Master Data You Must Organize Before “Improving Your Factory with AI”: Items, Processes, Equipment, and Personnel

Target audience: Factory managers, production control staff, IT promotion leads, and site directors at Japanese manufacturers with operations in Thailand and ASEAN. This article is written for those who face challenges such as “We are considering AI adoption but don’t know where to start” or “We are told to pursue DX, but our operational data is scattered and we can’t make use of it.”

The call to “improve factories with AI” is spreading rapidly among Japanese manufacturers based in Thailand. Vendor proposals, headquarters directives, industry seminar demonstrations — every one of them features the buzzwords “AI, IoT, DX.” Yet when we visit actual factory floors, most companies run into the same wall: the data that AI needs to consume is simply not in order.

Item master records are managed separately by department. Equipment codes follow different naming systems in the old and current platforms. Process names vary depending on which line leader you ask. Worker skills and qualifications are maintained in individual Excel files. In this environment, no matter how powerful an AI engine you deploy, what it produces is flawed recommendations based on garbage data. “Garbage In, Garbage Out” — this principle has not changed one bit in the age of AI.

This article explains, from a factory-floor perspective, how to organize the master data that Japanese manufacturers in Thailand must have in place before they can “improve factories with AI” — specifically item, process, equipment, and personnel master data — and why it matters so much. We also cover the recommended sequence for building out master data, how to frame the investment case for headquarters, and the areas where TOMAS TECH can provide support.

Why “Master Data” Is Being Questioned Now

Thai manufacturing in 2026 has shifted from a pure-growth story to one of selectivity and focus. The World Bank has issued cautious growth outlooks for Thailand, with external uncertainty, rising logistics costs, and increasing energy costs all squeezing manufacturer margins. At the same time, BOI (the Board of Investment of Thailand) has made it clear that it will actively support investment in automation, AI, data analytics, and enterprise management IT.

When considering AI and IoT investment in this environment, the most frequently overlooked element is the data foundation. Attaching sensors, building dashboards, running AI models — all of these presuppose that data is correctly defined. The underlying foundation for that prerequisite is master data.

Thailand-based operations face unique challenges. In some cases, systems from Japan headquarters cannot be transplanted as-is. In others, information asymmetry exists between Thai staff and Japanese expatriates. And in many facilities, years of operation have turned multiple Excel files into the de facto system of record. When AI adoption is attempted in such conditions, the result is typically “the system is running, but the numbers don’t match what’s happening on the floor.”

What Is Master Data: The Four Pillars

Manufacturing master data can be broadly divided into four categories. Each is interrelated and serves as the basis for judgment by AI and automation systems.

① Item Master

This assigns a unique code and attribute information to every product, component, raw material, and consumable. It includes part numbers, item names, units of measure, supplier codes, storage locations, lot management requirements, and inventory classification (A/B/C). A common problem at Thailand-based operations is having the same part registered under three separate entries — a Japanese item name, a Thai abbreviation, and an English model number — leading to double-counted inventory and missed purchase orders.

② Process Master

This standardizes and registers the process steps required to manufacture a product. It includes process codes, process names, standard cycle times, assigned lines and equipment, and quality checkpoints. Without a properly organized process master, it is impossible to compare actual man-hours against standards and identify which processes are generating losses. For AI to analyze “which processes have the highest defect rates,” the process definitions must first be unified.

③ Equipment Master

This is the ledger of all machines and equipment that make up the production line. It includes equipment codes, equipment names, installation locations, commissioning dates, maintenance intervals, assigned maintenance personnel, and specifications (rated output, processing capacity). When IoT sensors are attached to equipment to collect data, the equipment master is the key that links sensor signals to “which equipment” and “which condition value.” Without a proper equipment master, sensor data is nothing more than a string of numbers.

④ Personnel/Skill Master

This digitizes the skills, qualifications, and assignment information of operators, team leaders, and technicians. It includes which processes each person is qualified to handle, certifications held (forklift license, welding skills, etc.), work shifts, and multi-skill proficiency levels. A well-maintained personnel master enables use cases such as “automatically identifying available personnel when a specific piece of equipment has a fault” and “tracking progress of multi-skill development.” Given that personnel turnover is relatively high in Thailand, keeping skill data current is especially important.

