Indirect Method Cash Flow Forecasting builds the cash projection from forecast net income, then adjusts for non-cash items (depreciation, amortisation, stock-based comp) and changes in working capital (AR, AP, inventory, accrued liabilities). It is the standard approach for horizons beyond 13 weeks, especially 12 to 18 month strategic plans, annual budgets, M&A models, and covenant scenarios.
Direct method forecasting is built from receipts and disbursements: actual invoice cash-in dates, payroll runs, AP runs, debt service. That granularity makes it powerful for 13-week tactical planning but useless beyond about 90 days, because no one knows which specific invoices will exist in month seven. The indirect method solves the horizon problem by climbing one altitude higher. Instead of asking which dollars land when, it asks what will the P&L produce, and how will the balance sheet absorb or release cash around it. That is why every annual operating plan, every five-year strategic model, every lender covenant scenario, and almost every M&A target model is built indirect. It is the only method that scales past a quarter without collapsing into spreadsheet entropy.
The mechanics follow the standard cash flow statement reconciliation, but forward-looking. You start with forecast net income from the FP&A model. You add back non-cash items: depreciation, amortisation, stock-based compensation, deferred taxes, impairment charges, and unrealised FX. Then you adjust for changes in working capital accounts. An increase in accounts receivable is a use of cash (revenue booked but not collected). An increase in accounts payable is a source of cash (expense booked but not paid). Inventory builds consume cash. Accrued liability movements release or absorb it. Add capex, debt issuance or repayment, dividends, and buybacks below the operating line and you have a full indirect forecast.
In practice the working capital block is where 80 percent of the accuracy lives. Most teams drive it off DSO, DPO, and DIO assumptions tied to revenue and COGS forecasts. A common shortcut: hold DSO flat at trailing 12-month average. A common mistake: holding it flat when collections velocity is actually deteriorating because of an AR aging problem nobody flagged.
The textbook framing (direct is more accurate, indirect is easier) misses the point. Both methods are accurate for what they are designed to do. Direct is accurate for the next 13 weeks because you can see the invoices. Indirect is accurate for the next 12 months because you can model the P&L. Asking which one is better is like asking whether a microscope is better than a telescope.
What most mature treasury functions actually do: run both. The direct 13-week governs intraweek funding decisions, revolver draws, and short-term investment laddering. The indirect 12 to 18 month governs annual planning, covenant testing, capex pacing, and dividend policy. The two are reconciled at the 13-week boundary. If the direct method says cash at week 13 will be 84 million euros and the indirect method says month 3 cash will be 91 million euros, you have a 7 million euro gap to explain before either number gets used in a board meeting.
Indirect forecasts decay because they compound assumption error. Benchmarks suggest median forecast accuracy at month 1 sits around 92 to 95 percent for mature teams, drops to roughly 85 percent by month 6, and slides below 75 percent by month 12 in volatile sectors. Three drivers explain most of the slippage:
The fix is rolling refresh discipline: re-baseline the indirect forecast monthly with actuals, rerun the working capital sensitivities, and explicitly version the assumption set so variance analysis is possible.
The single largest swing factor in an indirect forecast is the AR balance, which is downstream of order-to-cash performance. Every day of DSO on a 500 million euro revenue base is roughly 1.37 million euros of trapped cash. Three DSO days is 4.1 million euros. This is why AR-led treasury teams treat the cash application close rate, deduction resolution velocity, and dispute aging as forecast inputs, not back-office metrics. Agentic AR platforms that close out cash application within hours instead of days do not just save FTE cost. They tighten the standard deviation on the AR assumption that feeds the indirect model, which directly improves month 6 through month 12 forecast accuracy.
Use indirect when the horizon exceeds about 90 days. For annual budgets, 12 to 18 month rolling forecasts, covenant scenarios, M&A models, and five-year strategic plans, indirect is the only method that scales. Use direct for the 13-week tactical window where invoice-level visibility is real.
Not if treasury owns liquidity. Indirect cannot tell you whether you need to draw the revolver on Tuesday. The two methods serve different decisions: indirect for strategic planning and covenant management, direct for intraweek funding and short-term investment laddering. Mature treasuries run both and reconcile them.
Benchmarks suggest 92 to 95 percent accuracy at month 1, dropping to about 85 percent by month 6 and below 75 percent by month 12 in volatile sectors. Accuracy improves materially with monthly re-baselining, explicit assumption versioning, and tighter AR data feeding the working capital block.
DSO, by a wide margin. On a 500 million euro revenue base, every DSO day equals roughly 1.37 million euros of cash. AR is also the most volatile working capital component because it is downstream of collections velocity, dispute resolution speed, and customer payment behaviour, all of which move quarter to quarter.
Monthly at minimum, with rolling actuals replacing the most recent forecast month and a fresh month added at the back end. Quarterly refresh is too slow: the model drifts 8 to 12 percent within a quarter in normal conditions, more in volatile ones. The refresh should include re-running working capital sensitivities, not just dropping in new actuals.
They change the inputs, not the math. Agentic cash application and AI-native AR platforms tighten the standard deviation on the AR balance that feeds the indirect model. That alone can lift month 6 through month 12 accuracy by 5 to 10 points. The reconciliation step (direct versus indirect at the 90-day mark) is also a natural automation target.
