Introduction
Overtime, rework, downtime and freight are real costs. If they only appear in the annual accounts, you are managing blind.
The annual costing problem
Most small manufacturers cost their products once a year. The sheet is built carefully, it is accurate on the day, and it decays from that point onward as material prices, labour rates and process efficiency all move.
By the time the annual accounts reveal that margin has drifted, three or four quarters have passed and the causes are hard to reconstruct.
What gets left out
Four costs are routinely absent from per-part costing because they are inconvenient to attribute.
- Overtime caused by a specific job or a specific breakdown
- Rework and rejection, often recorded as a quality statistic rather than a cost
- Machine downtime, absorbed into overhead rather than charged to the affected run
- Inbound and outbound freight, particularly on expedited shipments
Individually each is small. Together they are frequently the difference between the margin you think you have and the margin you have.
Costing from live data
If labour is captured against job cards, machine hours against runs, and material issues against batches, per-part cost becomes a calculation rather than an estimate. It also becomes current, which matters more.
The useful output is not a more precise annual number. It is knowing in week three that a particular job is running over.
Discounting is a cost too
Margin analysis that ignores the sales side is half an analysis. Orders discounted beyond the approved band are a direct margin cost, and in most companies nobody counts them because approval happened verbally.
Enforcing discount bands at quote time — with the margin visible to the person quoting — usually recovers more than any process improvement on the floor.
Asking the direct question
The most useful thing to be able to do is ask why margin moved and get an answer that spans purchasing, production, maintenance and sales at once. That is a data architecture question before it is an analytics question.
In summary
None of this requires a transformation programme. It requires deciding where each piece of data is created, making sure it is only created once, and letting everything downstream read from that record instead of keeping its own copy.
Meera Krishnan, Marketing Lead at easyto.work.
