Which job actually pays? The data layer that shows margin at quote time

The month closes, the statements come in, and one profit figure appears. If the figure is below expectation, the same question follows: which job did this? Usually there is no answer. Profit was measured across the sum of jobs, not on any job in particular.

This is not a failure of accounting. Accounting answers its own question correctly: what the company earned over the period. Sales and production ask a different one: what did this order earn, what does this customer leave behind, should we make this product again. The two questions do not feed from the same data. The first closes a period; the second chooses work.

Profitability is asked where a job starts, not where it ends.

The decision is made the moment a price goes out. The cost surfaces months later, in a lump. That lag is one of the most common reasons margin disappears. When a job is mispriced, seeing it requires the job to finish, the invoices to land and the period to close. By then several more quotes have gone out on the same reasoning.

So per-product profitability is not a backward-looking reporting exercise. It is a calculation that has to be placed inside a decision. And for that calculation to run, people first have to agree on what cost means.

The same cost items are always the ones forgotten.

Asked what a job costs, most people count two things: material and labour. Those are the visible items; they have invoices and timesheets. Margin usually melts in the items nobody counts.

Setup and changeover time. While a machine is being prepared it produces nothing but still consumes time. Die changes, calibration, first-article approval. On small batches this can run longer than the production time itself. If the per-unit cost does not divide that time across the batch, small orders will look cheaper than they are — systematically.

Rework and scrap. Machining a part a second time adds to the cost of the first pass. But the record usually says “made,” not “made twice.” Scrap gets buried in a general rate the same way. Which job the scrap came out of is part of that job's real cost.

Freight and shipping. A rushed shipment does not cost what a planned one costs. Neither does a half-empty truck. The difference accumulates per customer and is usually charged to nobody.

Shared overhead. Rent, energy, management, maintenance. These belong to no single job, so they are either left out entirely or spread evenly by revenue. Both mislead. Spreading by revenue makes cheap-but-laborious work look profitable and expensive-but-easy work look weak. Which driver the allocation follows — machine hours, labour hours, shipments — is a choice, and it should be written down.

Without one definition layer, no two screens agree.

Inside a company the word “cost” usually carries three meanings. Sales' cost, production's cost and accounting's cost are different numbers. Each is internally consistent. Put side by side in a meeting, the argument moves off the number and onto the definition.

The fix is not more reports. It is a single definition layer: which items enter cost, which driver they are allocated by, and at what moment they become final. That definition is written once, and every screen reads from it. The cost on the quote screen and the cost in the month-end report come from the same formula. Where they differ, the reason for the difference is visible in the same place.

The definition layer is not really a technical task; it is an agreement. Software only fixes it in place and makes it the single source.

Three things are needed before a quote screen can show margin.

First, unit costs. Material unit price, machine hour rate, labour hour rate, setup time. These need to be current, not perfect. A dated cost with a known date beats no cost at all.

Second, the structure of the job. What steps the quoted thing passes through: which machine, how long, how many people, how much material. Most companies hold this knowledge, but it sits away from the quote — on a drawing, in an old work order, in somebody's head. It has to come onto the same screen.

Third, feedback. When the job closes, the actual cost returns to the same record. Where the time assumed in the quote differs from the time spent on the floor, that difference is recorded. The next quote then rests on measured deviation rather than remembered instinct.

With those three, the quote screen stops being a calculator. As material, time and labour go in, the margin line updates. When a discount is applied, where the margin lands shows immediately. The decision is made before the price leaves the building.

Where to start.

None of this requires standing up full cost accounting. Pick one product family or one group of customers. Write the cost definition for that scope, collect the items, and let the quote screen work for that scope only. When the definition holds, extend it unchanged.

At Miletus the work is built in this order: define the cost, hold the definition in one place, then run the calculation at quote time. Profitability stops being a result learned at month end and becomes a number sitting on the screen while the price is being set.