Automation Failure Risks: Thai Factory Acceptance Gates

Reducing automation failure risks takes more than selecting a fast robot or a reputable machine builder. In Thai factory projects, the costliest problems often accumulate at the boundaries: between the buyer and system integrator, design and production, normal operation and abnormal recovery, or project delivery and maintenance ownership. If nobody owns the decision, acceptance evidence is vague, and jam or power-loss recovery is not designed, a machine can meet its headline speed and still fail as a production investment. This guide turns those gaps into stop/go gates for the RFP, design review, FAT, SAT and production handover.

Executive conclusion: most automation failure is acceptance-boundary debt

“The machine runs” is not a sufficient definition of success. A Thai factory must also be able to recover a misfed part, diagnose a dirty sensor, restart safely after a power loss, control access on night shift, restore PLC and HMI backups, source spares and coordinate upstream and downstream equipment. Peak machine speed does not measure any of those capabilities.

In this article, acceptance-boundary debt means project obligations omitted where one party or lifecycle phase hands work to another. It is a management concept, not an accounting term. A missing owner, unmeasurable acceptance statement, undefined exception, or unclear stop/go authority does not disappear. It returns later as a change order, shipment delay, production stoppage, unsafe workaround or dispute.

Use five rules to keep the debt visible:

1. Name one buyer-side gate owner and an authorized deputy for every gate.
2. Replace adjectives with a test procedure, inputs, expected result, logs and signatures.
3. Test jams, shortages, sensor faults, communications loss, power recovery and emergency-stop recovery—not only the normal automatic cycle.
4. Never close an exception as “later”; record its impact, interim control, owner, due date and retest date.
5. Make stop/go decisions from a pre-agreed evidence pack, not schedule pressure in the meeting.

For the broader investment envelope, see our guide to capital investment planning for Thai factories. For integration responsibility and commercial scope, see how to select a robot system integrator in Thailand. The present guide concentrates on making that investment contractually and operationally acceptable.

Why Thai factories should manage automation risks before signing a contract

The International Federation of Robotics reported that 542,000 industrial robots were installed worldwide in 2024 and Asia represented 74% of new deployments. Those figures show scale; they do not establish an individual factory’s success rate or ROI. As adoption expands, buyers need institutional capability for integration, acceptance, change and operation—not only model comparison.

Thailand’s BOI reported 132 applications worth approximately THB 17.2 billion under its Smart and Sustainable Industry initiative in the first half of 2026, covering machinery upgrades, digital technology, automation and robotics. The same release separately reports machinery, automation and robotics applications worth around THB 13.1 billion across 82 projects. These are application values, not completed investments, installed equipment or achieved benefits. This article also does not imply that a past or current incentive window is open. Confirm the latest eligibility and application terms directly with BOI when making a decision.

Market momentum and project success are different facts. When a company rushes an RFP because competitors are automating, baseline losses, exceptional work and responsibility boundaries are easily deferred. The answer is not indefinite caution. It is a short, disciplined rule for what evidence permits the project to advance.

The four forms of acceptance-boundary debt

Owner debt

Production may own throughput, EHS safety, Quality the product decision, Maintenance recovery, IT the network and Procurement the contract. Cross-functional participation is healthy; multiple final owners are not. If everybody is “accountable,” nobody will stop shipment when evidence is incomplete.

Assign one gate owner to collect evidence, maintain exceptions, convene the meeting and record the decision. That person does not override specialist veto rights. EHS and Quality can retain mandatory approval for their domains while the gate owner keeps the process coherent.

Evidence debt

“Stable operation,” “easy to use” and “adequate capacity” are not testable requirements. A valid criterion identifies the product variants, input condition, operator, measurement window, excluded stops, data source and pass rule. A photograph of the HMI cannot prove cycle time, safety-function validation, quality traceability or recovery time.

Build one evidence record from requirement ID, test ID, preconditions, performer, timestamps, input workpieces, expected result, actual result, logs or video, deviation and signature. Reuse the same requirement IDs at FAT and SAT so the buyer can see exactly which factory-condition differences still require testing.

