When a Thailand factory selects an inspection equipment manufacturer or system integrator, comparing camera resolution, quoted accuracy, cycle time and machine price is not enough. A useful RFP must also define what is measured, how a result becomes PASS or FAIL, whether each record can be traced to the product, recipe, equipment and time, and how performance will be maintained in production. In the first half of 2026, Thailand’s BOI reported 132 Smart and Sustainable Industry applications worth approximately THB 17.2 billion for machinery upgrades, digital technology and automation or robotics integration. As automation investment expands, inspection needs to be purchased as a measurement, decision, data and service system—not as an isolated machine.
This guide deliberately differs from our automated dimensional inspection guide. It focuses on selecting the equipment maker or SIer: the RFP, measurement system analysis (MSA), calibration, decision logic, data integrity, FAT/SAT and lifecycle support. The objective is to help quality, engineering and procurement teams compare proposals on the same basis and receive evidence that remains useful after mass production begins.
Select the business decision before selecting inspection equipment
Begin with the boundary of the investment decision, not the machine model. Inspection automation may support escape prevention, process feedback, traceable records, stable changeovers, customer audits and corrective action—not only labor reduction. The required measurement principle, handling, controls, data platform and service model depend on which outcomes become contractual acceptance conditions.
Put these questions at the front of the RFP:
- Who uses the result, and what decision does it authorize?
- Where will the risks of false acceptance and false rejection be controlled?
- Does the result trigger process feedback, segregation, or record creation only?
- Which characteristics require 100% inspection automation, and which remain audit or sample checks?
- Is PASS/FAIL enough, or are measured values, images, waveforms and reason codes required?
- Which responsibilities belong to the prime contractor, factory, fixture supplier and IT team?
Without those answers, bidders will price different systems under different assumptions. Apparent savings can later reappear as change orders for lighting, fixtures, masters, reject handling, networks, data interfaces or local commissioning.
Separate the component maker, machine builder and SIer roles
The term “inspection equipment manufacturer” may refer to very different scopes. A component manufacturer supplies cameras, sensors, optics or measurement software. A machine builder designs fixtures, handling, guards and panels. An SIer connects PLCs, robots, MES/QMS and databases. A metrology specialist may support calibration, uncertainty or measurement-method development.
| Role | Typical deliverables | Boundary to clarify in the RFP |
|---|---|---|
| Component manufacturer | Sensor, camera, optics, measurement head, software | Performance conditions, compatibility, lifecycle and technical support |
| Machine builder | Mechanics, fixture, handling, safety and control panel | Workholding, repositioning, maintainability, safety and drawings |
| SIer | Controls, decision logic, communications, data and commissioning | Interfaces, exceptions, audit trail and backup |
| Metrology/calibration body | Calibration, traceability, uncertainty and advice | Measurand, range, certificate and on-site/laboratory conditions |
| Factory | CTQs, boundary samples, process conditions and quality decisions | Master specification, approvers, material, training and change control |
Even when one company is the prime contractor, disclose its subcontract structure. Assign responsibility for measurement performance, fixtures, PLC stops, missing MES acknowledgements and overdue calibration in a RACI or interface-responsibility matrix.
Turn the RFP into verifiable commitments
A strong RFP is not a feature wish list. Requirements such as “detect scratches,” “store data” or “easy to operate” cannot be accepted objectively. Each requirement should combine the object, conditions, expected result, required evidence and accountable party.
Define the CTQ and measurand in one controlled statement
Whether the station checks dimensions, appearance, assembly, electrical behavior, leakage or load, define the measurand first. “Measure the hole” is incomplete. Specify the drawing feature, datum, orientation, thermal state, filtering and calculation used to produce the result. For visual inspection, define defect classes, locations, boundary examples, and how misses and over-detection are handled.
Keep one authoritative specification. If drawings, quality criteria, customer boundary samples and legacy operator judgments disagree, the buyer must resolve the conflict before delegating a decision to the vendor. Stable automated decisions cannot be built from ambiguous ground truth.