What Happens When Master Data Falls Into Disorder: Common Failure Patterns on the Floor

What kinds of problems arise when digitization and AI adoption proceed without properly organized master data? Here are the patterns most commonly observed on manufacturing floors in Thailand.

Pattern ①: Inventory counts don’t reconcile

Because item master records are maintained across multiple parallel systems, the inventory count in the system never matches the physical count. Investigation reveals that the same part was registered under multiple part numbers. The accounting team spends enormous time on monthly inventory adjustments.

Pattern ②: Utilization figures aren’t trusted

IoT sensors were installed to collect equipment operation data, but the “utilization rate shown by sensors” diverges greatly from the “gut feel of floor leaders.” Investigation reveals that the “rated operating hours” setting in the equipment master did not reflect reality, causing the OEE calculation formula to function incorrectly. The dashboard is running, but no one trusts the numbers.

Pattern ③: AI recommendations are “unusable”

A demand forecasting AI was deployed, but the purchasing recommendations it outputs diverge significantly from floor intuition. Root-cause investigation reveals that historical order data contained item code inconsistencies (the same product registered under multiple codes), causing the AI to learn them as separate products. The problem was not model accuracy — it was data quality.

Pattern ④: Digitization of forms stalls

An attempt is made to digitize daily reports, inspection sheets, and quality records, but the “item names,” “process names,” and “equipment names” that need to populate form selection lists vary by department, making it impossible to create a common form. Each department ends up maintaining its own form, and data cannot be aggregated or analyzed across departments.

In What Order Should Master Data Be Organized

Attempting to organize all four master datasets simultaneously makes the project too large and causes it to stall. The practical approach is to prioritize based on “the area where the numbers are most out of alignment” and “the area most directly tied to investment recovery.” The following is a general recommended sequence.

Phase	Target Master	Reason for Priority	Expected Benefits
Phase 1	Item Master	Foundation for inventory, purchasing, and stocktaking. Serves as the reference source for “item codes” in all other masters	Reduction of inventory discrepancies, visibility into excess stock, reduction of stocktaking man-hours
Phase 2	Equipment Master	Prerequisite for IoT and operation management. The key that converts sensor data into meaningful information	Accurate OEE calculation, foundation for predictive maintenance, standardization of maintenance planning
Phase 3	Process Master	Link to processes after item and equipment masters are solidified. Defines standard man-hours and quality checkpoints	Actual vs. standard comparative analysis, identification of defect-generating processes, common platform for form digitization
Phase 4	Personnel Master	Only by linking to the process master does it become clear “who can handle which process”	Tracking progress of multi-skill development, shift optimization, early identification of attrition risk

That said, the sequence should be adjusted based on your specific pain points. If equipment failures are occurring frequently and urgent reporting to headquarters is required, it is reasonable to prioritize Phase 2 (Equipment Master).

Practical Item Master Organization: Common Stumbling Blocks and Solutions in Thai Factories

Item master organization is not a one-time task — it is living data that requires continuous maintenance. Here are the stumbling blocks most commonly encountered in Thai factories and how to address them.

Stumbling block ①: Too much “garbage” in historical data
Systems used for years accumulate discontinued items, duplicate part numbers, and unused entries. The first step is to inventory all items and conduct a cleansing exercise to confirm “is this still in use?” and “are there duplicates?” This is painstaking work, but it cannot be skipped given its major impact on all downstream processes.

Stumbling block ②: Three languages — Thai, Japanese, and English — coexist
If you don’t decide upfront which language will serve as the official item name, searchability suffers. The recommended approach is to use “an English or alphanumeric code as the unique key, with Japanese and Thai display names stored as supplementary information.” If codes are unique, cross-language matching becomes straightforward.

Stumbling block ③: Floor staff take time to adapt to new part numbers
After master data is reorganized, floor staff may continue using old names and old part numbers. An effective transitional measure is to build an “old part number → new part number” conversion table into the system, allowing searches by both numbers during the migration period.

