When a factory or distribution center evaluates automated sortation, the first comparison often focuses on sorter top speed or robot count. Yet the investment is not decided by a catalog maximum. What matters is whether the system can meet an acceptance target after accounting for package variability, barcode read quality, recirculation, destination count, peak-flow shape, recovery from stops, maintainability, and the division of responsibility between the WMS and WCS.
This guide is for shippers, manufacturers and 3PL operators planning a sortation system in Thailand or Southeast Asia. It compares low-speed put-to-light and divert-arm concepts, medium-speed shoe and multibelt sorters, high-speed crossbelt sorters, and robotic picking on the same criteria. It then connects selection to the RFP, capacity validation, PoC, FAT/SAT and ROI. Published product values are useful indicators of technical potential, but they must not be treated as site guarantees.
The 2026 investment context for automated sortation
Thailand’s Board of Investment reported 132 applications worth approximately THB 17.2 billion under the Smart and Sustainable Industry measure in the first half of 2026. This does not prove the business case of any individual project, but it shows continued interest in investments that combine automation, digitalization and sustainability. Sortation should therefore be assessed as part of a chain that includes WMS, conveying, inspection, packing, shipping and maintenance—not as a standalone machine.
Before selecting equipment, management should agree on five questions:
- By what time must the peak backlog reach zero, rather than simply how many operators can be removed?
- Can the system absorb the maximum 15- or 30-minute inflow, rather than only the hourly average?
- How quickly will no-reads, damaged parcels and full destinations be recovered?
- What daily plan-attainment level is required after including stops and changeovers?
- Who diagnoses and restores failures across WMS, WCS, PLCs and scanners?
For the wider investment boundary, see our guide to automated warehouse costs in Thailand. The assumptions behind transport capacity are explained in conveyor system design for Thailand.
Do not begin sortation system selection with a product name
An envelope, carton, tote, polybag and long item need different handling surfaces. A system serving 12 destinations is not selected in the same way as one serving 800 chutes. A two-hour campaign peak is different from a flow that remains high throughout the day. If the mechanism is selected first, the RFP becomes a specification for one product and prevents fair comparison of alternatives.
Eight criteria to freeze before technology selection
- Load profile: minimum and maximum size, weight, center of gravity, friction, flexibility, bottom condition, protrusions, leakage risk and label location.
- Misreads and recirculation: where no-reads, duplicate reads, duplicate IDs, unknown destinations and full chutes are routed.
- Destination count: permanent destinations, wave-dependent allocation, parallel chutes and future expansion.
- Peak flow: 15-minute and hourly flow, burst pattern, SKU mix and package mix, not only a daily average.
- Availability: planned operating time, failure frequency, mean time to repair, degraded operation and bypass options.
- Maintenance: local spares, service response, preventive-maintenance windows, safe access, training and replacement procedures.
- System boundary: whether WMS chooses the business destination, WCS reallocates a physical chute, or PLC rejects the load.
- Acceptance: pass/fail equations for representative and worst-case loads in FAT and SAT.
With these criteria, vendors must respond with assumptions and exclusions instead of a single “items per hour” number.
Comparing four sortation approaches by load, flow and operation
The following table gives a practical positioning. Final selection still requires validation of the conveyor layout, induction spacing, chute design, information flow and operator movement.
| Approach | Suitable loads and work | Strength | Main constraint | Ask in the RFP |
|---|---|---|---|---|
| Put-to-light / divert arm | Piece handling after case opening, fewer destinations, human-assisted quality decisions | Simple concept and phased deployment | Operator pace, item orientation, unstable bags, destination growth | Sustained rate by operator, error prevention, fatigue, full-container handling, changeover |
| Shoe / multibelt | Cartons, totes and bags sorted continuously to a moderate number of destinations | Relatively gentle handling and good balance between rate and destinations | Load-surface compatibility, minimum gap, merges, maintenance points | Divert success by load class, noise, belt replacement, chute blockage |
| Crossbelt | Many destinations, high peaks and mixed small items carried individually | Supports large destination counts and high nominal throughput | Investment, synchronized induction, footprint, controls and maintenance skills | Induction success, recirculation, carrier failure, redundancy with real loads |
| Robotic picking | Extracting products from mixed bins and placing into destination containers | Automates upstream picking through vision and gripping | Unknown products, transparent or reflective packs, tangles and grip failure | SKU coverage, success definition, retries, exception handling and safety |
Low-speed: put-to-light and arm-based diverting
Low-speed does not mean inferior. Where destination count is modest, packages are manageable and people still make quality judgments, it can deliver value with the lowest complexity. Put-to-light is particularly useful when product variation makes full automation uneconomic: the system indicates a destination and quantity, while the operator confirms the action.