Exception and recovery debt

Design reviews and sales demonstrations naturally emphasize a clean automatic cycle. Real production loses time after a skewed workpiece, a vacuum fault, a blocked downstream process, a tool change, a power interruption or an emergency stop. Without a designed recovery state, the factory becomes dependent on one expert entering the cell and inferring PLC state.

OSHA’s robotics overview notes that many robot accidents occur during non-routine activities such as programming, maintenance, testing, setup and adjustment, when workers may temporarily be inside a robot’s working envelope. OSHA does not provide a project-failure percentage here; the point is that acceptance limited to normal automatic operation misses important exposure.

Decision debt

Shipping with an incomplete FAT moves supplier debugging into the Thai factory’s installation window. Accepting with an incomplete SAT can turn a design defect into a warranty-scope argument. A stop does not have to cancel the project. It can mean “do not ship until retested,” “run only the approved product family,” or “do not use this mode until the safety item is validated.” Define proportional hold conditions before schedule pressure arrives.

A six-gate factory automation process

Gate 0 — Freeze the business baseline

Do not begin with quotations. Begin by defining the present loss and constraints. Record:

• Product variants, materials and planned additions.
• Shifts, operating days, actual runtime, changeovers, cleaning and breaks.
• The cycle-time definition: instantaneous best, sustained average or good product at discharge.
• Sources for scrap, rework, micro-stops, waiting, absenteeism and overtime.
• Buffer and control responsibilities when upstream or downstream equipment stops.
• Hazards, worker access, maintenance practices and existing protective measures.
• Power, compressed air, floor, temperature, dust, washing and network conditions.

Do not inflate the baseline to make automation look attractive. If a value cannot be verified, mark it unknown and run a short observation. The go condition is an approved problem statement, scope, measurement method and decision owner—not a preferred machine concept.

Gate 1 — Freeze the RFP through responsibility boundaries

An RFP is a common basis for suppliers to price the same boundary. Include requirements, supply scope, exclusions, interfaces, acceptance plan, documents, training, spares, warranty and change control.

Do not write capacity as “60 pieces per minute” alone. State the variants, incoming condition, good-product definition, sustained test duration, treatment of planned and abnormal stops, boundary signals and measurement point.

For every material requirement, add columns for owner, verification location, evidence, exception condition and treatment of nonconformity. Require suppliers to answer compliant, conditionally compliant or noncompliant. “Site adjustment” must specify days, people, prerequisites and excess rates.

The go condition is not that every technical detail is known. It is that competing offers are comparable and the principal requirements have an owner, evidence method and failure treatment.

Gate 2 — Close safety, controls and maintainability in one design review

Approving only the mechanical layout allows guarding, access, sensors, panels and wiring to become late additions. Review mechanics, electrics, controls, safety, quality, IT/OT and maintainability against the same requirement register.

ISO 12100:2010 sets out machinery risk-assessment and risk-reduction principles, including hazard identification, risk estimation/evaluation across relevant lifecycle phases, documentation and verification. ISO says the 2010 edition was reviewed and confirmed in 2022 and remains current, while a replacement draft is under development. Confirm applicable editions and local legal duties at project start.

For industrial robotics, ISO 10218-1:2025 addresses safety requirements for industrial robots, while ISO 10218-2:2025 addresses robot applications and cells, including integration, commissioning, operation, maintenance and decommissioning. A conforming robot product does not automatically make the integrated cell acceptable. The end effector, fixtures, material, layout, access, peripherals and recovery behavior remain system-level responsibilities.

ISO 13849-1:2023 provides a methodology for designing and integrating safety-related parts of control systems, including software. It does not select a safety function or required performance level for a specific application. Those choices must flow from the hazard and risk assessment into a validation plan.

Stop the gate if the hazard list is incomplete, manual and maintenance modes are undefined, safety functions lack validation methods, replaceable parts cannot be accessed, boundary-stop ownership is unresolved, or control-source ownership and backups are absent from the contract.

Gate 3 — Use FAT to prove the system is ready to ship

FAT is not a customer tour. It is the last economical place to find repeatable defects before shipment. Approve the test scripts in advance and record the results. If production materials are unavailable, document the substitute, its limitations and each requirement that must be repeated during SAT.