A representative sample set needs more than obvious good and bad parts
Proposal-stage samples should include results close to specification boundaries, acceptable appearance variation and realistic differences in material, supplier, mold, cavity, upstream equipment, lot and surface condition. Artificial defects can support feasibility when real defects are scarce, but label them as artificial and document where they do not represent production failure modes.
Maintain a sample register containing the part ID, ground-truth label and rationale, storage condition, image, creator, approver and use history. Every bidder needs the same sample set and test sequence; otherwise the evaluation measures different test conditions rather than different solutions.
Required RFP chapters
| Chapter | Buyer input | Bidder response and evidence |
|---|---|---|
| Objective and KPI | Priority of escapes, false rejects, downtime, records and labor | Method, constraints and assumptions |
| Product and CTQ | Drawings, boundary samples, materials, variants | Inspection principle, blind spots and non-detectable conditions |
| Capacity and time | Production scenarios, changeover, retest and recovery | Cycle breakdown, buffers and bottlenecks |
| MSA and calibration | Evaluation policy, references and environment | Study plan, traceability and re-evaluation triggers |
| Decision logic | Limits, hold, retest and manual intervention | State transitions, reason codes and permissions |
| Data | Required fields, destination, search and retention policy | Schema, transport, missing-data behavior and audit trail |
| Mechanics and safety | Layout, infeed/outfeed, utilities and safety policy | Risk controls, access and recovery |
| FAT/SAT | Samples, scenarios, evidence and open-item handling | Procedures, logs, reports and correction plan |
| Support | Operating model, local language and replacement constraints | Service model, spares, remote support and end-of-life notice |
| Commercial | Scope, acceptance, payment and change control | Exclusions, options, boundaries and intellectual property |
Evaluate the measurement process, not only nominal specifications
Resolution and repeatability specifications matter, but they do not prove the factory result. Evaluate the complete measurement process: part location, fixture, lighting, temperature, vibration, contamination, surface reflectivity, operator action, recipe, algorithm, master, calibration and data processing.
ISO published ISO 10012:2026 in February 2026. Its public overview describes a framework for designing, operating, maintaining and continually improving measurement processes so results used in production, testing and other activities are fit for purpose and trustworthy. The practical procurement lesson is to manage confidence in results, not merely to buy a higher-performance gauge. Exact contractual application still requires review of the licensed standard and customer-specific requirements.
MSA is a selection test, not a ceremony after delivery
NIST’s Gauge R&R guidance identifies factors such as parts, operators, gauges and configurations, repeatability, reproducibility, stability, bias, resolution, linearity, hysteresis, drift and differences between geometries. Place MSA evidence across concept testing, design review, FAT, SAT and post-ramp confirmation rather than waiting until the final factory test.
Do not reduce the decision to a universal ratio copied from a presentation. The appropriate acceptance rule depends on the characteristic risk, process spread, tolerance, measurement use, representative samples and study design. The quality function should approve the rule before the supplier runs the study, while customer-specific requirements remain authoritative.
Give all bidders the same MSA challenge
Use the same parts, order, environmental scenarios and ground truth to examine:
- Changes after unloading, reloading, repositioning and repeat measurement.
- Effects of operator, shift, fixture, camera, head and recipe.
- Bias or linearity across the intended range, not only near the center.
- Stability under warm-up, environment, lighting degradation, dirt, vibration and continuous operation.
- Traceability of raw data, exclusions, retests and manual changes.
- Separate handling of FAIL, no-read, communication loss and unknown variant.
When a supplier says its standard method is sufficient, request the method, input data, equations, exclusions, software version and result files. Reproducible evidence is more valuable than an unsupported summary score.
Put calibration and metrological traceability into the contract
A calibration certificate for one component does not automatically make the station result traceable. NIST explains metrological traceability as a property of a measurement result established through a documented, unbroken chain of calibrations, each contributing to measurement uncertainty. NIST also emphasizes that calibrating an instrument alone is not sufficient.
The calibration specification should identify:
- Measurand, range, use points and environmental conditions.
- Reference standards, certificates, uncertainty and traceability chain.
- Whether sensors, masters, fixtures, software compensation and the integrated station are calibrated or verified.