Stumbling block ④: No designated maintenance owner
Without clearly defined “owners” for the item master (personnel with authority to approve additions, changes, and discontinuations), unauthorized entries accumulate indefinitely. Documenting master management procedures (the approval workflow for item additions, changes, and discontinuations) and enabling Thai staff to operate them as part of their daily work is the key to long-term sustainability.

Equipment Master and IoT: Turning Sensor Data into Meaningful Information

Attaching IoT sensors to equipment has become achievable at relatively low cost. However, the data those sensors continuously stream only acquires its operational meaning — “utilization rate,” “downtime,” “defect occurrence timing” — when properly linked to an accurate equipment master.

The minimum fields an equipment master should contain are as follows.

Equipment code: A system-wide unique ID (matched to IoT data tag names)
Equipment name and model number: Recommended to maintain in three languages — Thai, Japanese, and English
Installation location: Defined in a hierarchical structure of factory / building / line / station
Rated operating hours: Must be set accurately as the denominator for OEE (accounting for shift hours and planned downtime)
Maintenance interval and most recent maintenance date: Baseline for preventive maintenance planning
Assigned maintenance personnel: Linked to the personnel master

OEE (Overall Equipment Effectiveness) is calculated as “availability × performance rate × quality rate,” but the numerators and denominators of each component vary significantly depending on how the equipment master is defined. Organizing the equipment master accurately transforms OEE into a trusted management indicator that can serve as the basis for improvement activities.

Process Master and Form Digitization: Steps to Move Beyond “Paper Daily Reports”

At many Thai manufacturing sites, daily reports, work instructions, quality records, and equipment inspection sheets are still paper-based. When attempting to digitize these, there is a problem that must be solved before deciding “what form to build” — and that is unifying the process master.

Digital form templates require selection lists for fields such as “item name,” “process name,” “equipment name,” and “operator name.” Master data is the source of these lists. Without organized master data, the selection options vary by department, making it impossible to analyze collected data across organizational boundaries.

The three key points for process master organization are as follows.

Systematizing process codes: Build in a hierarchical structure such as “line number + process number”
Recording standard cycle times: The baseline for making productivity losses visible through comparison with actuals
Defining quality checkpoints: Embedding into the master “which inspection items are confirmed at which process” turns digital forms into “tools that prevent verification omissions”

Once the process master is in place, introducing a form digitization tool (e.g., i-Reporter) allows data from every process and every line to be collected in a unified format, making daily, weekly, and monthly aggregation and comparative analysis straightforward. The outcome is not just “paper replaced by digital” — it is “data that can actually be analyzed.”

Personnel Master and Multi-Skill Development: Preparing for the Mobility of the Thai Workforce

At Thai manufacturing sites, operator turnover rates are often higher than Japan headquarters anticipates. When a skilled worker who has mastered a specific process leaves, the quality and productivity of that process temporarily deteriorate sharply — this “key-person dependency risk” is a challenge shared across Japanese manufacturers in Thailand.

The personnel master is one mechanism for managing this risk. By digitizing “which processes each worker can handle and at what proficiency level,” the following operational capabilities become possible.

When a specific veteran worker is absent, substitute personnel can be identified immediately
Progress of multi-skill development can be visualized in a process × personnel matrix, making it easier to build training plans
“Single-operator processes” with high attrition risk can be identified early and backup training can be prioritized

Maintaining the personnel master requires updates every time there is a hire, assignment change, or departure, so it is important to create a system in which the HR and manufacturing departments collaborate on maintenance. Ensuring that Thai management staff are proficient in the system and can update it as part of their daily work is the key to sustainability.

Master Data and Headquarters Proposals: Building a “3-Year Payback” Argument

Master data organization, on its own, can make “the numbers easier to see,” but it is difficult to demonstrate a direct case for revenue growth or cost reduction, which makes it hard to secure investment approval from headquarters. In reality, however, master data organization greatly determines the ROI of subsequent investments (IoT, AI, form digitization).

The recommended approach for headquarters proposals is not to apply for “master data organization” as a standalone project, but to present it as an integrated project — such as “master data organization + inventory management system deployment” or “master data organization + equipment operation management + IoT” — and show the overall ROI.

