Supply chain delivery
Three years of a retailer’s orders, rebuilt as a Power BI model: what sold, what it earned, and whether it arrived when promised. The premium shipping mode is late on every single order, and the reason is the promise, not the warehouse.
The promise against the delivery
Shipped orders only. Each bar is the share of that mode’s orders taking that many days; gold bars are past the promise.
Late rate by mode
| Mode | Promised | Average days | Orders | Late |
|---|
By department
| Department | Net sales | Margin | Lines at a loss |
|---|
About one line in five loses money in every department: a flat rate the report shows rather than explains.
The file changes generator in October 2017
The market is a calendar
Page 06: the late-delivery calendar
September 2017 had 941 late orders from 1,640 shipped, and not one day of the month without a late order. Late orders sit on the day the order was placed, with the orders shipped that day in the corner of each cell. Buttons zoom out to months, quarters and years. A rise paints red here, because more is worse.
- One module, four reports. The page is generated by the same Python module that draws the calendar in three other reports. Each report supplies a short config: the measure, the wording, which way is good.
- A part year is never the subject of a claim. The first build’s Year view read “2018 down 89.5% on 2015”. 2018 is one month of data. Year and quarter comparisons now use whole years only, and the title says so.
- The check found what no render would. All 1,181 cells are compared with pandas. The one mismatch was 2018’s single month shaded
as the lowest band:
DIVIDE(BLANK, 0, alt)returns blank, notalt.
