Building a 13-Week Cash Forecast People Actually Trust

Profitable companies run out of cash regularly. It is not a paradox — it is arithmetic. When revenue grows thirty percent, receivables and inventory grow with it, and that capital leaves the bank well before the profit arrives.

The tool for seeing this coming is a rolling 13-week cash forecast. It is not complicated. What makes it work is not the model but the discipline of running it every week until the variance narrows enough that people believe it.

Why thirteen weeks

A quarter is long enough to see a real problem forming and short enough to forecast at weekly precision using actual data. Within thirteen weeks, most of what drives cash is already known: invoices are issued, purchase orders are placed, payroll dates are fixed, debt service is scheduled.

Push the horizon further and weekly detail becomes assumption rather than information. That longer view belongs to the monthly forecast, which is a different instrument answering a different question.

What goes in it

Start with the opening bank balance, then build weekly rows for the categories that actually move cash:

Collections. Not revenue — collections. Take the open receivables aging and lay each invoice into the week you expect payment, based on that customer’s actual behaviour rather than stated terms. A customer on net 30 who reliably pays in 47 days goes in week seven.

Disbursements. Open payables laid into the week you intend to pay them, plus recurring commitments: rent, insurance, subscriptions, utilities.

Payroll. Including the payroll tax deposits that follow, which are frequently forgotten and are never small.

Debt service. Principal and interest on the actual due dates, plus any revolver activity.

Capital expenditure. Committed purchases with expected payment timing.

Taxes. Estimated payments, sales tax remittances, property tax.

That is the whole model. Most versions live in a workbook, sourced from ERP data, and are entirely legible to anyone who wants to check the logic — which matters more than it sounds.

The part everyone gets wrong

The forecast fails when collections are estimated from terms instead of behaviour.

If you assume everyone pays on terms, week four looks fine and then the money does not arrive. Two weeks of that and the leadership team stops believing the model, which is fatal — a forecast nobody trusts has no function.

Build collections from the actual aging, customer by customer for anything material. Your largest twenty customers usually account for most of the receivable balance, and you or your controller already know how each of them behaves. Encode that knowledge instead of a policy nobody follows.

Making it credible

Run it every week without exception, and compare last week’s forecast to what actually happened. Not to assign blame — to find out which assumption was wrong.

Early forecasts can miss materially. Compare expected and actual receipts and payments each week, document the cause, and update the assumptions. Reliability should be assessed against the business’s own history; there is no universal accuracy percentage or six-week improvement guarantee.

That improvement curve is the whole point. The forecast becomes credible through repetition, not through sophistication.

What it changes

Once the forecast is trusted, three things shift.

Payment timing becomes a decision. Instead of paying whatever is in the queue on Friday, you can see that week nine is tight and sequence disbursements around it.

Collections get attention before they are a problem. When you can see that a slow-paying account creates a shortfall in six weeks, the collection call happens now rather than after the shortfall.

Covenant conversations happen early. If the forecast shows a projected balance approaching a covenant threshold, you have weeks to act — and a lender conversation held eight weeks ahead of a potential breach is a completely different conversation from one held the week it happens.

A note on tooling

You do not need software for this. A well-structured workbook sourced from ERP data can be a practical starting point, provided ownership and review are clear.

Automation tools help once the discipline exists and the assumptions are stable. Buying one first tends to produce an automated version of a forecast nobody believes.

Start with the aging report, a blank workbook, and a standing thirty minutes on the calendar every week. Use the first several cycles to test assumptions and decide what leadership can reliably use.