The FAT evidence pack should include requirement traceability, approved drawings, software versions, parameters, I/O tests, safety-function validation, capacity and quality tests, alarm list, exception-recovery tests, open items and retest results. A test passed only because the supplier’s engineer forced PLC states is not evidence that a normal operator can use the system.

Judge open items by severity, not count. A display typo and an unvalidated safety function must never be treated as equivalent items on a punch list.

Gate 4 — Use SAT to prove the Thai factory conditions

SAT includes actual utilities, floor, ambient conditions, material variation, people, networks, upstream/downstream equipment and shift changes. It does not merely replay FAT; it validates the differences between the supplier environment and the Thai plant.

Before capacity runs, confirm installation completion, calibration, grounding, utilities, guarding, asset registration, backups, access and work permits. Then test safe manual movement, homing, emergency-stop state, power recovery, material removal and lockout procedures.

Localization is more than translating an HMI. Alarms, recovery instructions, warnings, machine labels and training must use common equipment IDs and error codes. Require Thai operators to identify a fault, make the equipment safe and escalate without depending on an interpreter.

Gate 5 — Accept operating capability at production handover

Passing SAT is not enough if only the project team can run the cell. Production and Maintenance must demonstrate daily checks, variant changes, cleaning, jam removal, preventive maintenance, backup restoration, parts replacement and supplier escalation.

The handover package includes as-built drawings, risk assessment, validation records, version-controlled PLC/HMI/robot backups, credential-transfer procedure, parts list, recommended spares, maintenance plan, calibration, warranty, support contacts, training records and open items. The buyer should open, search and restore the files; their presence in a shared folder is not proof of usability.

The go condition is a named local owner, no critical nonconformities, controlled residual items and readiness to measure benefit with the original baseline definition. Link contractual acceptance and payment milestones to that evidence.

Make abnormal recovery the core of FAT and SAT

Exception testing is not random fault injection. Design safe test modes and simulations so state transitions can be verified without creating a new hazard. Cover at least:

• Wrong orientation, double feed, jam, dropped part and material shortage.
• Dirty, broken, out-of-range or chattering sensors.
• Grip failure, low vacuum and tool wear.
• Upstream stop, downstream full and buffer limit.
• Communications loss among PLC, robot, HMI and supervisory system.
• Loss and restoration of main power, control power and air.
• Emergency stop, guard opening and safety-device recovery.
• Manual mode, maintenance mode, changeover and cleaning.
• Inspection-device outage, master mismatch and data-write failure.

For each scenario, record detection, safe state, HMI message, operator action, authorization, residual energy, disposition of the workpiece, data integrity, restart conditions and first-piece confirmation. “Press reset” is not a recovery procedure; the procedure must show how the cause is removed and the safeguarded area is verified clear.

Do not evaluate recovery only with the supplier’s most experienced engineer. Observe a trained Thai technician: what information, tools and spares were needed, and when did escalation become necessary? If recovery is slow, distinguish poor diagnostics, inaccessible components, inadequate training and a fundamentally bad state model.

Small-start automation means a closed evidence loop

A small start is not simply a cheap machine. It is a bounded product family, task, interface and owner in which business benefit, safety, quality, recovery and operating evidence can be closed and reused for the next investment.

Good pilots have measurable current conditions and a clear comparison after implementation. Poor pilots attempt simultaneous multi-plant rollout before the standard is proven, or accelerate one process while leaving upstream and downstream ownership unresolved.

Use the same requirement IDs, change control, FAT/SAT discipline, backup ownership and continuation/removal rules as a production project. Temporary equipment does not justify weaker safety. At the same time, do not build every future feature. Specify expansion principles but accept the current use case on current evidence.

If a collaborative robot is a pilot candidate, the label “collaborative” does not itself determine application safety. Evaluate the tool, workpiece, speed, possible contact, surrounding machinery and task. Our guide to collaborative robot implementation in Thailand explains that planning layer.

Automation ROI: calculate the decision investment, not the machine quotation

Automation business cases often deteriorate because the denominator contains only machine price. Include fixtures, foundations, electrical and air work, rigging, shutdown work, guarding, adjacent-equipment changes, networks, training, spares, ramp-up loss, local tax/customs conditions and maintenance.