Do not base capacity on a skilled operator’s best minute. Measure a complete shift including breaks, replenishment, container changes, full destinations, label issues and newly trained operators. For mechanical arms and pushers, increasing conveyor speed also increases sensitivity to orientation, friction and center of gravity. FAT samples must include light bags, flexible-bottom cartons and off-center loads—not only stable boxes.
Medium-speed: shoe sorter and multibelt
A shoe sorter moves sliding shoes across the conveyor surface to divert a load. A multibelt sorter uses individually controlled belt sections for lateral transfer. Both can gently distribute cartons, totes and some bags across multiple destinations.
The governing variable is often not mainline speed but the gap that induction can consistently create. Effective capacity falls when upstream conveyors deliver clusters, item length varies, or loads must be reoriented for scanning. A full chute also prevents discharge even if the sorter itself is healthy. The WCS must know whether to assign an alternate chute, recirculate, or request operator clearance.
Daifuku lists 13,500 pieces per hour at a 450 mm load length and 165 m/min for one Jet Surfing Sorter example. These are manufacturer-published specifications useful for mechanism comparison, not guaranteed results for a user’s package mix, induction performance, read rate and chute occupancy. Require each vendor to provide a capacity calculation and test conditions translated to your data.
High-speed: crossbelt sorter
A crossbelt sorter uses a small belt on each carrier to discharge a parcel laterally at its assigned chute. Interroll publishes a range of 4,000–20,000 items/hour, a maximum load of 50 kg, a maximum speed of 2.7 m/s and less than 68 dB(A) for the MX-H. These are product specifications, not a promise that every site will achieve 20,000 items/hour.
Site throughput depends on empty-carrier percentage, induction success, no-reads, recirculation, long items occupying multiple carriers, full chutes, startup loss and downtime. For illustration, a nominal 20,000 items/hour multiplied by 95% induction success, 98% first-read success, a 97% factor after recirculation and 95% availability becomes approximately 17,200 items/hour before further allowance for load mix. This is an explanatory calculation, not a contractual formula; the project must agree its own guarantee equation.
On 1 May 2025, Dematic announced its next-generation Silky sorting technology for APAC, including Southeast Asia. More regional options are beneficial, but a new platform also makes local service coverage, spare-part lead time, installed references and software release management important RFP questions.
Robotic picking
Robots can extend automation upstream by recognizing items in a mixed container, gripping them and placing them into destination containers. Vendor comparisons for piece-picking systems are conditional on the SKU set, lighting, vision, gripper, retry logic and exclusions. A demonstration on ten favorable products cannot be expanded into a warranty for tens of thousands of SKUs.
Build PoC classes by difficulty, not just SKU count: clear, black, reflective, flexible, cylindrical, thin, perforated, tangled, overlapped and deformed packages. Measure first-attempt grip success, average retries, cycle time per item, exception rate, wrong-item rate, damage and regression after a model update—not only eventual pick success.
Turning catalog capacity into a site guarantee
Define five levels of capacity
| Capacity level | Meaning | Example |
|---|---|---|
| Mechanical theoretical | Derived from pitch and speed | Carrier count × laps per hour |
| Induction | Loads that can be placed at the correct spacing | Automatic plus manual induction |
| First-pass sort | Loads correctly read and diverted on first pass | Excludes no-read and recirculation |
| Sustained | Rate maintained for a defined period | Average of a two-hour continuous run |
| Operational completion | Ability to clear all work before the dispatch deadline | Zero residual work after wave close |
Separate a one-hour peak from a four-hour sustained rate. Accumulation can make a short test look strong, but the rate cannot continue if packing or loading is slower and destinations become full. Ask for an end-to-end bottleneck table.
Example capacity calculation
Assume a required operational peak of 12,000 items/hour, 98.5% first-read rate, 97% induction success, 95% availability including unplanned stops, and a 5% allowance for recirculation and exceptions. A rough nominal requirement is:
12,000 ÷ 0.985 ÷ 0.97 ÷ 0.95 × 1.05 ≒ 13,900 items/hour
The factors are not necessarily independent. A greater share of polybags may reduce induction and read success simultaneously. Final design should therefore validate historical package-mix scenarios, not rely only on multiplication of average factors.