Investment Item	Qualitative Cost Reduction Case for Headquarters	Logic for 3-Year Payback
Item master organization + inventory management system	Reduction of excess inventory, reduction of stocktaking man-hours, avoidance of production stoppages due to stockouts	Current inventory value × improvement in excess inventory rate + reduction of stocktaking man-hours × hourly labor cost
Equipment master organization + operation management system + IoT	Reduction of unplanned stoppages, optimization of maintenance costs, improvement of OEE	Reduction in opportunity loss: current unplanned downtime hours × production loss unit cost
Process master organization + form digitization	Reduction of daily report creation man-hours, rapid aggregation of quality records, traceability assurance	Daily report creation and transcription man-hours × number of personnel × hourly labor cost reduction
Personnel master organization + skill management	Reduction of key-person dependency risk, acceleration of multi-skill development, improved responsiveness to sudden absences	Reduction of production stoppage costs from attrition/absences and hiring/retraining costs

When leveraging BOI (Board of Investment of Thailand) preferential treatment for automation and IT investment, it is necessary to clearly state at the application stage “which system will be used for which process improvement.” Designing an integrated project that includes master data organization strengthens the BOI application and also makes it easier to obtain headquarters approval. For details, please refer to the BOI official website (https://www.boi.go.th/) or contact the relevant BOI office.

Failure Patterns in Master Data Projects and How to Avoid Them

Master data projects are among the most likely to stall midway. Here are the common failure patterns and how to avoid them.

Failure Pattern ①: Scope is too large

Plans to “organize all master data across the entire company in one go” stall in most cases — because floor-level cooperation dries up, assigned personnel run out of capacity, or priorities shift midway.
Countermeasure: Start with a pilot scoped to one warehouse, one line, or one process. Measure results before rolling out more broadly.

Failure Pattern ②: No master owner designated

Without a defined maintenance owner after the initial organization, master data quality deteriorates over time. Within a year, the data is “back to needing reorganization.”
Countermeasure: Document an “owner (responsible department and person)” and a “change approval workflow” for each master type. Design from the outset a structure that Thai staff can manage autonomously.

Failure Pattern ③: Progressing through IT only, bypassing the floor

When IT or administrative departments take the lead and design master data without incorporating floor-level input, floor staff find the system difficult to use and begin neglecting data entry. The result is divergence between “data in the system” and “what’s actually happening on the floor.”
Countermeasure: Involve floor leaders (team leaders, line leaders) from the master design stage. Design fields that fit naturally into daily work routines.

Failure Pattern ④: AI adoption races ahead before master setup is complete

AI models are introduced before master data organization is complete, producing problems of “poor accuracy” and “unusable recommendations.” The actual cause is data quality, but accountability becomes unclear between the AI vendor and the internal system team.
Countermeasure: Explicitly write “completion of master data cleansing” and “confirmation of historical data quality” as preconditions for AI adoption in contracts and project plans.

Phased Deployment: Starting Small and Embedding Change on the Factory Floor

Companies that have successfully achieved digitization and DX at their Thai operations share a common trait: they prioritize “accumulating small wins” over launching “large-scale projects.”

The recommended approach for the phased introduction of master data organization and subsequent systems is as follows.

Step 1 (Months 1–3): Current-state assessment and priority area identification
Conduct an inventory of current data management practices and clearly identify “which master is most disorganized” and “which area’s numbers are least trusted.” Measuring KPI baselines (inventory discrepancy rate, OEE calculation accuracy, etc.) at this stage is essential for measuring results later.

Step 2 (Months 3–6): Master data organization in the pilot area
Focus cleansing and organization on the highest-priority area (e.g., item master for key components, equipment master for key equipment). Finalize the master management procedures and responsible personnel at this stage.

Step 3 (Months 6–12): System deployment and results measurement in the pilot area
Use the organized master data as the foundation to deploy inventory management systems, operation management systems, and form digitization tools in the pilot area. Measure KPI changes (improvement in inventory discrepancy rate, improved accuracy of utilization visibility, reduction of daily report creation man-hours, etc.) and calculate actual ROI figures.