- How as-found and as-left states are recorded.
- How an out-of-calibration discovery triggers retrospective product-impact review.
- On-site versus laboratory work, downtime, loan equipment and transport protection.
- Master storage, cleaning, thermal conditioning, damage and replacement controls.
ISO/IEC 17025:2017 addresses competence, impartiality and consistent operation of testing and calibration laboratories. Thailand’s TIS 17025-2561 is identified by TISI as identical to ISO/IEC 17025:2017, and TISI publishes accredited laboratories and their scopes. The RFP should therefore verify whether the required measurand, range and uncertainty are inside the laboratory’s actual accreditation scope. General accreditation and technical fitness for this job are related but separate checks.
Do not reduce decision logic to PASS and FAIL
In automated dimensional inspection, a measured value close to a specification limit creates risk if uncertainty is ignored. ISO 14253-1:2017’s public abstract covers decision rules for verifying conformity or nonconformity, including values near specification limits while taking measurement uncertainty into account. The buyer and customer should agree on the decision rule, risk allocation, retest and hold process; a supplier should not infer them from a standard name alone.
Design the state model before programming
| State | Meaning | Example equipment action | Required record |
|---|---|---|---|
| PASS | Conforming under the agreed decision rule | Release to the next process | Value, recipe, time and equipment ID |
| FAIL | Nonconforming under the agreed decision rule | Reject or segregate | Failed characteristic, value, image and reason |
| HOLD | Boundary or unresolved ground truth | Route to controlled hold | Reason, approver and later disposition |
| NO READ | Measurement unavailable or invalid | Retest or controlled stop | Cause, attempts and equipment state |
| SYSTEM ERROR | Device, communication, clock or storage failure | Fail-safe action | Error, affected population and recovery |
| BYPASS | Approved temporary alternative | Route to alternative inspection | Approver, period, affected lots and release |
This prevents communication failure from being treated as PASS, no-read events from distorting defect rates, and manual reclassification from erasing original data. Align the state table with PLC signals, screens, database fields and reports.
Data integrity is part of 100% inspection automation
The value of 100% inspection automation is not that every part passes a sensor. The factory must explain how the inspected population maps to recorded results. Missing records, duplicate serials, clock drift, overwritten retests or unacknowledged MES messages undermine the evidence of “100% inspection.”
Require product/lot/carrier identity; start and end time; station, fixture and sensor ID; variant and recipe; threshold, algorithm and software version; measured value and unit; decision and quality flags; evidence file location; retest and manual-change history; calibration/master status; and MES/QMS delivery and retry status.
Do not copy a generic retention period or image-compression setting. The buyer should decide based on customer requirements, product life, warranty, audits, storage economics and confidentiality. Ask suppliers to show data-volume assumptions, search, backup, restore testing, time synchronization, access control, deletion and export.
Add OT security before the interface is frozen
Once the station connects PLCs, cameras, industrial PCs, remote support and MES/QMS, quality-data integrity and OT security overlap. NIST SP 800-82 Revision 3 provides guidance for securing OT while considering performance, reliability and safety; its examples include manufacturing, PLCs and physical measurement systems.
Request a network diagram, accounts and roles, default-password removal, patching approach, allowed communications, remote-access controls, logs, backup and recovery, vulnerability notification and end-of-support policy. Any deviation from factory standards needs an approved exception and compensating controls.
Compare inspection equipment manufacturers in three stages
First, apply non-negotiable Must conditions for safety, CTQs, factory interfaces, legal/import constraints, language support, confidentiality and data ownership. Do not let a low price compensate for a failed Must condition.
Second, run a common technical protocol. Use the same parts, ground truth, environment, duration and failure scenarios. A demonstration-room success is not production evidence if it cannot survive factory vibration, oil, dust, ambient light, temperature, loading variation and network conditions.
