PlantOps BI

On-Time Shipping

The newest month on an on-time chart often looks like the best one, and it is the least settled. On the demo orders, Aug 2026 reads 97.6% against 94.8% the month before, because 605 of its 1,411 lines haven't shipped yet.

fact_orders by order month, data through 2026-08-31. "Shipped only" leaves open lines out. "Decided" also counts open lines already past their promise date, as late.

Order monthLinesOpenOpen past promiseOn time, shipped onlyOn time, decided
May 2026 1,571 0 0 94.8% 94.8%
Jun 2026 1,527 0 0 95.0% 95.0%
Jul 2026 1,461 0 0 94.8% 94.8%
Aug 2026 1,411 605 14 97.6% 96.0%

Why the Newest Month Flatters

The lines that ship first from a young month tend to be the ones that went smoothly. The ones still open include every line that is stuck. Leaving them out is a survivor count: Aug 2026's rate is built from its easy lines only. Of its open lines, 14 are already past their promise date on 2026-08-31, so they will be late whatever happens next. Counting them brings the month to 96.0%.

That is still provisional. The other 591 open lines aren't due yet, and some of them will miss. Show the open share beside the rate, or grey out months that are still filling in, so nobody reads a young month as a good one.

Order Month or Ship Month

The same orders tell a different story depending on which date sits on the axis. By order month, the 2025 dip lands in May 2025 and Jun 2025. By ship month, late lines from those orders ship later, so the worst month moves to Jun 2025 (83.8%) and the recovery looks slower.

On-time share of shipped lines by order month (solid) and by ship month (dashed).
80%85%90%95%100% 202420252026

Neither is wrong; they answer different questions. Order month asks how well we kept the promises we made that month. Ship month asks how the dock performed that month. Pick one per visual and say which in the title. Mixing them across pages is how two reports end up disagreeing about the same quarter.

The DAX

The measures count decided lines rather than shipped ones, using a cut-off taken from the data itself rather than TODAY (), so a report refreshed last night and read this morning agrees with itself.

On-Time Lines =
CALCULATE ( COUNTROWS ( fact_orders ), fact_orders[on_time] = 1 )

-- The last day the data covers, whatever the report is filtered to
Data Cutoff =
CALCULATE ( MAX ( fact_orders[order_date] ), REMOVEFILTERS ( fact_orders ) )

-- Lines whose outcome is known: shipped, or still open past their promise
-- date, which makes them late whatever happens next
Decided Lines =
VAR Cutoff = [Data Cutoff]
RETURN
    CALCULATE (
        COUNTROWS ( fact_orders ),
        KEEPFILTERS (
            FILTER (
                ALL ( fact_orders[ship_date], fact_orders[promise_date] ),
                NOT ISBLANK ( fact_orders[ship_date] )
                    || fact_orders[promise_date] < Cutoff
            )
        )
    )

On-Time Ship % (Decided) = DIVIDE ( [On-Time Lines], [Decided Lines] )

-- Ship-month view: the order date relationship is the active one, so switch
-- to the inactive ship date relationship for an axis on dim_date by ship month
On-Time Ship % by Ship Date =
CALCULATE (
    DIVIDE ( [On-Time Lines], CALCULATE ( COUNTROWS ( fact_orders ), NOT ISBLANK ( fact_orders[ship_date] ) ) ),
    USERELATIONSHIP ( fact_orders[ship_date], dim_date[date] )
)

-- Show it next to the rate, so a young month reads as provisional
Open Lines % =
DIVIDE (
    CALCULATE ( COUNTROWS ( fact_orders ), ISBLANK ( fact_orders[ship_date] ) ),
    COUNTROWS ( fact_orders )
)

KEEPFILTERS keeps a filter already on the fact's own ship_date or promise_date column instead of replacing it. A filter from dim_date arrives through the relationship and is never removed by the FILTER. The ship-month measure assumes an inactive relationship from fact_orders[ship_date] to dim_date[date]; every line it counts has shipped, so no open lines need deciding. The figures come from the demo CSV when the site is built; the tests check that every line is counted once and that counting overdue open lines never raises a month's rate.

On-time delivery KPI · OTIF · About the demo dataset