Comparing forecast with actual receipts...
| Week | Period | Forecast | Actual | Difference | Diff % | Cum. forecast | Cum. actual | Paid on newer invoices |
|---|
Calculating forecast...
Each row answers three questions: how does this customer pay, what is open, and in which week do we expect the money.
| Class |
Structural pays predictably (at least 6 invoices, small spread); always uses its own median. Normal enough history, normal spread. Irregular large spread, or the last 3 months clearly slower than before. Little history fewer than 6 paid invoices; uses the profile of its DBUnit (or its own, if slower). |
|---|---|
| # Paid | Paid invoices in the history period ( months). |
| Median | Half of the invoices were paid within this many days after the due date. Negative = paid early. |
| P75 | 75% of the invoices were paid within this many days after the due date. |
| Used | The delay this forecast uses: , plus buffer days, plus the AI correction if applied. |
| AI | +20d 75% means: expect 20 days extra delay and that 75% of the balance will be collected. ok means AI reviewed the customer and agrees with the statistics. Hover over the badge for the reason. A large gap between median and average delay (in the detail screen) usually means a few large invoices are paid very late. |
| Balance | Net open balance: all open items added up (invoices minus credit notes and unapplied payments). |
|---|---|
| Overdue | Only the overdue invoices, without credit notes or payments. It can therefore be higher than Balance. Credits are netted against the customer's earliest expected receipts. |
| W.. columns | Expected receipt per week, already corrected for the chance of payment. Expected date = due date + Used. Items already past that date are expected in week 1 to 4 after the reference date; the further past, the later. |
|---|---|
| Later | Expected after the last week of the horizon ( weeks). |
| Doubtful | The part not expected to come in: items overdue more than days count for %, more than days for 0%, and the AI collectability percentage on top. |
| Total horizon | Sum of the week columns. |
Check: W columns + Later + Doubtful = Balance (for customers with a positive balance).
A customer is Normal with median +3d and P75 +5d. AI says +20d 75%, so Used = 5 + 20 = +25d. Open invoices 200.000, credit notes and unapplied payments 130.000: Balance 70.000, Overdue can still be 175.000. Expected 75% × 70.000 = 52.500, spread over the weeks in which due date + 25 days falls; Doubtful 17.500.
Click a customer name for the open items with expected date, chance and amount, and a chart of days late per month.
| Backtest | Recalculates the forecast as it would have been made on a past reference date: the items open then and the payment history up to then, with the settings above. AI corrections are not applied. |
|---|---|
| Snapshot | The forecast as it was stored that day (every Monday at 06:30, or with "Save snapshot now"), with the default settings. "With AI" includes the AI correction of that moment. |
| Actual | What came in on the items that were open on the reference date, in the week each item was closed. Receipts on invoices posted later are shown separately. |
| WAPE | The weekly differences (absolute) added up, divided by what actually came in. 20% means the weekly figures were on average 20% off; a negative total difference means the forecast was too optimistic. |
| Document | Type | Entry date | Due date | Age | Amount | After netting | Chance | Expected date | Expected |
|---|