North Carolina manufacturing intelligence

Sources observed Oct. 1–29, 2026

Supply · Original economic model

When assembly is fast and test is full

Two production routes can cost almost the same and consume very different amounts of the resource needed to ship. In the original scenario below, making 2,000 accepted circuit boards costs about $79,400 with flexible test and $79,900 with dedicated contacts and an automated test sequence. The flexible route needs 218 station-hours; the dedicated route needs 55. If the release window contains 80 available test-station hours, the small price difference conceals the decisive production difference.

That is a useful starting point for North Carolina manufacturers, buyers and people considering new capability. A busy test station can point toward test engineering, reliable contacts, programming or diagnosis before another assembly machine. Which intervention matters depends on the accepted output of the whole route.

This is CCN’s original hypothetical analysis. The quantities, costs, yields and times are assumptions, not supplier prices, observed Carolina demand or evidence of a local shortage. The mechanism is worth studying now; its inputs can be replaced as operating evidence becomes available.

A common board can become several different products

Imagine a small indoor controller with one circuit-board revision, one populated bill of materials and four firmware versions. Each version has a different approved timing or message behavior, while staying inside the same assumed power and input/output envelope. The campaign requires 500 accepted boards of each version. No radio, safety-critical function or certified environmental rating is assumed.

The physical commonality can preserve material purchasing, stencil, placement-program and solder-process work. Verification has its own boundary. Shared board geometry does not establish that the same stimulus, software state and measurement limits demonstrate each product’s behavior.

The reusable test interface therefore includes more than contact locations. It includes how the board is supported, powered and programmed; what states are reachable; which responses are measured; and which firmware, script and limits belong together. A connector moved on the board can require new contacts. A firmware change can require a new sequence without moving a component.

This is where late customization can consume the flexibility it preserved upstream. The factory may have a large common assembly batch and several smaller verification jobs.

Count attempts and test executions separately

Compare two candidate test architectures against the same acceptance plan:

  • Route A uses programmable probing plus connected manual functional work. Initial development costs $1,200; each complete test execution costs $6

  • Route B uses dedicated contact tooling plus an automated electrical and functional sequence. Initial development and tooling cost $9,200; each complete execution costs $2.40

  • Both routes share $32 of board, component, assembly and optical-inspection cost per attempted board

  • First-pass acceptance is 96%. Every initial failure receives one $12 rework and one full rerun; 75% of those failures recover

The four firmware scripts are included in initial development. Programming is included in each execution. Both routes are assumed to provide equivalent acceptance evidence, with no false acceptance or false rejection. That equivalence is a modeling condition, not a result established for a real product.

The final accepted fraction is 96% + 4% × 75% = 99%. Delivering 2,000 accepted boards consequently requires 2,000 ÷ 0.99 = approximately 2,020 attempted boards. Each attempt gets its first test, and the 4% that fail get another. Total executions are 2,020.20 × 1.04 = approximately 2,101.

Common assembly and rework consume ($32 + 0.04 × $12) × 2,020.20 = $65,616. Route A then adds $1,200 + $6 × 2,101.01, bringing its total to $79,422. Route B adds $9,200 + $2.40 × 2,101.01, bringing its total to $79,859. A is cheaper by $436 in this first campaign.

These are route costs, not selling prices. They exclude enclosure and cable assembly, whole-product qualification, freight, taxes, general overhead, financing, exceptional repairs and field failures. Fractional units are expected planning arithmetic; a real release needs whole units and allowance for variability.

The station can reverse the apparent choice

Now give A six station-minutes per complete execution and B 1.5 minutes. Include ordinary loading, programming and recording in those invented effective times. Add eight setup hours across the four versions on A and two on B.

A consumes 8 + 2,101.01 × 6 ÷ 60 = 218.10 station-hours. B consumes 2 + 2,101.01 × 1.5 ÷ 60 = 54.53 hours.

With 80 otherwise uncommitted workable station-hours, B fits the modeled test-stage window. A has 72 run hours after setup, enough for 720 executions and approximately 685 expected accepted boards under this rework model. Extra boards arriving from assembly would join the test queue.

The relevant alternatives include another staffed station, a faster equivalent test sequence, better fixture reliability, improved first-pass acceptance or a different release schedule. A second assembly line would matter only if assembly were also limiting accepted delivery.

Resource hours do not establish calendar lead time. Engineering development, fixture fabrication, material shortages, commissioning, repair labor and approval can consume the same window. A fast fixture delivered late may still miss the release. Transfer lots can let test start before the entire assembly campaign ends, but they do not erase the required station work.