Provide a peak waveform, not an average
Give vendors 8–12 weeks of anonymized data showing 15-minute arrivals, day-of-week variation, month-end peaks, campaigns, dispatch cutoffs, SKU classes and package mix. If the data is unavailable, perform one or two weeks of direct measurement first. Modeling actual arrival concentration and wave-release logic is safer than assuming a generic peak factor such as 1.5 times average.
Barcode quality, misreads and recirculation are design inputs
GS1’s guidelines cover topics including barcode verification, Scan4Transport and EPCIS. In automated sortation, better scanners alone are not enough. Label generation, print quality, placement, packaging reflection and the timing of data registration may all require improvement.
Classify scan results into four states
- Good read: one identifier is read and matches the expected master data.
- No-read: the code cannot be detected or does not meet the quality threshold.
- Multiple read: several codes are visible and the system cannot select the correct ID.
- Valid but unknown: format is valid, but no corresponding work exists in WMS.
If only no-reads are counted, a wrong code that was successfully scanned will be missed. Define a discharge point, rescan, manual correction and audit trail for each state. After manual correction, event IDs and state transitions must prevent duplicate shipment.
Recirculation is not a free buffer
A loop can absorb a temporary no-read or full chute, but unresolved loads consume carriers and may circulate indefinitely. WCS rules should include maximum laps, maximum dwell time, priority exception discharge, full-chute release and operator notification. Track reason and lap count, as well as the percentage ultimately sent to manual handling.
Write the WMS/WCS/PLC responsibility split into the RFP
The longest disputes in sortation projects often concern interfaces rather than mechanics. Specify normal flow and the behavior for timeouts, retries, out-of-order events, duplicates, network loss and WMS outage.
| Function | Typical WMS responsibility | Typical WCS responsibility | Typical PLC/equipment responsibility |
|---|---|---|---|
| Release | Order, wave and business-destination policy | Convert work for equipment execution | Execute commands |
| Destination | Choose customer/logistics destination | Map dynamically to active chutes | Perform commanded divert |
| Tracking | Retain business status | Track zone, conveyor and carrier | Control sensors and actuators |
| Exceptions | Hold, cancel and re-ship decisions | Control no-read, full chute and recirculation | Safe stop and physical jam detection |
| Actuals | Confirm shipment and update inventory | Aggregate sort events | Send time-sequenced sensor results |
| Recovery | Approve business restart order | Reconcile state, resend and reroute | Restart after safety confirmation |
The interface specification should include message ID, load ID, timestamp, state, destination, error code, retry count and correlation ID. Rather than asking vaguely for “exactly once,” test idempotency: a duplicate message must never produce a duplicate business confirmation.
For cybersecurity, define IT/OT segmentation, allowed traffic, remote-access authentication, audit logging, backup, patch ownership and end-of-support migration. If vendors connect remotely, specify the access window, approver, session record and emergency disconnect.
Safety must be tested together with performance
US OSHA 1926.555 addresses conveyor warning signals, emergency stops, prevention of unintended restart, guarding and lockout during maintenance. A Thailand project must confirm applicable local law, machinery-safety standards and plant rules with qualified specialists, but these topics are useful inputs to the RFP.
Performance testing must never require an open guard, bypassed sensor or automatic restart after emergency stop. Recovery from a safety trip is part of real availability. SAT should cover:
- Audible and visible pre-start warning coverage
- Emergency-stop reach, stopping zone and prevention of unintended restart after reset
- Guards for nip points, crushing, falling items and chute access
- Lockout/tagout for jam clearance, cleaning and belt replacement
- Safe recovery from fire, power failure, network loss and low air pressure
- Thai-language signs, training, competency checks and near-miss reporting
What an RFP for automated sortation should require
Business and data conditions
Provide an anonymized load master, sample photographs, size and weight distributions, barcode types, destination count, 15-minute flow, operating calendar and exception reasons. Every bidder should receive the same data set.
