How to forecast collections and predict your cash flow

A collections forecast predicts how much of your open receivables will turn into cash, and in which week or month it will arrive. You build it by grouping open invoices by expected payment date, applying a realistic collection rate to each group based on how those customers actually pay, then updating the numbers as money lands.

The point is not a single tidy figure. It is a weekly view of incoming cash you can plan against, so finance is not guessing whether payroll, a tax payment, or a supplier run is covered. Done well, a collections forecast is the most reliable short-term cash predictor a business has, because receivables are cash you have already earned.

Why collections forecasting matters

Most cash flow misses come from the AR side. Revenue was booked, the invoice went out, and the cash simply arrived later than anyone assumed. A forecast that maps expected collections week by week closes that gap. It tells treasury what to expect, flags shortfalls early enough to act, and shows which accounts carry the most timing risk.

It also keeps collections honest. When you forecast a customer to pay in week three and they slip to week six, the variance points straight at an account that needs attention. The forecast becomes a worklist, not just a number for the board pack.

Inputs you need to forecast

You need four things, and the third is the one teams skip:

• Open AR by invoice. Every unpaid invoice with its amount, customer, issue date, and due date. Your aging report is the starting point.
• Your billing pipeline. Invoices not yet raised but scheduled, recurring contracts, milestones, and renewals. Cash you have not billed still belongs in a forward forecast.
• Payment behavior by customer. How each account actually pays relative to terms. A customer on net 30 who reliably pays on day 47 should be forecast at day 47, not day 30.
• Known events. Promises to pay, active disputes, deductions, and partial payments. These override the historical pattern for specific invoices.

Without payment history, you are forecasting off due dates, and due dates lie. The difference between forecasting on terms and forecasting on behavior is usually the difference between a forecast you trust and one you quietly ignore.

Building a collections forecast (with template)

Work invoice by invoice, then roll up. The simplest reliable model assigns each open invoice an expected pay date and a probability, then sums the expected cash into weekly buckets.

Expected cash (invoice) = Invoice amount x Collection probability
Expected pay date       = Due date + Average days late for that account

A worked example. You hold $400,000 in open AR across three segments:

• On-time payers, $150,000, who pay within terms. Forecast at the due date, 98% probability.
• Slow but reliable payers, $200,000, who run 15 days late on average. Forecast at due date plus 15, 95% probability.
• At-risk accounts, $50,000, already 60-plus days overdue. Forecast at 60% probability, spread over the next eight weeks.

Expected collections come out to roughly $147,000 + $190,000 + $30,000, or $367,000 of the $400,000 book, with the timing spread across the weeks each group is due. The $33,000 gap is your expected shortfall against face value, and the at-risk segment is where you focus collections effort.

A working template needs five columns: invoice, amount, expected pay date, collection probability, and expected cash. Add a sixth for the week bucket and pivot the expected cash by week. That weekly total is your forecast.

Accounting for payment behavior and risk

A flat collection rate across the whole ledger hides the accounts that matter. Segment instead. Group customers by how they pay, not by size, and assign each segment its own average days-late figure and probability. Reliable slow payers are predictable and low risk. Erratic payers and aging balances are where variance lives, so weight them down and watch them closely.

Layer known events on top of the segment defaults. An invoice in active dispute should not sit in the forecast at 95% until the dispute clears. A promise to pay moves a specific invoice to a specific date. Keeping a tight collections cadence and surfacing disputes early both make the forecast more accurate, because they shrink the pool of invoices behaving unpredictably.

Updating forecasts as cash lands

A forecast is only useful if it stays current. Re-run it weekly. Each week, mark what actually collected, compare it to what you forecast, and read the variance. Cash that landed early means a customer's behavior improved or your assumption was conservative. Cash that slipped means an account is running later than its pattern, which is a signal to act before it ages further.

Track forecast accuracy as its own metric. If you consistently over-forecast a segment, adjust its collection rate down. Over a few cycles the model tunes itself to your actual book, and the weekly variance shrinks. This feedback loop is what separates a forecast you act on from a spreadsheet that gathers dust. A faster collection process tightens the loop further, because cash lands closer to when you predicted it.

How Rex improves forecast accuracy

Rex forecasts off behavior, not due dates. Because it works every account on the ledger continuously, it knows how each customer actually pays, how they responded to the last reminder, and which invoices carry an open dispute or promise. It uses that live picture to predict payment timing invoice by invoice, then refreshes the forecast as cash applies and accounts age, so the number reflects today rather than last month's export.

The forecast and the work stay connected. When Rex predicts an invoice will slip, it does not just lower the number, it chases the account to pull the cash forward, and escalates the cases that need a human call. The forecast becomes both a prediction and a plan, with the routine follow-through already running underneath it. Lower bad debt follows from the same early action.

See how Rex turns your ledger into a live cash forecast and works the accounts behind it.