What is a 13-week cash flow forecast, and why 13 weeks?

Structure: one column per week, receipts on top, disbursements below, ending cash and revolver availability at the bottom. The AP inputs, in order of importance: scheduled payments (weeks 1 - 2), open approved invoices with modeled timing (weeks 2 - 5), received-but-unapproved invoices with approval-lag assumptions (weeks 3 - 6), the recurring-payment register (all weeks), and open PO conversion estimates (weeks 5 - 13). Payroll, debt service, rent, and taxes come from their own calendars. The discipline that makes it work: every week, record actuals against last week's forecast, explain the variances, and roll the model forward - variance review is where forecasts improve.

Automate the data pulls (open payables, scheduled payments, bank actuals), keep the model structure stable so updates are data-refresh rather than rebuild, and timebox the variance review to the ten largest misses. Teams that spend a day on it are re-assembling data manually - the model isn't the problem, the plumbing is.

Direct method always (cash in, cash out - no accrual adjustments); record actuals weekly and track forecast-vs-actual by line; investigate variances over a set threshold; roll the forecast forward every week so week 14 becomes week 13; and keep one owner accountable for the whole model with input owners per line.

You're probably forecasting weeks 5+ from due dates on invoices you haven't received yet - i.e., from hope. Weeks 5 - 13 need different inputs: PO commitments, recurring-payment registers, and category run-rates with seasonality. Also check for systematic bias: if week-6 forecasts always miss in the same direction, the assumption is wrong, not the world.

Convert open PO value to expected invoices using historical PO-to-invoice lag by category, then apply payment terms for cash timing. Add known unbilled commitments (contracts without POs) from a maintained register. It's modeling, not measurement - label it as such and tighten the conversion assumptions with each variance cycle.

Commonly cited practitioner targets: within roughly 5% on total disbursements in weeks 1 - 2, within 10 - 15% by weeks 3 - 6, and directional (15 - 25%) in the back weeks - with accuracy improving over the first few months as variance reviews tune the assumptions. Track your own accuracy by week rather than chasing generic targets; the trend matters more than the level.

Excel is fine - and standard - while one person can maintain it and data assembly is under an hour weekly. Graduate when entity count or volume makes assembly the bottleneck, when multiple stakeholders need live views, or when key-person risk becomes real. The upgrade priority is automating the AP-side inputs, since that's where the manual work concentrates.

Granularity and governance tighten: daily detail for weeks 1 - 2, named owner per disbursement line, a payment-prioritization tier list (payroll, taxes, critical vendors, everyone else), and the forecast becomes the agenda of a weekly cash meeting with authority to delay or release payments. In a crunch, the 13-week model stops being a report and becomes the operating system.

The AP side of a 13-week model is only as good as your visibility into the payable pipeline. Stampli provides real-time status on every invoice and payment - received, in approval, approved, scheduled, paid - so the weekly refresh draws from live workflow state instead of a reconstruction. Because Stampli AI processes and validates the underlying transactions against ERP structure with human review, the disbursement data feeding the model is the same data that will post - not a parallel spreadsheet that drifts from reality.