Third, compare total ownership risk. Break out fixtures, lighting, validation, logistics, training, calibration, masters, software licences, storage, spares, travel, downtime, new variants, cybersecurity and decommissioning. Do not borrow prices from an article; require every bidder to answer the same cost breakdown and label inclusions, exclusions and variable charges.
| Evaluation area | Question | Evidence of a mature response |
|---|---|---|
| Measurement | Is there evidence under changing conditions? | Raw data and failure conditions are disclosed |
| Integration | Are PLC/MES/QMS boundaries clear? | Signals, APIs, retries and exceptions are documented |
| Operation | Can the plant recover, change over and clean safely? | Time, roles, procedures and training are demonstrated |
| Support | Is suitable support available in Thailand? | Contact, escalation and spare-part path are specific |
| Change | Are recipe and software changes controlled? | Versioning, approval, rollback and revalidation exist |
| Commercial | Is non-performance handled clearly? | Acceptance, correction and termination map to deliverables |
For mechanical and workholding responsibility, see our jig design and manufacturing RFP guide. If the project includes line-wide PLC, robot and MES work, align responsibilities with the FA system integration vendor guide.
Turn FAT/SAT from “it ran” into acceptance evidence
FAT and SAT are not the same test at different locations. FAT checks design, assembly, controls, measurement and documentation before shipment. SAT proves operation with production parts, utilities, handling, operators, environment, factory networks and upstream/downstream systems.
Before FAT, freeze the approved specification and requirements traceability matrix; drawings, BOM, electrical and interface documents; software, PLC, recipe, model and parameter versions; calibration evidence, master and sample registers; test cases and expected evidence; and operating, cleaning, recovery, backup and restore procedures.
| Test area | FAT | SAT |
|---|---|---|
| Function | Normal, abnormal, restart and changeover | Production operation and neighboring equipment |
| Measurement | Agreed samples, interim MSA and repeatable decisions | Production parts, site environment and final MSA |
| Capacity | Simulated flow and bottleneck breakdown | Real supply variation and accumulation |
| Data | Fields, clock, search, retry and audit history | Factory network, MES/QMS and recovery |
| Safety | Safety functions, guards and interlocks | Installed layout, work and maintenance access |
| Maintenance | Fault injection, replacement and backup | Thailand support, spares and operator training |
| Documents | Document set and open-item list | As-built set, training records and final report |
Manage conditional acceptance, redesign, retest, workaround and later-phase items separately. Each punch-list item needs product impact, interim control, owner, due condition, retest method, evidence and effect on acceptance or payment. Do not rely on a verbal promise to “fix it later.”
Also separate SAT from full production acceptance. A controlled ramp period that experiences shifts, variants, lots, cleaning, maintenance and restart makes it easier to distinguish test success from sustainable production capability.
A station that cannot be maintained is not production capability
Lighting ages; lenses become dirty; fixtures wear; sensors and PCs are replaced; recipes, models and databases change. Select a system that can continue to explain its results years after commissioning.
The support agreement should state the Thailand contact, language, escalation path, secure remote-access process, spare strategy, replacement and revalidation method, licences and recurring charges, required diagnostic logs, end-of-support notification, change approval, regression testing, source/configuration backup and credential handover. Response or parts-delivery commitments must be based on site, stock, customs and contracted service hours—not on unsupported universal promises.
Common RFP failures and corrections
“Detection rate” without a denominator, defect class, boundary sample, ground truth, retest and no-read rule is not comparable. Request class-level confusion data, the sample register, raw results and false-decision review.
Different sample sets for different suppliers invalidate comparison. The buyer should control a common set, blind order and decision rule. If production samples are insufficient, contract a data-collection phase before locking the solution.
Running MSA for the first time at FAT exposes the project too late. Split evidence across feasibility, design review, FAT and SAT. Saving only PASS/FAIL removes the ability to investigate or re-evaluate later; retain the needed values, evidence, recipe version and decision reason. Comparing only machine price hides integration, calibration, training, spares, downtime and new-variant work. Finally, unrestricted threshold changes destroy the inspection basis; apply role-based access, approval, reason, version, rollback and revalidation.
Pre-award checklist
- CTQs, measurands, authoritative specifications and boundary samples are approved.
- False acceptance, false rejection, no-read and hold risks are separated.
- MSA design, data, approval rule and re-evaluation triggers are agreed.
