Cash forecasting projects when cash will arrive and leave so treasury can fund gaps, invest surplus, and hold the right buffers. It is the visibility engine inside liquidity management — not a substitute for cash pooling or an in-house bank, but the input those structures need to run well.
A forecast that operations will not maintain is decoration. Design for the shortest horizon people will actually update, then layer longer views that finance already produces.
Direct vs indirect
Direct (receipts / disbursements) builds from expected cash movements: collections, payroll, tax, debt service, CapEx, dividends. Best for daily to ~13-week views when bank and ERP data are timely.
Indirect (P&L / balance-sheet bridge) starts from projected earnings and working-capital changes, then bridges to cash. Best for monthly to annual planning when the question is funding capacity, not tomorrow’s overdraft.
Most groups run both. Short horizon stays direct; medium and long horizon stay indirect or a hybrid. See direct vs indirect cash forecasting for the choice frame.
Horizons treasurers actually run
| Horizon | Typical method | Decision it supports |
|---|---|---|
| Same-day / next-day | Direct, bank + TMS | Sweeps, release funding, overdraft avoidance |
| Weekly / 13-week | Direct or hybrid | Pool targets, short-term investment, line draws |
| Monthly / quarterly | Indirect or hybrid | CapEx, dividends, debt calendar |
| Annual / multi-year | Indirect | Structure, contingency lines, strategic buffers |
Daily accuracy dies if every entity rebuilds from scratch. Centralize bank actuals and variance rules; push only material drivers to the field.
Drivers and variance
Forecast quality is driver quality:
- Collections — invoice aging, dispute rates, customer seasonality (working capital management)
- Payables — payment runs, early-pay discounts, CapEx gates
- Payroll / tax / debt — calendar-driven and usually high confidence
- Intercompany — IC invoices, loans, and nets before pool sweeps (intercompany netting)
- FX and trapped cash — corridors where cash cannot move on the forecast date (cross-border cash pooling, country notes such as China)
Run a variance loop: forecast vs actual by driver, not just one total miss. Without that loop, models drift and teams stop trusting the number.
Systems map
| Layer | Role in the forecast |
|---|---|
| Bank / multi-bank reporting | Opening balances, same-day actuals |
| ERP | Open AR/AP, payment proposals, CapEx commitments |
| TMS | Aggregation, scenarios, funding recommendations |
| Pool / IHB books | Header vs participant positions after concentration |
| [[payment-factory | Payment factory]] |
The TMS does not invent cash. It consolidates bank and ERP truth and applies treasury policy. If ERP cutoffs and bank MT940/camt.053 timing disagree, fix the data path before adding scenarios.
How it feeds liquidity structures
- Pooling — daily/weekly forecasts set sweep targets and header investment appetite (cash pooling)
- Cross-border designs — forecast must flag trapped or delayed cash, not assume every surplus is movable (cross-border cash pooling)
- IHB / netting — IC settlement calendars change entity-level cash even when group cash is flat (in-house bank, intercompany netting)
- Payment factory — release calendars and POBO corridors only work if funding is predicted early enough
Minimum design package
- Horizon stack (daily / 13-week / monthly) with an owner per layer
- Method choice per horizon (direct, indirect, or hybrid) — see direct vs indirect
- Driver dictionary and variance cadence
- Bank + ERP + TMS data contracts and cutoff times
- Link to concentration and funding playbooks under liquidity management
- Scenario set (base / downside / trapped-cash) that leadership will actually use
Forecasting earns its keep when it changes a funding, investment, or buffer decision — not when it produces a prettier spreadsheet.