Separate:

• Initial fixed cost: design, machinery, tooling, integration, construction, validation and training.
• Uncertainty allowance: real-workpiece variation, brownfield differences, installation conditions and extra testing.
• Ongoing cost: maintenance, licences, communications, calibration, consumables, spares and energy.
• Transition loss: downtime, ramp-up scrap, parallel work and training time.
• Residual exposure: supplier dependency, obsolescence, exchange rate and product-change frequency.

Clearly labelled illustrative ROI calculation

The following assumptions are for explanation only. They are not a TOMAS TECH average, Thai factory benchmark or performance guarantee.

Assume the factory’s own baseline shows THB 180,000 per month in losses attributable to manual handling and micro-stops. If a conservative 40% is addressable, gross benefit is THB 72,000 per month. Added maintenance, power and consumables are THB 22,000 per month, leaving THB 50,000 verified monthly net benefit after SAT and ramp-up.

The installed quotation is THB 2,400,000. Add THB 360,000 for exception work and uncertainty, producing a THB 2,760,000 decision investment.

Simple payback = 2,760,000 ÷ 50,000 = 55.2 months.

Excluding contingency would display 48 months, understating the decision exposure by 7.2 months. The example excludes tax, financing, depreciation, inflation, exchange rates, residual value, revenue growth and safety benefits. Apply company cash-flow rules in a real investment decision.

Most importantly, never reuse the 40% or THB 50,000 as a benchmark. Measure the baseline at Gate 0, verify capability in FAT/SAT and recalculate after stable production. If baseline, addressable share or ongoing cost cannot be substantiated, the correct gate result is “insufficient evidence,” not optimistic payback.

Protect automation investment payback with formal change control

Products, takt, layouts, communications and inspection requirements change after RFP. Do not pretend change can be prohibited. Measure each difference from the approved baseline and assess price, schedule, safety, capacity, documentation and retesting.

A change record should hold the requirement ID, reason, requester, old and new conditions, impact analysis, quotation, schedule, risk assessment, approver, implemented version and retest. Even a supplier’s free change must be recorded if it changes software or drawings.

Link payment milestones to evidence such as design approval, FAT pass, installation completion, SAT pass and handover. This article does not recommend a universal retention percentage; the terms depend on the transaction and applicable law. Clarify the warranty start as well. A warranty starting at shipment can be consumed before production if site readiness is delayed.

Define the Japan–Thailand–integrator operating boundary

When Japan specifies, the Thai factory operates and an SIer builds, local requirements tend to arrive last. Set the boundary at every gate:

Language is part of the design. A common-language approval record can coexist with Thai work instructions, but equipment IDs, alarm codes and part numbers must remain stable. Maintain a controlled terminology list so the local team can identify the current document without relying on a Japanese master alone.

Acceptance clauses to copy into an RFP checklist

Requirements and scope

• Every requirement has an ID and priority; bidders answer compliant, conditional or noncompliant.
• Product set, incoming condition, good-product definition and test window are explicit.
• Supply, buyer-furnished items, exclusions, brownfield modifications and boundary signals are diagrammed.
• Thai factory utilities, environment, access and shutdown dates are confirmed.

Safety and quality

• Applicable laws, standards, company requirements and editions are listed.
• Risk assessment, review, residual risk and validation evidence are supplier deliverables.
• Manual, maintenance, cleaning, jam removal and restart safety functions are defined.
• Defect disposition, quarantine, master management and inspection-outage behavior are specified.

FAT and SAT

• Test-script approval, witnesses and required workpieces have dates and owners.
• FAT substitutes map directly to SAT repeat tests.
• Evidence format, logs, video permission, signatures and retest process are controlled.
• Severity-based stop/go rules connect punch items to shipment and acceptance.

Software and data

• Source, version, licence and backup ownership are clear for PLC, HMI, robot and vision.
• Credentials, service accounts, remote access, logs and contract-end revocation are defined.
• Local behavior and resynchronization during supervisory-system loss are tested.
• As-built drawings, code, parts list and instructions are updated after changes.

Handover and service

• Operator, changeover, maintenance, troubleshooting and safety training include competence checks.
• Spares, lead times, obsolescence, special tools and calibration are listed.
• Warranty start, support window, local/remote escalation and charges are explicit.
• The owner, period and method for post-production benefit measurement are assigned.

