The Cost of a Tighter Tolerance

A tighter tolerance is not a property of the part alone. It is a purchased service: the supplier must measure more, reject more, and document more to guarantee that the distribution of delivered units sits inside a narrower band. The price difference between a ±5 percent and a ±1 percent specification on the same nominal function is therefore mostly the cost of the work that the narrower band imposes, not a difference in the raw material consumed.

This is the same arithmetic that governs a margin in any operating business. What is paid for is not the value requested but the labor the request forces onto the process. A distributor that quotes a control component at one price and a second distributor that quotes the same function at twice that price is usually not describing two different markets; it is describing two different test and screening regimes behind the same part number. The gap is legible once the test flow is written down.

For readers who want the commercial version of this reasoning, a Canadian advisory site covering small and medium business management, Action Strategies, treats B2B pricing and margins as a question of cost structure rather than of asking price. The parallel is direct: a tolerance, like a price, is only defensible when the work behind it is named.

Where does the tolerance get paid?

The cost of a narrow tolerance appears at four stations, and each one is measured in a different unit. At incoming inspection, the supplier pays in gauge time and in the labor of an operator or an automated handler. At test, the supplier pays in seconds per unit multiplied by the number of units that must be tested, which for a screened specification approaches one hundred percent. At sort, the supplier pays in the handling of units that fail, including the ones that are reworked and retested. At scrap, the supplier pays in material that was purchased, processed and then discarded because it fell outside the band.

A specification written as a single number hides this structure. A specification written as a distribution, with a stated test method and a stated sampling or screening rule, exposes it. Two suppliers can agree on the nominal value, agree on the temperature range, and still quote prices that differ by a factor of two, because one is testing a sample per lot and the other is testing every unit at two temperatures.

Why do two suppliers quote the same function so far apart?

The divergence is usually traceable to five variables, none of which is the nominal part. The first is the test method: a bench measurement at a single frequency and a swept measurement across a band are not the same operation, and the second costs more per unit. The second is the screening level: lot sampling, AQL-based sampling, and one hundred percent test produce different escape rates and different labor content. The third is the yield at the specified band, which depends on the process distribution the supplier actually holds, not the one printed in a datasheet. The fourth is the documentation package, since a certificate of conformance, a test data sheet per unit, and a traceable calibration chain are three different deliverables. The fifth is the qualification burden, which is the cost of the screens the part must survive before it is allowed into the specification at all.

A buyer comparing two quotations for the same function is therefore comparing two test flows. The quotation that looks cheaper is often the one with the lighter flow, and the difference reappears later as incoming inspection, field returns, or a qualification failure that forces a redesign.

What does a screening screen actually cost?

The following table lists common screens applied to control components, the parameter each one constrains, and the cost driver that the screen introduces. Figures are given as ranges because the absolute cost depends on unit volume, handler throughput and the number of temperature setpoints.

Cost drivers for common screens, expressed as the parameter constrained and the operation that consumes time or material.

Screen	Parameter constrained	Cost driver
One hundred percent functional test	Insertion loss, isolation, switching	Handler time per unit, plus test equipment amortization
Temperature cycling	Parameter drift over range	Chamber occupancy, number of setpoints, dwell time
Burn-in	Early-life failure rate	Fixture count, power, duration, post-burn-in retest
Fine leak and gross leak	Package hermeticity	Helium consumption, operator time, reject handling
X-ray or acoustic inspection	Internal voids, die attach	Imaging time per unit, image review labor
Group qualification	Design margin	Sample destruction, test house fees, schedule

The right column is the part that a price quotation usually omits. A screen that consumes chamber occupancy for forty-eight hours has a cost that scales with the number of chambers available, not with the number of units inside them, so the per-unit cost falls steeply with volume and rises steeply when a lot is split.

How does yield loss enter the price?

Yield loss enters twice. It enters once as material that is purchased and processed and then discarded, and it enters again as capacity that is consumed by units that never ship. A process running at seventy percent yield inside a narrow band does not merely lose thirty percent of its material; it also occupies thirty percent of its test and handling capacity on units that generate no revenue.

Suppliers respond to this in one of two ways. They widen the process distribution, which requires tighter process control and often a capital investment, or they price the narrow band to cover the expected loss. The second response is more common at low volume, and it is the reason a specification that is easy to meet at ten thousand units per year can be expensive at one hundred. The buyer who asks for the tighter band without asking for the volume assumption is asking for a price that has no stable basis.