The Objectives Grid plans outcomes and the sensitivities say how drivers move them. This section is about the quality of the forward figures themselves — how wrong the forecasts have been, where payment risk sits in the book, which products are about to run out, and what the next thirteen weeks of cash look like on the evidence the business already holds.
Forecast error, at the grain you manage. Every forecast a promoted Analytics Lab duty publishes carries its own backtest error with it. That error is now a governed measure in its own right — forecast error — and it can be read by product and by horizon step, not only as one number for a whole run: the forecast for your best product three weeks out may be tight while the same duty's ten-week view of a slow line is loose, and the measure says so. Like any governed measure it can be targeted on the Objectives Grid, so "get forecast error under 12%" is a plan the business can steer by.
What is not served: forecast bias. Bias — whether the forecasts chronically run high or low — cannot be computed from what the Lab records. Backtests record unsigned error (how far out, never in which direction) at every grain, so BlossomERP names bias unavailable rather than approximating a signed figure from unsigned evidence. If a surface ever shows you a bias number, it will be because a signed record exists — never because one was inferred.
Payment risk is a score with its workings showing. Payment risk (over the open receivables book) is a governed measure scored 0–100, computed deterministically from the book itself: half the score is the overdue share of open AR, half is the amount-weighted average days past due across the whole book, with lateness capped at 90 days. The weights and the cap are declared on the measure itself. It is a MODEL ESTIMATE in the same sense as the likely-donor score: transparent arithmetic that ranks and warns — not a credit classifier, and it claims no accuracy. A customer with nothing open contributes nothing; an empty book scores 0, so read the score beside the open-invoice count, never alone. Ask can rank customers by it, and it can be targeted on the Objectives Grid.
Which products run out, and when. Ask "which products will run out?" and the answer composes four governed reads: the published demand forecast for each product (promoted Lab duties only — experiments never appear), the stock on hand and the quantities already on open purchase orders at the latest inventory snapshot. Walking the forecast forward against that cover gives each product's run-out date — the earliest date the forecast exhausts everything the business has or has ordered (on-order stock counts as available from today; supplier lead times are not netted, and the answer says so). Every product's line carries its forecast's own backtest error beside it. And the coverage is stated the way OTIF states it: the answer covers only stocked products that carry a published demand forecast, and it counts and names the stocked products it cannot assess rather than dropping them quietly.
The 13-week cash outlook — a filed duty, not a chart. Turn on 13-week cash outlook (Company → Features → Money) and a duty files the outlook every Monday at 06:00, your company's time, into the Document Library and by email to everyone holding view cash-outlook; the same page (Accounts → Cash Management → 13-Week Cash Outlook) files one on demand and holds the history. Thirteen weekly buckets, all in your functional currency:
- Opening cash is the ledger balance of your active bank and cash accounts — read from the general ledger itself, because that is where cash truth lives.
- Expected receipts place each open invoice at its due date plus that customer's average settlement delay over the last twelve months (the company average where a customer has no history — the filing states what share of the book carried its own history).
- Expected payments mirror it, at your average delay paying each supplier.
- Committed purchase orders place each approved order's un-invoiced value in the week of its required date. An order with no required date is excluded — and named in the filing's coverage, with its value.
- Payroll and other recurring outflows are declared assumptions: a monthly amount and a day of the month, declared on the outlook page by anyone holding update cash-outlook. Undeclared means no line is projected — the filing says so in plain words; the system never estimates a payroll for you.
Each week's projected closing is arithmetic — opening plus receipts, less payments, orders, payroll and recurring — becoming the next week's opening. Amounts expected beyond the thirteen weeks, and any amount whose frozen exchange rate cannot be resolved, are excluded and counted in the filing's coverage: the outlook always shows what it left out and why. Every filing freezes the assumptions and bases it was computed under, so a filing from March is judged by what March declared, not by today's settings.
Who can do what. Reading the outlook needs view cash-outlook (accountants, finance managers, managers and company admins hold it); filing on demand needs create cash-outlook and declaring assumptions needs update cash-outlook (managers and company admins). The scheduled duty files only while the company switch is on.