A 30-minute evidence pack for stop/go meetings

Do not reteach the design during the gate meeting. Circulate a one-page decision sheet with evidence links:

1. Gate and decision requested.
2. Counts of passed, conditional, failed and untested mandatory requirements.
3. Critical safety, quality, schedule and cost exposure.
4. Changes since the previous gate.
5. Interim control, owner, due date and retest for each deviation.
6. Recommended Go, Conditional Go or Stop and named approvers.

Never mix “untested” with “passed.” Use Conditional Go only where the condition is specific, authority and due date exist, and advancing does not increase the risk. Schedule pressure is not a valid condition for incomplete critical safety validation.

The evidence pack is not paperwork for its own sake. Requirement-to-drawing-to-test-to-deviation-to-handover traceability finds disagreements early and creates reusable assets for the next line.

Summary: do not let undefined work cross the next boundary

Automation risks cannot be reduced to zero, but they do not have to remain hidden until commissioning. Define owners, measurable evidence, exception recovery and stop/go authority before the RFP. Measure the baseline at Gate 0, freeze boundaries in the RFP, close safety and maintenance in design, prove shipment readiness at FAT, prove Thai-site fitness at SAT and accept operating capability at handover. Evaluate ROI against the decision investment, including uncertainty and ongoing cost, rather than the machine quotation alone.

TOMAS TECH can support RFP structuring, boundary review of existing quotations, FAT/SAT planning and production handover at Thai factories before a specific machine has been selected. If you want to convert a concept into owner/evidence/gate requirements, contact us for an initial discussion.

FAQ: preventing automation failure risks

What should be checked first when an automation project may fail?

Check the baseline-loss method, project scope, acceptance owner and exceptional work before equipment capability. Without those four items, quotations do not share benefit or responsibility assumptions. A one-page Gate 0 baseline is a practical first deliverable.

What is a safe factory automation process?

Use stop/go gates for baseline, RFP, design review, FAT, SAT and production handover. Safety is not a final inspection item: connect hazards, manual and maintenance work, safety functions and validation evidence from design onward. Confirm local law and applicable standards for the actual installation.

What should be small in small-start automation?

Bound the product family, task, interfaces and ownership—not only the machine price. Close a complete evidence loop for benefit, safety, recovery and operation, and retain requirement IDs, test records and changes for reuse.

How should automation ROI be calculated?

From factory-measured baseline losses, estimate only the demonstrably addressable share, subtract ongoing maintenance, energy and consumable costs, and compare it with a decision investment that includes integration, construction, training, ramp-up and uncertainty. Recalculate with the same formula after SAT and stable production.

Does a longer payback mean the project should be cancelled?

Not automatically. Safety, supply continuity, quality, labour availability and capacity may matter alongside financial return. But do not fill evidence gaps with optimistic benefits. If the baseline or ongoing cost is unknown, measurement should be the next gate condition.

What is the difference between FAT and SAT?

FAT tests what can be reproduced at the supplier before shipment. SAT tests actual Thai factory utilities, materials, operators, adjacent equipment and networks. Every FAT substitute should map to a specific SAT repeat test.

Can equipment run with open acceptance items?

It depends on severity and interim controls. Unvalidated safety, legal or critical-quality requirements must not be accepted for schedule alone. Where limited operation is justified, document the permitted scope, monitoring, owner, deadline, retest and stop authority.

Primary sources

• ISO 12100:2010 — https://www.iso.org/standard/51528.html
• ISO 10218-1:2025 — https://www.iso.org/standard/73933.html
• ISO 10218-2:2025 — https://www.iso.org/standard/73934.html
• ISO 13849-1:2023 — https://www.iso.org/standard/73481.html
• OSHA Robotics Overview — https://www.osha.gov/robotics
• IFR World Robotics 2025 release — https://ifr.org/ifr-press-releases/global-robot-demand-in-factories-doubles-over-10-years
• Thailand BOI 1H 2026 press release — https://www.boi.go.th/index.php?_module=news&from_page=press_releases2&language=en&page=press_releases_detail&topic_id=139075