Production Management Systems for Small Factories: Excel, Custom Build or Package?

When a small factory evaluates a production management system, the conversation often starts with product names. The decision is more fundamental: which work should be standardized, who owns each dataset, and what evidence will constitute acceptance. Continuing with Excel, building a custom application, buying a package, or implementing in stages can each be rational. A broad replacement without a clear process boundary can burden a small organization more than the problem it was meant to solve.

This guide helps small and medium-sized manufacturers in Thailand compare the four paths and convert the decision into an RFP, proof of concept (PoC), total cost of ownership (TCO), and user acceptance testing (UAT). Prices and benefits are not presented as universal facts. Worked examples clearly state their assumptions. The goal is not “install the fewest features quickly,” but “establish reliable data and ownership in the smallest useful business scope.”

Executive answer: compare four options on the same seven dimensions

Use these seven dimensions instead of counting features:

1. Operational pain: stockouts, WIP congestion, slow promise dates, late reporting, or weak lot traceability.
2. Scope: one line, one product family, one factory, or several sites.
3. Standardization: retain current work or adopt the package’s standard process.
4. Changeability: who can safely change products, routings, forms, or customer rules later.
5. Data governance: where the authoritative item, BOM, routing, inventory, and production records reside.
6. Operating capability: who manages users, support, training, backup, and recovery.
7. TCO: implementation plus five years of subscription, support, changes, downtime, internal labor, and exit.

Excel has low initial friction and is easy to change, but simultaneous use, version control, auditability, and key-person dependency become constraints. A custom system can fit distinctive work, but its code, documentation, security updates, and succession become the factory’s responsibility. A package offers standard functions, support, and a roadmap, but requires real decisions about adopting standard processes. A staged approach limits investment and change, but without a target architecture it can merely multiply spreadsheets and point solutions.

Compare Excel, custom development, packages, and staged implementation

“We are small, so Excel is enough” and “future growth means ERP from day one” are both incomplete. Five users in a food or automotive supplier may need strong audit and lot genealogy. Fifty workers may require only a few named planners if shop-floor events are captured automatically. Define scale through transaction volume, concurrency, traceability granularity, change frequency, and the consequence of downtime—not headcount alone.

When governed Excel remains valid

Keeping Excel is an active operating choice. Store the authoritative workbook in one controlled location. Fix column meaning, units, codes, and date formats; protect formulas; identify approved versions; and assign an owner. Do not let every department maintain its own copy of the item master. Back up the files and test restoration.

Measure the cost of extending Excel when several of these symptoms appear:

• production, inventory, and accounting re-enter the same event;
• daily consolidation delays a promise date until the next day;
• a formula or inserted column stops reporting;
• a critical workbook cannot be maintained when one person is absent;
• lot, WIP, scrap, and rework history cannot be reconstructed;
• an order change does not reach planning and the line at the same time.

Recording labor, correction, premium freight, stockout, and excess-inventory events creates a defensible investment baseline.

When a custom production management system is sustainable

A developer alone is not a product team. Sustainable custom development needs a process owner, product owner, development, testing, infrastructure, security, and user-training responsibilities. It also needs master-data approval. If one developer is essential, the plan must explain holiday coverage, departure, database upgrades, vulnerability remediation, and incident response.

Custom development is strongest when a distinctive production method is genuinely competitive and forcing it into a package would destroy value. Even then, the factory does not need to reinvent authentication, user administration, audit logs, and backup. A hybrid can retain custom production logic while using mature components for commodity capabilities.

When a production management package is the better fit

Evaluate a package through scenarios, not labels. “Inventory management included” is less useful than: “After a customer changes quantity and due date, the system recalculates requirements, shows potential shortages, and sends an approved purchase request.” Use representative items, substitutes, subcontracting, split deliveries, rework, scrap, and lot splits.

Do not turn every Fit & Gap finding into customization. Classify each gap as a legal or customer requirement, a quality requirement, competitive differentiation, or habit. Change habits where practical. For true requirements, consider configuration before add-ons and integrations. Heavy customization increases regression testing for every upgrade.

When small-start implementation is disciplined

A small start is a testable business boundary, not an arbitrary feature cut. One product family’s “order → plan → issue → report → finished receipt” can be an initial scope, with a daily summary to accounting and equipment integration later. Item codes, lots, units, timestamps, and user identifiers should still match the future design.

Our production management system comparison guide covers the broader product-selection criteria. The small-start implementation guide explains phase design. For cost categories, see the production management system cost guide.

