Skip to content
Predict.ai

Decisions

Cash flow forecasting: receipts, not a P&L in disguise

Treasury cares when money lands, not when revenue was recognized. A cash forecast is collections, payroll, tax, and drawdowns on a calendar — messy, lumpy, and allergic to MAPE if a day can be negative or zero.

Updated Aug 26, 2026·7 min read

Not the revenue forecast

Treasury cares when money lands, not when revenue was recognized. A cash forecast is collections, payroll, tax, and drawdowns on a calendar — lumpy, sometimes signed, allergic to MAPE. If you stretch the P&L path into the bank, you will be right on profit and still bounce a payment.

Lumps are the point

Tax days, payroll, a big receivable, a capex wire: the series is not supposed to be smooth. Calendar those. Do not make a neural net invent a board vote. One-off M&A is a scenario or a known-ahead flag, not a pattern.

  • Forecast receipts and outflows separately if they have different drivers.
  • Collections often lag billings; that lag is a driver, not a nuisance.
  • FX and interest are their own series if they actually move the cash number.

WAPE gets weird on signed cash

Days can be negative. Zeros happen. MAPE is a trap. MAE in money, or a scaled error the treasurer feels, beats a percentage that explodes. Bias still matters: systematically late collections are a process, not a model aesthetic.

A goal on the cash signal

Push collections and outflows as their own signals in Predict.ai. The platform will not assume they behave like units sold. Pick a metric that survives signs. Driver discovery can find the billing series, a season, a payday calendar. The live path sits in the same workspace finance already peeks at — not a shadow model in a treasurer’s laptop.

What-if a large customer pays 14 days late. That is a scenario on the champion, with an interval, not a new spreadsheet tab named FINAL_v7.

FAQ

Can I forecast cash from the P&L forecast?
You can derive a crude bridge. You will miss timing. Treasury is a timing product. Forecast the bank, or you will be right on profit and still bounce a payment.
What about one-off M&A or capex?
Do not make the model invent them. Put them in as known-ahead covariates or a scenario. A tournament cannot see a board vote.

Keep going

Try it on your data.

Connect an outcome and its history. Predict.ai finds the drivers, races the models, and keeps the forecast live — in the workspace, over the API, and through agents.