Two quotations for PCB assembly services can both say “100% tested” while describing different work. One supplier might check that every board starts up. Another might apply signals to the inputs, run outputs under specified loads, and exchange data through the communication interfaces.

Both could test every unit. The evidence behind their pass results would be different.

In the purchase specification, “100% tested” should mean that every delivered assembly completes and passes the agreed test procedure. What that procedure can detect depends on its coverage—the faults and functions addressed by the inspection methods and electrical checks it contains.

The percentage becomes useful only when the buyer knows what each board must pass.

Establish What Happens to Every Board

First, separate bare-board testing from assembly testing. A bare-board electrical test checks continuity and isolation in the circuit board’s connections. It takes place before the subsequent assembly work and therefore provides no evidence about components or firmware installed later.

A quotation should identify which production stage its “electrical testing” charge covers.

Then check the quantities. If every assembly receives optical inspection but only five samples undergo functional testing, the order includes full-batch optical inspection and sample-based functional testing. Those commitments should appear separately.

Specify the inspection scope as well. “X-ray on every board” still leaves a question about which packages or solder joints are examined. An agreed program and inspection area give the supplier a defined task and the buyer something to verify.

Match the Method to the Fault

Automated optical inspection, or AOI, examines visible assembly features such as component presence, placement and exposed solder joints. X-ray inspection can examine joints hidden beneath packages such as ball grid arrays. These methods provide evidence about assembly construction rather than demonstrating the board’s operating performance.

In-circuit testing, or ICT, makes electrical checks for faults including shorts, opens, incorrect component values and certain orientation errors. Coverage depends partly on access to the relevant circuit nodes. Flying-probe equipment makes measurements without a dedicated bed-of-nails fixture, although the board design and test program still determine what gets checked.

Functional testing operates the powered assembly under defined conditions. Its scope might include supply measurements, input responses, loaded outputs and communication checks. The next section illustrates what those requirements could look like for a specific board.

A useful coverage review connects each method to the fault it is intended to find. For example, confirming that a connector is fitted and confirming that a signal passes through it require different evidence. The review should identify any gaps rather than assume that a long equipment list closes them.

Give Functional Testing a Measurable Scope

Consider a hypothetical 24 V industrial control board with a regulated 5 V supply, an analog input, a switched output and an RS-485 interface.

A startup check and an unloaded supply measurement would leave its input accuracy, loaded output behavior and external communication untested.

The table below shows the information needed to define those checks. It is a planning example, not a complete acceptance procedure.

Function	Details to establish before production
Power supply	Input voltage, applied load, measurement location, settling time and acceptable output range
Analog input	Signal levels, injection point and permitted measurement error
Switched output	Connected load, commanded states and acceptable output behavior
RS-485 interface	External test device, bit rate, cable and termination arrangement, test duration and permitted errors

Suppose the approved specification requires the 5 V rail to remain within ±3% at a 0.5 A load. The corresponding product limits are 4.85–5.15 V under that condition. “5 V present” would lose the requirement that matters.

The test engineer still needs to select suitable measurement equipment and define how results near those limits will be judged. A working sample can help check the setup; its measured values should not become the specification by default.

Include firmware identification in the procedure. Where manufacturing diagnostics use a separate image, require a final check that the approved delivery firmware has been installed and starts correctly.

These details also support an accurate quotation. PCBCool’s PCB testing service identifies product functions, firmware, connection methods and acceptance criteria among the inputs needed to establish functional testing. The scope should be settled before production begins.

Follow the Signal Through the Fixture

Return to the control board’s analog input. Suppose the fixture injects its signal at a test pad downstream of the input connector and protection network.

The downstream circuit could return the expected reading even with an open connection at the input connector. That result would support the measurement path actually used by the fixture, not the complete path available to the customer.

Mark the bypassed section in the coverage review and identify which other check, if any, addresses it.

Physical access deserves the same attention. Test-pad placement and nearby component heights affect probe contact. On high-speed signals, adding a branch for test access can also disturb the circuit being measured. These decisions belong in PCB layout review, while changes remain practical.

The program itself needs evaluation before release. Controlled fault simulation can reveal gaps in test patterns; suitable known-defective samples can provide further evidence. Passing a good board confirms only one side of the task. The intended defects also need to produce the expected failures.

Keep the First Failure in the Record

Suppose a hypothetical batch contains 500 assemblies. Of these, 480 pass the specified test immediately. The other 20 pass after documented rework and retesting.

Every board was tested, and every delivered board eventually passed. The first-pass yield at that operation was 96%.

First-pass yield excludes units that need rework or retesting to meet the requirements. Reporting only “500 passed” would hide the initial production outcome.

A retest can be legitimate. What matters is whether its reason is documented and the release follows an approved process. A confirmed fixture-contact problem should remain distinguishable from a repaired solder defect.

Agree on the records before the batch reaches testing. Each unit’s history should identify the test-program revision and relevant hardware and firmware configuration, with failed attempts retained alongside subsequent results. Test-data systems can link measurements to the unit, station, equipment and conditions used during each run.

For numerical checks, retain the measured value as well as the pass/fail result. In the 5 V example, that would allow a later reviewer to distinguish a reading near the acceptance boundary from one comfortably inside it.

Separate Production Testing From Qualification

Design qualification addresses whether a design meets its requirements under the relevant operating and environmental conditions. Production acceptance uses selected checks to establish whether manufactured units conform to that approved design.

For the hypothetical controller, a room-temperature test at 24 V provides evidence for that condition. Operation at other specified temperatures and supply voltages needs separate supporting evidence.

The plan should distinguish routine unit tests from sample-based checks and design-level verification. There is no basis for assuming that every delivered board repeats the entire qualification program simply because the quotation says “100% tested.”

Put the Procedure in the Order

An order should reference the approved procedure and revision, supported by the agreed inspection scope and reporting requirements.

An illustrative instruction could read:

Every delivered assembly must pass functional test procedure FT-017, Revision C. Results must be linked to the assembly’s serial number and identify the firmware and test-program versions used. Failed attempts, rework and retest results must remain traceable. Skipped tests or departures from the approved procedure require documented customer approval.

The procedure number is an example. Its real counterpart must contain the connections, conditions and limits needed to execute the test consistently.

Compare quotations against that scope, separating fixture and program-development charges from the recurring cost of testing each unit. Confirm which records will accompany the shipment and which will remain with the manufacturer.

A defined startup check gives the buyer a limited scope that can be evaluated. A documented functional test provides more evidence where the product requires it. The purchasing decision should rest on that evidence—not on the percentage printed beside “tested.”