Build a minimum business model before writing requirements

Map the current and target flow on one page: sales order, production planning, purchasing, material receipt, issue, production reporting, inspection, finished receipt, and shipment. For every step, record the input, output, responsible role, cutoff, exception, and current form.

Exceptions are decisive. A normal-item demo cannot prove fit. Test situations such as:

• customer quantity or due date changes during execution;
• material substitution, lot split, or return;
• a routing step is skipped, repeated, or outsourced;
• nonconformance becomes rework, scrap, or concession;
• one material lot feeds several finished lots;
• offline events are buffered and synchronized without duplicates.

Divide the model into initial scope, future scope, and exclusions. Explicit exclusions protect delivery dates and accountability.

Write an RFP around scenarios, not a feature catalogue

Business volume and non-functional assumptions

State factory, line, user, and concurrent-session counts; item count; BOM depth; manufacturing orders and event volume; storage locations; and retention period. If unknown, measure logs for one to four weeks and state the estimation method.

“24/7 availability” is not sufficiently precise. Define service hours, whether planned maintenance is excluded, the measurement period, recovery time objective (RTO), recovery point objective (RPO), and support hours. For a small factory, a local buffer or controlled paper fallback during a network outage may create more practical resilience than an expensive high-availability stack.

Master data and traceability

Define create, edit, approve, and retire permissions for items, BOMs, routings, equipment, users, partners, warehouses, units, and reason codes. Effective dates should prevent a later master change from rewriting historical orders.

“Lot traceability available” is not an acceptance condition. Specify forward and backward lookup time, behavior after issue correction and lot split, links to certificates, and the applicable retention period. Derive retention from customer contracts and regulation rather than a generic article.

Interfaces and error handling

For each ERP or accounting field, decide which system is authoritative for items, orders, purchasing, inventory valuation, completion, and shipment. Define API or file protocol, retries, idempotency key, failure notification, closed-period correction, time zone, and encoding.

Direct database coupling can be fast but breaks when an upgrade changes tables. Prefer a contracted API or versioned file interface. Each interface needs a sender success condition, a receiver duplicate key, and an operator procedure for resuming unprocessed records.

Roles, audit, and backup

Apply least privilege to planning changes, event correction, inventory adjustment, master approval, and user administration. In a small team, high-risk actions can use dual approval or monthly exception review. Avoid shared identities and retain who changed what and when.

“Daily backup” is incomplete. Define objects, frequency, location, encryption, retention, owner, restore method, and restore-test frequency. A SaaS provider’s platform backup is not the same as the customer’s export for accidental deletion or contract termination.

A PoC is a decision experiment, not a free demo

A PoC should test the uncertainties that could reverse the buying decision. Examples include whether an operator can report within 20 seconds, existing barcodes can be read, events survive an outage without duplication, or day-end closes within 30 minutes. These thresholds are examples; set values from the factory’s real process.

The PoC plan should include:

• hypothesis and pass/fail threshold;
• representative items and exception scenarios;
• users, devices, network, scanners, and printers;
• test-data creation, anonymization, and disposal;
• period, ownership, and escalation;
• artifacts reusable in production and artifacts to discard;
• return or certified deletion of PoC data.

A perfect vendor dataset proves little. Run at least one exception such as quantity correction, substitute material, rework, or a lost connection.

Define UAT acceptance before the contract is signed

UAT is the evidence for payment and go-live, not a tour after configuration. Agree test cases, expected results, severity, retest, pass criteria, and treatment of unresolved items in the RFP or contract.

“No critical defects” can hide many smaller defects. Set limits by severity, accepted workaround, fix date, and the person authorized to defer. Include a stabilization period and support through the first monthly close.

A 30/60/90-day implementation decision plan

Days 1–30: measure reality and freeze the evaluation scope

Measure Excel and paper transaction volume, re-entry labor, corrections, promise-date lead time, inventory variance, and emergency purchases. Assign owners to items, BOMs, routings, and inventory. Profile duplicates, missing data, and inconsistent units.

Deliver a current-state map, problem priority, minimum business model, data dictionary, option longlist, initial TCO, and first RFP. Management approves the scope and decision criteria. Shop-floor leaders select representative scenarios. IT documents interfaces, security, backup, and continuity.

Days 31–60: use RFP, demonstrations, and PoC to reduce uncertainty

Give every shortlisted provider the same scenarios and data. Score process fit, usability, integration, governance, operations, security, five-year TCO, and delivery capability. Retain evidence instead of relying on presentation quality.

Run a PoC only with leading candidates and focus on entry time, exceptions, offline behavior, labels, and difficult ERP/accounting interfaces. Review data ownership, SLA, price adjustment, and termination assistance in the draft contract. Keep governed Excel as a legitimate outcome if the evidence supports it.

