Weight each receivable
By that customer’s own payment history rather than by stated terms. A customer who has paid at 71 days for three years is modelled at 71, not at 30.
AI capability
Most cash forecasts age receivables on invoice terms, which assumes customers pay when they agreed to. They do not, and the variance is not random — each customer has a stable habit that predicts better than the contract does. Using the habit instead of the term is most of the accuracy improvement available.
What it does
By that customer’s own payment history rather than by stated terms. A customer who has paid at 71 days for three years is modelled at 71, not at 30.
Approved payment runs, open purchase orders, subcontract commitments, and recurring vendor charges — the money already promised but not yet moved.
The largest and most predictable outflow in most service businesses, modelled from the payroll calendar rather than smoothed monthly.
Subscription and retainer billing from the contract records, including ramps and known renewals or terminations.
Real balances across every bank account and legal entity, with intercompany movement shown separately rather than netted invisibly.
A lost account, a delayed collection, a hiring plan, a slower quarter — modelled against the real cost base rather than a percentage haircut.
A standard forecast takes each open invoice, adds the payment terms to the invoice date, and assumes cash arrives on that day. Every finance person knows this is optimistic, and most compensate with a blanket haircut — assume everyone pays fifteen days late.
That blanket adjustment is where the accuracy goes. Your customers do not deviate uniformly. One pays on day 28 reliably, another on day 74 reliably, and a third is genuinely erratic. Applying the same fifteen-day slip to all three is wrong three times in different directions, and the errors do not cancel because the amounts differ.
Every forecast rests on assumptions and the useful thing is to see which. Ours marks each line with its basis: behavioural where there is history, terms-based where there is not, contracted where it comes from an agreement, and assumed where somebody entered it.
The practical effect is that a CFO reviewing a thirteen-week forecast can see that forty-one percent of week nine rests on terms-based estimates for new customers, and discount accordingly. A forecast presented as a single confident line hides exactly the information needed to judge it.
Thirteen weeks is the standard operating horizon because it is long enough to act on and short enough to be meaningfully accurate. Beyond a quarter, the dominant variable becomes new business rather than collection timing, and that is a pipeline question rather than a cash question.
We produce longer views and label them as directional. A twelve-month cash forecast for a growing business is a planning artefact rather than a prediction, and presenting it with the same confidence as a four-week view would be misleading.
Limits
Every capability page on this site carries one of these, because a feature described without its boundaries is a claim rather than a description.
Pipeline is an input you weight, not a forecast it produces. Deals not yet signed are modelled as an explicit assumption you can turn off.
They are modelled on terms and flagged as an assumption. In a fast-growing business a large share of the book may be like this, which limits accuracy honestly.
Draws, covenants, and availability are yours to manage. We show the cash position that informs those decisions; we do not model the facility itself.
Questions
Six months of AR, AP, and bank data is enough to run a forecast as of six months ago and grade it.