Skip to content

Home / Guides

Building a Test Strategy for Box Build: From ICT to Outgoing AQL

BoxBuild VN editorial team · 16 October 2026

Key points
  • A box-build test strategy layers board-level tests (ICT, flying probe, boundary scan) under unit-level functional test, burn-in, safety tests and final inspection.
  • Functional test fixtures and software are normally paid by the buyer as NRE, and the test specification you write decides what they check.
  • Outgoing inspection is usually sampled to ISO 2859-1 AQL levels agreed in the contract, on top of 100 percent functional test.
  • Factories can support FCC, CE and RoHS pre-compliance, but formal certification is done by accredited labs and the legal responsibility stays with the brand owner or importer.

A box build is only as good as the last test it passed before it was packed. The PCBA may meet IPC-A-610, the harness may pass continuity and the enclosure may fit, but the buyer receives a finished product and judges it on whether it powers up, performs to specification and looks right out of the box.

This article walks through the test layers a box-build program typically uses: board-level tests, firmware programming, functional test, burn-in, safety tests, calibration, final inspection and outgoing sampling. It also covers the parts buyers most often leave undefined: who designs and pays for fixtures, what the test specification must contain and where factory pre-compliance support ends and formal certification begins.

The practices described are what EMS and box-build factories typically do. Each factory has its own methods, and the scope and cost must be confirmed in their quotes.

The layers of a box-build test strategy

Testing is layered so that defects are caught as early and as cheaply as possible. A fault found at board level costs minutes; after final assembly it means opening an enclosure; in the field it means a return.

Test layer What it catches Fixture needed Typical use
AOI and X-ray Placement and solder defects, BGA voiding No product fixture Every board, inline after reflow
Flying probe Shorts, opens, wrong or missing values No fixture, program only Prototypes, NPI, low volume
ICT (in-circuit test) Shorts, opens, component values, some powered checks Bed-of-nails fixture per board Stable designs at volume
Boundary scan (IEEE 1149.1, JTAG) Interconnect faults on BGAs and dense boards Connector access only Boards with limited probe access
Firmware programming Wrong or missing firmware, keys, IDs Programming jig or in ICT/FCT Every board or unit
FCT (functional test) Circuit and product-level function Custom FCT fixture and software Every board and/or finished unit
Burn-in or run-in Early-life failures, thermal issues Racks, power supplies, chamber 100% at launch or sampled later
Hipot and ground bond Insulation and earth faults in mains products Safety tester Every mains unit
Final visual and cosmetic Scratches, gaps, label errors, missing parts Inspection station, limit samples Every unit
Outgoing AQL audit Escapes from all of the above None Sample per lot

Not every product needs every layer. A battery-powered tracker may skip hipot entirely; an industrial controller with a mains supply cannot.

Board-level tests: ICT, flying probe and boundary scan

ICT uses a bed-of-nails fixture with spring probes that contact test points on the board. It measures component values, detects shorts and opens and can run limited powered tests in seconds. The limits are fixture cost, fixture lead time and the need for test points in the layout, which must be designed in, not added later.

Flying probe testers move a small number of probes across the board under program control. No fixture is needed, so it suits prototypes, new product introduction and low-volume runs. Each board takes minutes rather than seconds, so it becomes a bottleneck at volume.

Boundary scan, defined in IEEE 1149.1 and commonly called JTAG, uses test logic built into compatible ICs to check interconnects that probes cannot reach, such as BGA-to-BGA nets. It needs only a connector and is often combined with flash programming through the same port.

A typical path: flying probe during prototypes and pilot runs, then ICT once the design is frozen and volume justifies the fixture, with boundary scan wherever access is limited.

Firmware programming and serial number logging

Programming is easy to overlook until it becomes the bottleneck. Decide:

  • Where it happens: pre-programmed devices from the distributor, in-system programming on a jig, or as part of ICT or FCT.
  • What gets written: application firmware, bootloader, unique IDs, MAC addresses, calibration data, security keys.
  • How it is verified: checksum or version readback, logged against the serial number.
  • How versions are controlled: a released firmware file with a version and checksum, and a written rule for what happens when you release a new one mid-production.