- Calibration scope, traceability, uncertainty and out-of-calibration action are defined.
- Decision rules near limits and customer-specific requirements are confirmed.
- Workholding, cleanliness, environment, PLC, safety and MES/QMS boundaries are included.
- PASS, FAIL, HOLD, NO READ, SYSTEM ERROR and BYPASS are distinct.
- Network loss, storage failure, clock error, restart and power loss will be tested.
- FAT, SAT, controlled production and final acceptance are separate gates.
- Thailand support, spares, calibration, training, end-of-life and change control are contracted.
- If BOI or another incentive is assumed, current eligibility is reconfirmed with the official authority when applying.
Conclusion: select a partner that can keep proving the result
Inspection automation in a Thailand factory requires more than nominal accuracy and a successful demo. Design the requirements traceability, representative samples, MSA, calibration, boundary decision rules, data integrity, FAT/SAT, change control and local support as one procurement system. Even 100% inspection automation is not quality evidence unless every result maps to the product and missing data, retests and manual changes remain explainable.
Give every candidate the same challenge and evidence format. Compare Must conditions, technical proof and total ownership risk in that order. A mature proposal does not promise that nothing can fail; it states the operating envelope, failure modes and recovery evidence, and keeps the measurement result reproducible and traceable in production.
TOMAS TECH can help from the concept stage with supplier-neutral RFPs, sample plans, PLC/MES boundaries and FAT/SAT evidence for Thailand factories. You are welcome to contact us even before the inspection method or vendor shortlist is final.
FAQ about inspection equipment manufacturer selection
Should we hire an inspection equipment manufacturer or an SIer?
A component manufacturer may be enough for evaluating a measurement head or camera. When the scope includes fixtures, handling, PLC, safety, data and on-site commissioning, a machine builder or SIer needs integration responsibility. The key is not the label but the contractual boundary across measurement, mechanics, controls, data, acceptance and support.
How should an automated dimensional inspection RFP specify accuracy?
Define the measurand, range, part orientation, fixture, temperature, vibration, cleaning, repositioning, calibration, decision rule and verification evidence. Derive the need from product tolerance, process behavior, customer requirements and decision risk; do not insert an unsupported universal ratio.
Does 100% inspection automation eliminate sampling?
Not automatically. The factory still needs measurement-system monitoring, master checks, audits, calibration, data reconciliation and revalidation after change. Whether sampling can be removed depends on customer-specific requirements, regulation, process risk and the independence of the measurement system.
Can the supplier’s standard MSA acceptance rule be used?
It can be an input, but the buyer approves the final rule based on use and risk. Study design, sample representativeness, process variation, range, reproducibility, stability and bias matter. Obtain raw data and calculation conditions, and prioritize customer-specific requirements.
Is an ISO/IEC 17025 calibration certificate sufficient?
Not by itself. Confirm that the laboratory scope covers the required measurand, range, method and uncertainty. The factory must also maintain the chain from the certificate through masters, fixtures, software compensation and actual production results.
What is the difference between FAT and SAT?
FAT verifies the station and documentation before shipment under controlled or simulated conditions. SAT verifies operation at the destination with production parts, utilities, neighboring equipment, users, environment and factory systems. Both need agreed cases, expected results, evidence and retest rules.
Which costs are commonly omitted from inspection-equipment quotations?
Fixtures, lighting, samples, MSA, calibration, logistics, utilities, networks, MES/QMS integration, training, spares, travel, licences, storage, new variants, cybersecurity and future upgrades are common gaps. Ask every bidder to complete the same WBS and cost sheet.
References
- ISO 10012:2026 — Requirements for measurement management systems
- ISO 14253-1:2017 — Decision rules for conformity and nonconformity
- ISO/IEC 17025:2017 — Testing and calibration laboratories
- NIST Engineering Statistics Handbook — Gauge R&R studies
- NIST — Metrological Traceability FAQ and Policy
- NIST SP 800-82 Revision 3 — Guide to OT Security
- TISI — TIS 17025-2561
- TISI — Accredited Calibration Laboratories
- Thailand BOI — H1 2026 investment applications