Days 61–90: contract, design UAT, and prepare the first release

Select the option and scope, then create a responsibility matrix. Baseline configuration, migration, training, UAT, cutover, and stabilization. Start data cleansing and schedule migration rehearsals. Approve roles, audit, backup, local buffer, and incident contacts.

The factory does not need every function live on day 90. The appropriate exit is a defensible investment decision, agreed contract, usable data, acceptance criteria, and operating ownership—ready for safe initial deployment.

Compare TCO, including costs outside the quotation

TCO includes setup, subscription, maintenance, customization, infrastructure, interfaces, migration, training, internal labor, downtime risk, and exit. Use a period appropriate to the expected product life and contract; the example below uses five years.

Worked example with explicit assumptions

This is a method example, not a quoted market price or promised result:

• configuration, migration, and training: THB 1,200,000;
• subscription and support: THB 45,000 per month;
• additional changes: THB 300,000 in year 1 and THB 200,000 in year 3;
• internal administration: 40 hours/month at THB 700 loaded cost per hour;
• devices, network, and backup: THB 350,000 over five years;
• termination export and migration preparation: THB 200,000.

The five-year TCO is:

1,200,000 + (45,000 × 60) + 300,000 + 200,000 + (40 × 700 × 60) + 350,000 + 200,000 = THB 6,630,000

The example excludes tax, foreign exchange, price escalation, and outage. Replace every assumption with supplier terms and internal cost. Excel also has consolidation, error correction, audit preparation, backup, and dependency costs. Custom development includes requirements, testing, monitoring, security maintenance, technology upgrades, incident response, and succession.

Benefits need assumptions too. If three people each spend 45 minutes on daily consolidation, 22 workdays/month, loaded labor is THB 500/hour, and 70% of that effort is truly avoidable, the annual labor benefit is:

3 × 0.75 hours × 22 × 12 × THB 500 × 70% = THB 207,900/year

Do not add a generic “20% inventory reduction.” For inventory, service, or schedule benefits, state the baseline, mechanism, value at risk, and accountable owner.

Specify data ownership and the exit before implementation

“The customer owns its data” does not define what can be recovered. List transactions, configuration, masters, attachments, audit logs, calculations, dashboards, and API-created records. An export needs keys, relationships, units, code tables, timestamps, encoding, attachment structure, and a data dictionary—not CSV files alone.

Ask:

• Can the factory export regularly without additional fees?
• Which fields, history, and audit events are available through the API?
• What are the notice period, delivery time, and cost at termination?
• When and how does the provider delete retained copies, and is evidence available?
• Can configurations and custom reports be transferred?
• What assistance is available if the product or provider ceases service?

Test a sample export monthly in another environment. Discovering missing fields only at termination leaves little leverage.

Right-size operations, RBAC, and backup for a small factory

Small organizations need fewer layers, not weaker accountability. The process owner controls rules and priorities; the data owner controls master quality; the administrator controls users and settings; and the provider delivers contracted support. Use one service desk and distinguish incidents, questions, and enhancements.

Build roles from tasks, not titles: plan, approve plan, report, correct event, adjust stock, change master, and administer users. Set deadlines for disabling leavers and review access periodically. Shared terminals should still preserve individual authentication.

Test restoration in cloud and on-premises arrangements. A local buffer maintains work during disconnection but creates sync and device-failure responsibilities. Derive RPO and RTO from tolerable data loss and downtime. After restoration, reconcile item counts, open orders, inventory totals, and audit records.

A practical small-factory architecture

Separate four layers: Shop floor → edge/local buffer → application/database → ERP/accounting. Scanners, terminals, and machines send events through a buffer where continuity requires it. The application manages orders, progress, WIP, lots, and genealogy. ERP/accounting receives the agreed financial or logistical granularity.

Apply RBAC, audit, and backup at each appropriate layer and preserve a regular export. Avoid direct writes into internal ERP tables. Constrain the blast radius so a cloud or interface outage does not stop every physical task. A direct-to-cloud design may also work, but only after checking plant connectivity, latency, cybersecurity, residency, and continuity requirements.

The supplier should explain:

• which tasks continue offline and for how many records;
• how duplicate or out-of-order events are resolved after reconnection;
• when master changes reach each layer;
• who can read audit logs or restore backups;
• when and in what format the complete dataset can be exported.