Revisions and repeats change the commitment

In the first campaign, B commits $8,000 more and saves $3.60 per execution. With 1.04 ÷ 0.99 executions per accepted board, its cost crossover is approximately 2,115 accepted boards. That threshold belongs to this defined family and test scope.

Suppose a pre-production layout change has a 30% probability of requiring $3,000 more contact-tool rework on B than on A. The expected differential is $900, moving the crossover to about 2,353 accepted boards. A firmware-only revision is a separate scenario: physical contacts might survive while script, state or limit work changes.

A stable repeat has another comparison. If usable tooling and programs are already paid for and accessible, and both routes incur the same repeat release charge, B saves approximately $7,564 in execution cost at 2,000 accepted boards. The original development expense still belongs in lifetime economics. Tool ownership, maintenance, revision validity and access must survive the reorder.

Quality also changes resource demand. If first-pass acceptance drops from 96% to 92%, with the same 75% recovery, final yield becomes 98%. The campaign then needs approximately 2,041 attempts and 2,204 executions. A’s station load rises to 228.41 hours; B’s rises to 57.10. This controlled sensitivity does not assume automation improves yield. It shows why failures, diagnosis and reruns belong in capacity planning.

Price the capability that would remove the constraint

A possible buildout is a shared test-development, programming and release cell. Other possibilities are contracted engineering, customer-owned fixtures at an existing operation, or purchased test from a specialist. Each has different responsibility, lead-time and capital boundaries.

Consider an independent annual model. Assume an installed automated station costs $150,000 and incurs $60,000 annually including capital recovery, allocated staffing, space, maintenance, calibration and software. Compare purchased test at $6 per execution with the cell’s $2.40 variable cost, holding execution scope and quality equivalent. The installed cost is represented through capital recovery; it is not added again to annual expense.

At a $3.60 saving per execution, break-even is 16,667 annual executions, or approximately 15,865 accepted boards under the original yield and rerun assumptions. This counts compatible workload, not every electronics unit produced in the region.

Now suppose five new incompatible families each require an additional $8,000 fixture and development program during the year. Annual fixed burden becomes $100,000 and break-even rises to approximately 26,442 accepted boards. Family proliferation consumes engineering and tooling capacity even when station run utilization looks modest.

For a capacity check, assume 250 eight-hour days and 75% effective run availability: 1,500 run hours. At 40 executions per hour, the ceiling is 60,000 executions, approximately 57,115 accepted boards. The original break-even uses about 28% of that ceiling before additional family setups, commissioning and other omitted workload. A large theoretical ceiling says little about whether the right demand arrives in the right windows.

Preserve the evidence behind an accepted result

Testing has two questions: whether a recorded measurement is dependable, and whether the chosen measurements adequately support the product’s acceptance decision. NIST’s metrological traceability policy connects a measurement result to a reference through documented calibrations and their uncertainties. It also explains that traceability alone does not establish fitness for purpose; measurement uncertainty must suit the particular need. That supports keeping measurement provenance separate from test adequacy.

A useful production record should connect:

  • Board and populated-BOM revisions, including actual substitutions

  • Firmware identity, programming responsibility and allowed operating states

  • Contact, fixture, script and acceptance-limit revisions

  • Measured results, failure cause, repair action and rerun identity

  • Station workload, engineering hours and the release window

  • Ownership of acceptance and the scope left for whole-product qualification

Those are proposed intelligence fields, not a claim about a supplier’s system or a live CCN integration. They let a buyer, manufacturer or software agent ask what changed and which earlier evidence remains usable.

For a North Carolina route, compare a local cell, an existing service and a wider-state specialist using the same boundary: sample transfers, engineering iterations, freight, queues and responsibility for release. Distance can affect those costs; it cannot establish the best route by itself.

The useful supply question is precise: which missing capability would increase accepted deliveries for a recurring, compatible workload, and what cheaper alternative could do the same? Conditional models can expose that question before a live customer project exists. Observed demand, actual coverage and total resource costs determine whether the eventual answer is engineering, operational improvement, shared service or equipment.

Sources and model boundary

NIST, Policy on Metrological Traceability, policy items 1 and 5. Page created February 12, 2010; updated June 17, 2024; checked October 8, 2026. The factual paraphrase above concerns measurement traceability and suitability only. All campaign, revision, quality, capacity and buildout calculations are original CCN scenarios with explicit assumptions. No named supplier, current order, regional shortage or investment return is established.

Original analysis by Carolina Capacity Network, prepared with AI assistance.

Cite this page: Carolina Capacity Network, “When assembly is fast and test is full”, carolinacapacitynetwork.com/blog/assembly-test-constraint/ (sources observed Oct. 8, 2026).