The question this page answers
How much of my forecast can I trust, where is it weakest, and is it getting better? Forecast engine status was the page for that question. It is hidden for now while it is redesigned, and the same record is read on Demand forecasting instead.
What you see
- 1The performance summary, scored against the by-hand method you pick.
- 2All methods opens the weekly error chart and the method table.
Nothing opens at this page's addresses today. Both of them land on Demand forecasting: the page's own address, and the old Forecast accuracy address, which pointed at it. The rail no longer lists Forecast engine status, the Home figure that opened it is not drawn on any plan, and Demand forecasting no longer carries a door to it. Forecasting runs as it always has, and so does the weekly scoring: every Monday the forecast is scored against what sold, and the choice of forecast approach for each product is re-run straight after.
On Demand forecasting, the comparison in the head says how the forecast's error compares with a by-hand method over the scored weeks. Pressing it opens the Forecast performance drawer, which is where that record is read today:
- The performance summary for the scored weeks, against the by-hand method you pick.
- All methods: the weekly line of the error ratio, then the method table. Tightly is the first row and every by-hand method follows it, scored on the same products and weeks. The columns are Method, Error per 100 sold, Units missed, Over / under and Error ratio. The row the comparison divides by is marked compared, and Tightly's own row reads Not compared under Error ratio, because every other row is divided by it. An Error ratio below 1 means Tightly missed less than that method did. The footing names how many by-hand methods were scored and the units sold that Units missed is worked out from.
- Forecast status: the forecast engine's own sentence, shown only where there is something to act on, such as a run that never ran, is stale, is partial or failed.
What to do
- Open Demand forecasting and press the comparison in its head.
- Read the method table to see whether Tightly beat each by-hand method on the same weeks, and by how many units.
- Press Explore in Ask to open Ask Tightly beside the page and ask where the forecast missed.
Nothing in the drawer is a write, and nothing you press changes a score: scores move only when the Monday run lands. A bookmark or a saved link to Forecast engine status still works; it opens Demand forecasting, so there is nothing to update.
When it is empty or failed
The Forecast performance drawer says why it has no record, in one line:
No week has been scored yet.
No scored weeks in this date range. Choose a different date range.
No full week measured yet.
While the record loads the drawer reads Loading performance. Where it could not be read it reads Performance unavailable, with Try again beside it.