invoices were paid. A client who pays 12 days late on net 15 really pays in 27 days. The DSO calculator gives the same idea across all clients.
• Acceptance rate. Accepted proposals divided by proposals with an outcome (accepted, declined or expired) over the last year. 6 accepted out of 15 is 40%.
• Days to accept. The median number of days between sending a proposal and the client accepting it. Use 7 if you have no history.
• Expenses. Total the last 3 full months of business expenses, divide by 3 for an average month, and spread it evenly: multiply by 12 and divide by 52 for a week. Mark the ones that repeat every month, like software and rent.
• Taxes. Put your quarterly estimated payments in the weeks they're due. How much to set aside for taxes and the quarterly estimated taxes guide cover the amounts.
• Starting balance. What's in your business account today. Without it, the forecast shows only the change, starting from zero.

Build it in a spreadsheet, step by step

1. Make 13 columns, one per week, starting with the current Monday.
2. List every unpaid invoice with its amount and its expected week (due date plus that client's delay). Anything expected in a week that already passed goes in this week.
3. Below them, list accepted work not yet invoiced and the next retainer invoices, each in its expected week.
4. Below that, list open proposals at amount × acceptance rate, in their expected week.
5. Add a money-out row: the weekly expense average in every column, plus tax payments in their weeks.
6. Add three running balances from your starting balance: sure minus money out, likely minus money out, possible minus money out.
7. Highlight any week where the sure balance drops below one month of expenses. That's the week to act on now.

Read the forecast for tight weeks

Look at the lowest point of each line, not the end. A forecast that ends positive can still go negative in week 6 if a big invoice lands in week 7. When the sure line dips, you have three moves: bring money in sooner (invoice accepted work now, ask for a deposit on new work, chase the late client before the due date), push money out later (move a non-urgent purchase), or hold more cash. Each move is visible in the forecast before you make it. If billable hours are waiting to be invoiced, the unbilled time calculator shows how much is sitting there.

What AI should and shouldn't do in a forecast

A forecast is arithmetic on your own records, so code should do it: the same inputs always give the same result, and every number traces back to an invoice or an expense. A language model asked to “forecast my income” will produce plausible numbers it can't trace. Where AI helps is reading the finished forecast back to you in plain words: which weeks are tight, what moves the result most, and which assumption to check. It shouldn't advise you on loans, investments or taxes.

The 13-week forecast in SoloWOS

In Invoicing, the Cash flow tab builds the forecast from your records, with no AI: sure from sent and viewed invoices at the due date plus each client's usual days late (clients who pay late are marked At risk, and invoices more than 60 days overdue are listed apart), likely from the next recurring invoices and proposals accepted in the last 90 days that aren't invoiced yet, and possible from open proposals at your acceptance rate. With fewer than 5 proposals that have an outcome, the rate is 0% and the page says so. Money out is your average month of expenses over the last 3 full months, spread by week, with repeating vendors marked, plus the estimated tax payments due in those weeks from the tax estimate in Expenses. Billable time not invoiced yet is shown apart, and each currency gets its own forecast; nothing is converted.

You can type a starting balance, which stays in your browser and never reaches the server. Explain with AI recalculates the forecast and sends only the weekly totals, the balances from zero and the assumptions, with no client names, invoice numbers or vendors; it returns 3 to 5 sentences on gaps and tight weeks, and isn't saved. It uses one AI draft each time it works (3 a month on Basic, 200 on Solo). SoloWOS uses Claude models from Anthropic. Under Anthropic's commercial terms, the data sent for a request isn't used to train their models, and the SoloWOS privacy policy lists what each AI feature sends. The SoloWOS AI assistant answers “how much will come in next month?” from the same numbers.

The forecast is only as good as the records behind it, so send invoices clients pay by card and log business expenses with receipts in the same account. For the terms that move the sure line, see net 30 vs due on receipt.