Thirteen Week Cash Forecast

Thirteen-Week Cash Forecast

When to use

Use whenever near-term cash visibility matters more than annual or monthly views — a business managing tight liquidity, preparing for a lender conversation, or simply wanting an early warning system for cash rather than finding out about a shortfall after it happens. A monthly cash view can show a comfortable average while hiding a specific week where the balance actually goes uncomfortably low.

What it does

Projects the cash balance week by week from actual and projected inflows and outflows via a bundled calculator (the direct method, tracking real cash movements rather than deriving cash from an accrual P&L), and flags the exact week and dollar amount by which the balance would breach a stated minimum cash threshold, if any week does.

Method

1. Build inflows and outflows from real, specific sources for each of the 13 weeks — actual collections expected (tied to the AR aging), actual payments scheduled (tied to AP aging and payroll dates), not a smoothed average of a monthly number divided by 4.3.
2. Use the direct method: track actual cash receipts and disbursements, not net income adjusted for non-cash items — the 13-week forecast's entire value is showing real cash timing, which an indirect, accrual-based method obscures.
3. State the minimum cash balance the business needs to maintain — whatever operational or covenant-driven floor actually matters, not zero.
4. Run the bundled calculator (scripts/thirteen_week.py) to get the week-by-week balance and an explicit flag for the first week (and every week) the balance would drop below that minimum.
5. Treat a breach flag as needing a specific action before that week arrives, not just an observation — draw on an available credit line, accelerate a specific receivable, delay a specific payable, or cut a specific planned outflow, named concretely rather than "we'll figure it out."
6. Refresh the forecast weekly, rolling the horizon forward, since a 13-week forecast built once and left untouched loses its main value, which is catching a problem with enough lead time to act.
7. Compare this week's actuals against last week's forecast for the same week as a standing accuracy check — a forecast that's consistently off in the same direction needs its underlying assumptions revisited, not just this week's numbers.

Inputs

• Starting cash balance
• Weekly inflows and outflows for the 13-week horizon, built from real AR/AP timing and payroll dates
• The minimum cash balance the business needs to maintain
• Config saved as JSON matching the format documented at the top of scripts/thirteen_week.py

Output format

Week-by-week table of inflows, outflows, net cash movement, ending balance, and headroom versus the minimum; explicit flag on the first breach week if any week's balance drops below the stated minimum, with the shortfall amount.

Example

A monthly cash view shows the business comfortably cash-positive all quarter. The 13-week direct-method forecast reveals that in isolation, week 9 through week 13 show declining headroom as a large payroll and a vendor payment cluster in the same window while a major customer payment lands a week later than usual, though in this run the balance stays above the $200K minimum throughout, ending week 13 with $235,980, still $36K of headroom, a genuinely tight but not breaching result the monthly view alone would never have shown with this level of week-specific detail.

Common pitfalls

• Building the forecast from a smoothed monthly average instead of real weekly inflow and outflow timing, hiding a specific tight week inside a comfortable average.
• Using the indirect, accrual-based method instead of tracking actual cash receipts and disbursements.
• Building the forecast once and not refreshing it weekly, losing the early-warning value that's the entire point of a 13-week view.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: Natan-Mohart
• Source: Natan-Mohart/24-finance-skills-for-claude
• License: MIT

Install and usage instructions live in the source repository linked above.