Required response format
| RFP item | Required vendor response |
|---|---|
| Guaranteed capacity | Load mix, measurement boundary, duration, excluded time and confidence basis |
| Sort accuracy | Denominator and treatment of mis-sort, no-read and recirculation |
| Availability | Formula, planned stop treatment, upstream/downstream stops, MTBF and MTTR |
| Noise/environment | Measurement position, loads present, temperature, humidity, dust and floor condition |
| Maintenance | Preventive hours, parts list, local stock, response time and required skills |
| IT/OT | Architecture, responsibility matrix, API, logs, time sync and backup |
| Safety | Risk assessment, applicable standards, verification records and training |
| Expansion | Unit prices to add destinations, throughput, lines and software licenses |
MHI makes a 2026 Annual Industry Report available. This article does not infer or reproduce statistics we could not verify; it only identifies the report as a relevant industry resource. Use confirmed market data and your own operational history in the RFP.
Use gated measurement, design, PoC, FAT/SAT and acceptance
Gate 1: measurement and data quality
Measure flow, load profile, reads, task time, exceptions and stop reasons before purchasing equipment. Clock drift between WMS and PLC can reverse the apparent order of causes. Define time synchronization, timezone handling and retention.
Gate 2: concept design and simulation
For every shortlisted mechanism, model the flow waveform, buffers, induction, chutes, operators and failure modes. Evaluate recovery time after the largest wave, not only average capacity. Overlay columns, egress routes, floor loading, fire requirements and future expansion on the layout.
Gate 3: proof of concept
A PoC is an experiment to reduce unknown risk, not a demonstration that “it moves.” Bring real difficult loads for robotic picking, flexible packaging, reflective labels and dense induction. Sign success, failure and retest rules before the test.
Gate 4: factory acceptance test
FAT integrates mechanics, controls, WCS, exceptions and safety before shipment. If the full length cannot be assembled, combine an emulator with representative equipment sections to validate interfaces, state transitions and load.
Scenarios should include normal peak, sustained run, no-read, multiple read, unknown destination, full chute, failed sensor, failed motor, communications loss, WMS delay, duplicate message, power recovery and emergency stop. Retain raw logs, video, sample lot, software version, open items and corrective-action deadlines.
Gate 5: site acceptance and operational acceptance
SAT uses the actual building, power, network, operators, WMS and packages. Passing mechanical SAT does not complete business acceptance if dispatch deadlines are missed. Pass capacity, accuracy, availability, safety, maintenance, training and documentation separately. Link conditional acceptance to deadlines and retained payment.
Make availability and maintainability measurable
Sortation KPIs beyond OEE
| KPI | Recommended definition | Caution |
|---|---|---|
| Good sort rate | Loads correctly discharged on first pass / inducted loads | Do not mix in success after recirculation |
| No-read rate | Unread loads / loads presented for scanning | Separate label and presentation causes |
| Recirculation rate | Loads entering a second lap / inducted loads | Analyze by reason and number of laps |
| Mis-sort rate | Loads physically discharged to the wrong destination / inducted loads | Separate wrong WMS instruction from equipment error |
| Availability | Available time / planned operating time | Report upstream and downstream ownership separately |
| MTTR | Average time from failure to safe restart | Show waiting for parts or approval |
| Completion lag | Time from wave cutoff until residual work reaches zero | Directly measures deadline attainment |
Fix the contractual availability formula. If “outside equipment responsibility” is excluded without limit, reported availability can look excellent while operations fail. Use common stop-reason codes for equipment, WCS, WMS, upstream, downstream, utilities and operations, and report both whole-system and responsibility-specific values.
Maintenance in Thailand
For Thailand maintenance, distinguish a global 24-hour help desk from technicians who can reach the site. Classify critical parts by production impact, lead time, substitutability and shelf life. Obtain bills of materials and replacement procedures for PLCs, drives, scanners, belts, carriers, wheels, sensors, PCs, UPSs and network devices.
Training should be role-based for operators, supervisors, maintenance, IT and administrators. Include refresher training after one month, night-shift rollout, materials for new employees and fault-recovery drills.
Evaluate ROI beyond labor reduction
Cost overview
Capital cost includes the sorter, induction, merges, conveying, chutes, platforms, guarding, power, networks, WCS, WMS changes, building work, fire compliance, testing, training, spares, freight and tax. Operating cost includes power, wear parts, support, software, cloud services, parts, calibration, cleaning, training and manual fallback during outages.
Benefits fall into four groups:
- Direct labor: sorting, walking, rechecking, supervision and temporary peak labor.
- Quality: wrong shipment, redelivery, returns, customer service and inventory variance.
- Capacity: peak outsourcing, overtime, missed carrier cutoffs, sales constraints and avoided expansion.
- Safety and continuity: heavy handling, incidents, recruitment difficulty and absence sensitivity.