Secure keys and certificates deserve their own procedure: who provides them, how they are stored at the factory and how unused keys are accounted for.

Serial number tracking ties every other test together. The common practice is to print or laser a serial label early, scan it at each test station and store results in the factory’s production database. Ask for the data in an agreed export format per lot, so you can trace any field return back to its test records.

Functional test (FCT) fixtures

FCT checks that the board or finished unit performs its function: power rails in range, interfaces communicating, sensors reading, radios transmitting, displays and audio working, buttons and LEDs responding. It is where your product knowledge matters most, because only you know what ‘working’ means.

An FCT station typically has three parts: a fixture (pogo pins for boards, or mating connectors and clamps for finished units); instruments such as power supplies, meters and RF testers; test software that sequences the steps, applies limits and records results against the serial number. Board-level FCT and unit-level FCT are often both used: the first catches faults before assembly, the second confirms the complete product including harnesses and mechanical parts.

Burn-in and run-in

Burn-in and run-in aim to find early-life failures, the left side of the reliability ‘bathtub curve’, before the customer does.

  • Run-in operates the product powered, often cycling through functions, at room temperature for a set period.
  • Burn-in adds stress, commonly elevated temperature and sometimes power cycling, to accelerate early failures.

Durations commonly quoted range from a few hours to 48 hours, depending on the product and risk. These figures are indicative; the right duration comes from your reliability data and must be priced by the factory, because racks, power and floor space scale with it. Many programs run 100 percent run-in at launch and move to sampled burn-in once failure data are stable.

Hipot, ground bond and other safety tests

Mains-powered products, and products with external power supplies built in house, need production-line safety tests. The usual set is dielectric withstanding voltage (hipot) between mains circuits and accessible parts, and ground bond or earth continuity for products with a protective earth. Test values and durations come from the product safety standard (for example IEC 62368-1 for audio, video and IT equipment) and the certification body. Certification bodies’ factory follow-up programs commonly require these routine tests and check the records during factory inspections.

Calibration with data logging

Products with sensors, measurement or RF output often need calibration: the station measures the unit against reference instruments, calculates coefficients, writes them to the device and verifies the result. Two requirements belong in the spec: reference instruments must themselves be calibrated with traceability, typically through ISO/IEC 17025 accredited calibration labs, and every coefficient and verification result must be logged against the serial number.

Final visual and cosmetic inspection

Cosmetic disputes are among the most common quality arguments in box build, because ‘no visible scratches’ means different things to different people. A usable cosmetic specification defines surface zones (for example A for the front face, B for sides, C for hidden surfaces), the allowed defect size and count per zone, viewing distance, viewing time and lighting. Approved limit samples, signed by both sides, settle most questions faster than any written rule.

Final inspection also checks labels, country-of-origin marking, accessories, manuals, packaging inserts and the retail box.

Outgoing inspection and AQL sampling

On top of 100 percent functional test, finished lots are usually audited by sampling. The common reference is ISO 2859-1, sampling procedures for inspection by attributes, which sets sample sizes and accept/reject numbers based on lot size, inspection level and AQL. Commonly quoted settings at general inspection level II are zero accepted critical defects, 0.65 or 1.0 for major and 2.5 or 4.0 for minor, but the values must be agreed per product. The defect classification, what is critical, major or minor, belongs in the inspection specification. Independent third-party inspection is an option many buyers use in addition.

Pre-compliance vs formal certification

Factories can often support FCC, CE and RoHS pre-compliance: EMC scans in a chamber or with near-field probes, XRF screening of materials for restricted substances, collecting supplier declarations and preparing samples. That is useful for catching problems early. It is not certification.