Step 4 (Month 12 onward): Horizontal rollout and next-phase master organization
Use the knowledge and actual ROI from the pilot to secure additional investment approval from headquarters and proceed with rollout to other lines and departments. Simultaneously begin work on next-phase master data (process master, personnel master).

The Path to AI Utilization: The Possibilities That Organized Master Data Opens Up

Once master data is accurately organized, AI and machine learning utilization for the first time functions as a realistic investment. In an environment with well-organized master data, the following AI applications move actual factory numbers.

Demand forecasting and order optimization: By combining accurate item master data with historical inbound/outbound records, purchase quantity recommendations that optimize the balance between excess inventory and stockouts function with high accuracy.
Predictive maintenance: With the equipment master and accurate sensor data properly linked, a model that “detects deviations from normal vibration patterns and recommends maintenance timing” operates at a practical level.
Defect root-cause analysis: With the process master and quality record data integrated, AI can identify “at which process, under which conditions, and with which operators does the defect rate increase.”
Shift optimization: By combining skill data from the personnel master with required skill profiles for each process, automatic generation of optimal shift plans aligned with production schedules becomes possible.

These are AI applications that can be realized “if master data is organized.” Conversely, deploying AI without organized master data will not generate these values. It is most accurate to view master data organization as a “prerequisite for AI investment.”

TOMAS TECH Perspective

TOMAS TECH provides Japanese manufacturers in Thailand and ASEAN with the inventory management system PEGASUS, the paperless application i-Reporter, an equipment operation management system, and a smartwatch system. Here is a brief overview of how each solution addresses the challenges described in this article, from the perspective of master data organization and subsequent system deployment.

PEGASUS as a data foundation for inventory management:
The inventory management system PEGASUS incorporates a mechanism for centrally managing the item master. By routing inbound, outbound, and stocktaking data through PEGASUS, it becomes the foundation for resolving the challenge of “system inventory counts not matching physical counts.” Deploying PEGASUS alongside item master cleansing makes it possible to achieve both data accuracy and operational continuity.

i-Reporter for form digitization and process data accumulation:
The paperless application i-Reporter, by designing form templates linked to the process master, enables collection of daily reports, inspection sheets, and quality records in a unified format across all processes and all lines. Beyond reducing paper transcription man-hours, the data accumulated serves as the foundation for process-level quality and productivity analysis. The UI design that allows Thai staff to operate it intuitively on the factory floor has also been recognized for improving adoption rates.

Equipment operation management system and OEE visualization:
By deploying the operation management system on top of an accurately organized equipment master, OEE (availability, performance rate, quality rate) can be tracked in real time for each piece of equipment. Recording the occurrence and cause of unplanned stoppages enables use of the data as the basis for maintenance planning.

Smartwatch system for improved floor responsiveness:
By receiving alerts for equipment abnormalities and process anomalies via smartwatch, floor leaders can notice and respond to problems immediately. Combined with skill data from the personnel master, it is also possible to implement a system that delivers alerts directly to “personnel who are qualified to respond.”

TOMAS TECH has adopted as its standard approach starting small — from “one warehouse, one line, one process” — embedding the change on the floor, and then rolling out more broadly. We can support the entire process from master data organization through system deployment, results measurement, and headquarters reporting. Please contact us at https://tomastc.com/contact.

Summary

The intent to “improve factories with AI” is sound and the investment direction is appropriate. However, AI requires accurate data to function, and accurate data cannot be generated without organized master data. Only when item, process, equipment, and personnel master data are in order does IoT sensor data become “utilization rates,” form digitization data become “the basis for quality analysis,” and demand data become “material for order optimization.”

For Japanese manufacturers in Thailand to navigate the environmental changes of 2026 and beyond — slowing growth, rising costs, escalating quality expectations — what is needed is not DX as a buzzword but DX that actually moves the numbers on the factory floor. Master data organization is the first and most important step toward that goal.

There is no need to launch a large-scale project all at once. Start with one warehouse, one line, or one process: organize the master data, deploy the system, and measure results. Accumulating these small wins is the most reliable path to making DX a reality at your Thailand operations.