Sensitivity analysis
Build downside, base and upside cases by varying volume, wages, availability, ramp-up, maintenance cost and exchange rates. The first year should include a learning curve. Do not count uncontracted future orders as certain benefits. Link payments and warranty remedies to design approval, PoC where required, FAT, SAT and stable operation.
Common failures and how to prevent them
Shortlisting by top speed
Speed only has meaning with load profile, induction, chute occupancy and recirculation. Require every vendor to simulate the same data and state a site-specific guarantee.
Treating barcode defects as a scanner problem
Root causes include label generation, printing, application and packaging. Use GS1 guidance and create upstream-quality KPIs.
Postponing the WMS/WCS boundary
If interfaces are finalized after mechanics, exceptions will fail at FAT. Approve state transitions and the responsibility matrix during basic design.
Assuming exceptions will be unmanned
No-reads, damage, full chutes and cancellation will occur. Design a safe exception lane, screen workflow, authorization, audit log and sufficient manual capacity.
Staffing maintenance after startup
Involve local maintenance in FAT and require fault-injection and recovery. Acceptance should prove that the team can restore service, not merely that a manual exists.
FAQ: sortation system selection
What should be compared first when selecting a sortation system?
Compare load profile, destination count, 15-minute peak, barcode quality, recirculation, availability, maintenance and WMS/WCS boundaries before comparing products. Equal hourly ratings are not equal if test loads and exclusions differ.
At what volume should a crossbelt sorter be introduced?
There is no universal threshold. Crossbelt is strong for many destinations and high peaks, but induction or chute capacity can dominate. Compare it with shoe, multibelt and distributed robotics using your waveform, expansion plan, service capability and ROI.
How should conveyor transport capacity be specified?
Use 15-minute peak inflow, occupied length, minimum gap, merges, scanning and restart behavior—not daily average. A faster mainline adds no business capacity if upstream induction or downstream packing remains the bottleneck.
Can a logistics automation case study be used directly for ROI?
It can reveal possible benefit categories, but cannot be copied. Labor rates, volume, package mix, destinations, building, shifts and service targets differ. Recalculate using your own operation.
What is the difference between FAT and SAT?
FAT verifies integrated mechanics, controls and software before shipment. SAT verifies the installed system with the actual building, WMS, workforce and load mix. Both should cover capacity, exceptions, safety and recovery.
Can a manufacturer specification become a contractual guarantee?
Only if its assumptions match the site. Normally, the parties create a separate guarantee defining load mix, duration, read rate, recirculation and stop exclusions, and use it as the FAT/SAT pass equation.
Conclusion: design automated sortation backward from acceptance
The success of automated sortation is determined not by top speed but by whether the operation clears real package waves before the business deadline. Put-to-light and arms, shoe and multibelt, crossbelt, and robotic picking each have an appropriate application. Write load profile, misreads and recirculation, destination count, peak flow, availability, maintenance and the WMS/WCS boundary into the RFP. Then reduce risk gate by gate through measurement, design, PoC, FAT and SAT. Use catalog values to shortlist; contract against a site-specific guarantee and acceptance test.
TOMAS TECH can help compare sortation concepts, prepare an RFP, validate throughput and plan FAT/SAT in Thailand, even before the equipment specification is fixed. Share your current flow and constraints through our contact page.
References
- Thailand Board of Investment, Smart and Sustainable Industry investment applications: https://www.boi.go.th/index.php?_module=news&from_page=press_releases2&language=en&page=press_releases_detail&topic_id=139075
- Interroll, MX-H Horizontal Crossbelt Sorter: https://www.interroll.com/products/sorters/mx-h-horizontal-crossbelt-sorter
- Daifuku, Sorting and Picking Solutions: https://www.daifuku.com/us/solution/list/sortingpicking/
- Dematic, next-generation sorting technology for Southeast Asia, 1 May 2025: https://www.dematic.com/en-au/newsroom/press-releases/2025/dematic-launches-next-generation-sorting-technology-to-enhance-logistics-efficiency-across-southeast-asia/
- OSHA, 1926.555 Conveyors: https://www.osha.gov/laws-regs/regulations/standardnumber/1926/1926.555
- GS1 Guidelines: https://ref.gs1.org/guidelines/
- MHI: https://www.mhi.org/
- Dematic, Piece Pick Systems: https://www.dematic.com/en-us/solutions/suite-of-intralogistics-systems/piece-pick-systems/