Treat public support as a financing input, not the reason to buy

Thailand BOI reported continued use of the Smart and Sustainable Industry measure for existing-operation upgrades. In Q1 2026, BOI reported 61 applications with THB 7,071 million of investment, covering areas such as energy saving, renewable energy, machinery upgrades, automation and robotics, and digital efficiency. This is market context, not evidence that a small production-management project automatically qualifies.

In 2026, depa published initiatives and cooperation related to digital adoption by SMEs, standardized digital services, trials, advice, and access to finance. OSMEP’s BDS presents a mechanism connecting SMEs and registered business-development service providers. Application windows, eligible entities, eligible expenses, matching ratios, caps, and pre-approval requirements can change. Confirm the latest conditions directly with BOI, depa, OSMEP/BDS, or the relevant official office before relying on support. The base TCO should remain viable without a subsidy.

Use ISO 22400 as a common language for KPI definitions

ISO 22400-1:2014 provides an industry-neutral framework for defining, composing, exchanging, and using KPIs for manufacturing operations management and was confirmed current in 2025. Its concepts help align a KPI’s purpose, formula, elements, time behavior, unit, and user. The public preview of ISO 22400-2 discusses time, logistical, quality, energy, and other elements in KPI descriptions.

Do not merely display OEE or schedule adherence. Document the formula, planned-time boundary, stop categories, good-unit definition, cutoff, and recalculation. Two parties can use “availability” while calculating different denominators. This article does not claim conformity or certification; formal requirements should be checked against the licensed standard and competent advice.

Frequently asked questions

At what company size is a production management system necessary?

Headcount is not decisive. Consider transactions, concurrent changes, traceability, promise-date speed, audit obligations, and the consequence of error. A small regulated supplier may need stronger control than a larger but simple operation.

What should a small manufacturer implement first?

Choose the smallest closed workflow tied to the priority problem. “Order → plan → issue → report → finished receipt” is often more testable than planning or inventory alone. Align codes with later quality and maintenance phases.

Is a custom production management system cheaper?

It can appear cheaper when only initial screen development is counted. Compare five years of change, testing, monitoring, backup, security, staff turnover, and incidents. Custom is more defensible where unique operations create value and a permanent internal capability exists.

How should production management packages be compared?

Use the same factory data and exception scenarios. Score integration, authorization, recovery, export, implementation team, and five-year TCO in addition to function. Preserve demonstration evidence.

Does a small-start implementation create local optimization?

It can, unless the target architecture, authoritative data, common codes, and phase exit criteria are agreed first. Define both the first scope and the date or condition when temporary workarounds end.

Are RFP, PoC, and UAT all necessary for a small project?

They can be lightweight, but their roles differ: RFP makes offers comparable, PoC tests uncertainty, and UAT accepts contracted results. Short documents are fine; undefined pass/fail conditions are not.

Can a Thai SME use public support?

Possibly, depending on entity, project, timing, and eligible cost. Do not apply an old announcement directly to a current purchase. Confirm the latest conditions through BOI, depa, OSMEP/BDS, or the responsible official channel.

Summary: start small, but design data and exit for the long term

A production management system for a small factory should not be selected by instinct. Compare governed Excel, custom build, package, and staged delivery through one business model, one scenario set, RFP, PoC, five-year TCO, and UAT. The first release can be small, but master data, ownership, RBAC, audit, backup, export, and ERP/accounting principles should be established from the outset.

TOMAS TECH can help at the early stage—testing whether Excel has reached its limit, identifying processes suitable for a package, and choosing the first PoC boundary. We support factories in Thailand with comparison matrices, RFPs, and practical 30/60/90-day plans that respect existing ERP and accounting systems. Contact us.

References

• Thailand BOI, Q1 2026 investment applications and Smart and Sustainable Industry activity: https://www.boi.go.th/index.php?language=en&module=news&page=boi_event_detail&topic_id=138788
• Thailand BOI, H1 2026 investment update: https://www.boi.go.th/un/boi_event_detail?language=de&module=news&topic_id=139075
• depa, digital transformation cooperation for Thai SMEs, 17 July 2026: https://www.depa.or.th/en/article-view/20260717_04
• depa, SME Spring Board / BDS workshop, 30 June 2026: https://www.depa.or.th/th/article-view/20260630_01
• OSMEP SME Scoring: https://osmepscoring.sme.go.th/
• OSMEP BDS information: https://bds.sme.go.th/News/Detail/103
• ISO 22400-1:2014 overview and status: https://www.iso.org/standard/56847.html
• ISO/DIS 22400-2 public preview: https://www.iso.org/obp/ui?_escaped_fragment_=iso%3Astd%3Aiso%3A22400%3A-2%3Adis%3Aed-2%3Av1%3Aen