Why naive cash forecasts fail.
The naive method: take open AR, slot each invoice into the week it is due, add expected new sales by their due dates, subtract scheduled AP, plus payroll and tax. The result is wildly optimistic on the inflow side because customers do not pay when they are supposed to. By week 4 the forecast is off by 25 to 40 percent. By week 13 it is fiction.
The fix is to model actual collection patterns, not invoice due dates. If your last 12 months show that 30-day-term customers actually pay at an average of 47 days with 70 percent within 60 days, that is the pattern to model forward. The forecast becomes probabilistic but accurate, instead of deterministic and wrong.
The collection-curve method.
Step 1: pull the last 12 months of invoice-to-payment data. Step 2: bucket invoices by stated terms (immediate, 7 days, 14 days, 30 days, 60 days). Step 3: for each bucket, calculate what percent collected by week 1 after due, week 2, week 3, week 4, beyond. Step 4: apply this curve to current open AR plus expected new sales. The result: dollarized expected cash inflow per week for the next 13 weeks.
Worked example. A Toronto-based B2B distributor, 30-day-term customers. Historical curve: 15 percent paid by due date, 30 percent within 1 week of due, 25 percent within 2 weeks, 15 percent within 3 weeks, 10 percent within 4 weeks, 5 percent beyond. C$200,000 of 30-day-term invoices outstanding due in week 4: forecast week 4 inflow C$30,000, week 5 C$60,000, week 6 C$50,000, week 7 C$30,000, week 8 C$20,000, beyond C$10,000. Total adds back to C$200,000 but spread realistically.
Customer-level overrides.
The collection curve is the default. Specific customers override it. Customer A is a government department that takes 90 days reliably; customer B is a multinational that pays on day 30 to the day; customer C has been late and shaky for 3 months and probably pays in 60. Each gets an explicit collection assumption that overrides the curve.
Maintain customer-level collection profiles for the top 20 customers (typically 70-80 percent of revenue for SMBs). The remaining 20-30 percent of revenue uses the curve. This combination is more accurate than either approach alone, and it forces you to think about each big customer's payment behavior, which is good discipline regardless.
The 13-week forecast structure.
Inflows: cash from open AR (using curves and overrides), cash from expected new sales, other receipts (loan drawdowns, refunds, capital injections). Outflows: scheduled AP payments, payroll, taxes (income tax instalments, GST/HST, payroll remittances), rent, utilities, capex, debt service. Balance: opening cash, plus net inflows, equals closing cash. Refresh weekly, ideally Monday morning.
Why 13 weeks? It is long enough to see the next quarter's tax and dividend obligations, short enough that the forecast is meaningfully accurate, and matches the typical operating cadence of an SMB. Beyond 13 weeks, switch to monthly long-range forecast which uses different assumptions.
- Inflows from open AR using collection curves
- Inflows from expected new sales by week
- Other receipts (loans, refunds, capital)
- Outflows: AP, payroll, taxes, rent, capex, debt
- Balance: opening + net = closing per week
Stress testing the forecast.
A single forecast is a point estimate and gives a false sense of certainty. Run three scenarios: base case (your collection curves as-is), downside (collection lags by 2 weeks across the board), upside (collections improve by 1 week with active dunning). Watch the closing cash balance in each. The downside scenario is your real risk picture.
A practical rule: if downside-case minimum cash dips below 4 weeks of operating expenses, you are in a danger zone and should pull a lever — accelerate collections (offer early-payment discount), defer non-critical AP, draw on a credit line, or postpone capex. The forecast is not a prediction; it is a decision tool.
Common pitfalls.
Pitfall 1: forecasting in isolation from sales. The sales pipeline must feed the forecast. New deals expected to close get factored in at probability-weighted amount. Pitfall 2: ignoring tax timing. Payroll remittances are due monthly or biweekly in most jurisdictions. GST/HST/VAT is monthly or quarterly. Corporate tax instalments are quarterly. These are real cash outflows, not accounting events. Pitfall 3: stale assumptions. Re-baseline the collection curve every quarter using the latest 12 months of data.
Pitfall 4: treating the forecast as a one-person spreadsheet. The owner, finance manager, and key sales lead should all see the forecast and contribute. Sales knows which deals are slipping. Finance knows the obligations. Owner makes the calls. Keep one shared, version-controlled document, not multiple personal spreadsheets.
How Nonari builds the 13-week forecast.
Open AR feeds in automatically with stated terms. The collection curve is calculated from the trailing 12 months of payment data per term bucket. Customer-level overrides can be set on the customer master. Expected new sales pull from the pipeline if integrated, or from a sales forecast input. Scheduled AP, payroll, and tax obligations come from the bills, payroll module, and tax calendar.
The forecast refreshes daily as new invoices, payments, and bills land. Three scenarios (base, downside, upside) generate automatically. Alarms fire when downside-case cash dips below threshold. Owner gets the weekly summary on Monday morning. The forecast is decision-ready, not a quarterly exercise.