  • FCC: equipment authorization runs through accredited test labs, with a telecommunication certification body (TCB) granting certification; some devices may use the supplier’s declaration of conformity route instead.
  • CE: the manufacturer or importer draws up the technical file and signs the EU declaration of conformity based on test reports; a notified body is involved only for some directives and conformity routes.
  • RoHS: compliance with Directive 2011/65/EU as amended, supported by supplier declarations, screening and lab tests where needed.

Legal responsibility for compliance stays with the brand owner or importer, not the assembly factory.

Who designs and pays for fixtures

Test fixtures and software are product-specific, so they are normally quoted as NRE, separate from the unit price. Common arrangements: the factory designs and builds to your test specification; you supply a finished fixture and software; a third-party test house builds it. Commonly quoted budget ranges run from a few thousand USD for a simple FCT fixture to the tens of thousands for ICT or complex RF test; these are indicative only and must be confirmed by factory quotes. The cost article, box build cost in Vietnam, shows how NRE sits beside unit cost.

Agree in writing who owns fixtures and software, whether they can be moved to another factory and who pays for maintenance and spare probes.

What to define in the test specification

Section Content
Scope Which tests at which stage, 100% or sampled
Test steps and limits Each measurement with nominal, limits and units
Equipment Instruments, accuracy, calibration requirements
Firmware File, version, checksum, programming method
Burn-in or run-in Duration, temperature, power profile, pass criteria
Safety tests Hipot and ground bond values from the safety standard
Data Fields logged per serial number, export format, retention
Failure handling Retest rules, failure analysis, reporting
Cosmetic and AQL Zones, limit samples, defect classes, AQL levels
Ownership Fixture and software ownership, change control

For the rest of the RFQ package, see the box build RFQ checklist. For workmanship classes see IPC-A-610 classes, and for cable tests see wire harness assembly in Vietnam. When vetting a factory, ask to see a live FCT station and a sample test record, as described in how to vet an EMS provider in Vietnam.

boxbuildvn.com is a matching service, not a factory, and does not run tests. Send your BOM, drawings and test specification through the box-build RFQ form and get matched with up to 3 Vietnamese EMS factories within 24 business hours. Free for buyers, NDA on request. More on scope is on the box build assembly in Vietnam page, or go straight to the RFQ form.

Frequently asked questions

Who pays for functional test fixtures in a box build program?

In most programs the buyer pays for product-specific fixtures and test software as NRE, quoted separately from the unit price. The factory may design and build the fixture to your test specification, or you may supply it. Agree in writing who owns the fixture, whether it can be moved to another factory and who maintains it. Budget figures are indicative and must be confirmed by factory quotes.

Is burn-in necessary for every box build product?

No. Burn-in and run-in add time, space and energy cost, so they are used where early-life failures are expensive: power supplies, products with high thermal load, industrial controllers or anything installed in hard-to-reach places. A common approach is 100 percent run-in during launch, then a move to sample-based burn-in once field and test data show a stable failure rate.

Can a Vietnamese EMS factory certify my product for FCC or CE?

No. Factories can support pre-compliance, such as EMC scans, RoHS screening and preparing samples, but formal FCC equipment authorization runs through accredited test labs and certification bodies. For CE, the manufacturer or importer signs the declaration of conformity based on test reports, with a notified body involved only for some directives and routes. Legal responsibility stays with the brand owner.

What AQL levels are commonly used for outgoing inspection?

Commonly quoted levels under ISO 2859-1, general inspection level II, are zero accepted critical defects, 0.65 or 1.0 for major defects and 2.5 or 4.0 for minor cosmetic defects. The right values depend on your product and market. Define what counts as critical, major and minor in the inspection specification, otherwise the AQL numbers have no practical meaning.

What is the difference between ICT and flying probe testing?

Both test the assembled board for shorts, opens and component values. ICT uses a custom bed-of-nails fixture and tests a board in seconds to a minute, which suits high volume but needs fixture NRE and test points. Flying probe uses moving probes and no fixture, so it suits prototypes and low volumes, but each board takes several minutes. Many programs start on flying probe and move to ICT as volume grows.