Yield management: knowing how much good product a process really gives you

Fix the definition at the step or the number means nothing

Three measures matter and they are not interchangeable. Output yield counts good units leaving a step against units entering it, which is what capacity planning needs. First-pass yield counts units that passed without rework, which is what tells you how capable the process is. Material yield counts usable output mass against input mass, which is what matters where the input is expensive and the loss is offcut, trim or evaporation. Write down which one each report uses, at which step, and stop the recurring conversation where an operations figure and a finance figure differ and nobody can explain why.

Rolling yield across a route exposes what step figures hide

Individually respectable step yields multiply into something much worse across a long route. A route of many operations, each losing a little, can deliver a strikingly low proportion of started material as good finished product, and each department can still show a creditable local number. Calculating the rolled figure for the whole route is often the moment a management team realises how much of its input never reaches a customer. It also reorders priorities honestly, because the step with the worst local yield is not always the one destroying the most value once material cost at that point is considered.

Standard yield is a commercial number as much as a technical one

The yield assumption embedded in the bill of material and the cost model determines how much material is planned, what the item is quoted at and what variance the plant reports. Set it optimistically and production runs short, buys expedited material and reports adverse variance every month. Set it pessimistically and the plant carries surplus, hides improvement inside the allowance, and quotes uncompetitively. Review standards against recorded performance on a defined cycle, with engineering and finance in the same room, and change them deliberately — the worst arrangement is a standard nobody believes that everybody still reports against.

Drift, and the ratchet of quietly accepted loss

Yield rarely collapses; it slides. A die wears, a supplier's material sits at the edge of specification, an operator adopts a compensating adjustment, a check is skipped because it never found anything. Each accommodation is reasonable and the cumulative effect becomes the new normal, defended as how the process runs. The defence is a control chart on yield by product and line, reviewed at a fixed interval against the standard rather than against last month, plus an explicit trigger for investigation. Comparing only with recent performance guarantees that a slow decline never registers as a problem.

Attributing loss to something a person can act on

A yield figure by itself directs no effort. The loss has to be split into categories with owners: setup and dial-in, in-process defect by mechanism, material out of specification on arrival, handling damage, trim and offcut inherent to the geometry, and testing consumption. Each has a different route to improvement and a different owner — engineering, purchasing, maintenance, product design. The category that most often turns out to dominate, and most often surprises people, is inherent geometric loss, which no amount of shop-floor discipline will address because it was designed in long before production started.

Frequently asked questions

Should rework count as good output when calculating yield?

Not in first-pass yield, which is the measure that shows process capability — counting reworked units there hides the defect that generated the work. Include them in final output yield, since they are saleable and capacity planning has to reflect what actually ships. Reporting both figures side by side is the point: a large gap between them tells you the plant is buying its quality with rework labour, which is a cost that rarely appears anywhere obvious.

How do we measure yield when the input is a continuous material?

Work in mass or length balance rather than unit counts: what entered the process, what left as saleable product, what left as recoverable offcut, what left as unrecoverable loss. The balance has to close, and where it does not, the gap is usually unmeasured moisture, dust, purge or unrecorded reprocessing. Building that balance for a single line is tedious and frequently uncovers a material stream nobody had accounted for at all.

Is a yield improvement worth more than a throughput improvement?

It depends on where the money is. Where purchased material dominates the cost of the product, a small yield gain typically outweighs a large output gain, and it needs no extra capacity. Where the plant is capacity constrained and material is cheap relative to conversion cost, throughput at the constraint wins. Cost the two in the same units before choosing, because improvement effort tends to flow toward whichever is easier to measure rather than whichever pays more.