When selecting a shuttle AS/RS, a claim of high speed or high density does not tell a factory how many orders it can actually ship. A shared lift, merging conveyor, picking station or WMS interface may limit the entire system before the shuttles reach their rated capacity. This guide is for buyers handling totes and cases in Thai factories and distribution sites. It covers shuttle versus stacker-crane selection, peak throughput, usable storage density, common failure points, maintainability and acceptance through RFP, FAT and SAT. The aim is to make competing bids comparable and turn the promised service into a verifiable contract.
Define the storage service before selecting a shuttle
A typical tote shuttle AS/RS combines shuttles moving horizontally inside racks, a lift between levels, infeed and outfeed conveyors, racking and control software. Dematic lists shuttle, rack, conveyor, lift and software as the core components of its Multishuttle. Swisslog describes its CycloneCarrier as a small-load store, a feeder for goods-to-person stations and a consolidation buffer. These are vendor examples, not proof that every shuttle product has the same capabilities.
The buyer should specify more than the number of slots. Which orders must reach a picking port before the shipping cut-off? How many component totes must reach a production line before its local stock runs out? What output must remain when one lift is unavailable? A system sized to daily averages can queue during a short peak. An unsupported extreme forecast, on the other hand, can buy shuttles and lifts that remain idle.
This article concentrates on totes and cases. Pallet shuttle systems have different load-handling devices, floor loads, fire and building requirements, and throughput units. Do not transfer a tote calculation directly to pallets.
How this guide differs from the stacker-crane selection article
Our stacker-crane AS/RS selection guide covers the crane-centred technology choice and building constraints. Here we examine what happens after a shuttle option survives the initial shortlist: how many shuttles and lifts are needed, how a real peak is served, and what remains available after a failure. Give each proposed technology the same SKU, load, order, peak and availability data before comparing simulation results.
The scope boundary matters. A tote shuttle may still need a separate solution for raw-material pallets. Retrieving a tote from the rack creates no customer value if the outfeed conveyor is full. Delivering it to a picking port may not raise shipments if staffing and replenishment do not match the demand wave. An RFP limited to the shuttle vehicle leaves the performance owner unclear.
Shuttle versus stacker-crane AS/RS: comparison questions
The table is a set of questions, not a universal ranking. Obtain project-specific figures from each supplier and verify them in testing.
| Decision factor | Shuttle proposal: ask for | Stacker-crane proposal: ask for |
|---|---|---|
| Load unit | Tote, case or pallet suitability; size and weight range | Mini-load or unit-load configuration and load handler |
| Motion | Parallel horizontal moves and shared lift waiting time | Travel, hoist and fork cycle time |
| Peak flow | Limits by level, aisle, lift, conveyor and port | Limits by aisle, crane and input/output point |
| Density | Usable slots after depth, SKU rules and maintenance access | Usable slots after aisle, height and depth constraints |
| Failure | Separate effect of one shuttle and one lift failure | Effect of one crane failure on an aisle |
| Expansion | Whether lifts and ports can expand with shuttles | Space and building provision for another aisle or crane |
| Maintenance | Shuttle retrieval and safe access inside racks | Crane parking, aisle closure and recovery route |
Parallel shuttles can create useful capacity, but a common lift or conveyor can cap both capacity and availability. A crane often serves a defined aisle, yet dual cycles, load handlers and aisle layouts change its effective performance. Neither “shuttles are always faster” nor “cranes are always cheaper” is a sound procurement rule. Compare the response to the actual demand curve.
Design throughput from order lines, not a machine’s headline cycle
Extract daily and hourly receipts, dispatches, order lines, SKU velocity, urgent requests, replenishment, returns and empty-tote movements from the WMS. Include ordinary weeks and the busy season. Preserve differences by shift, weekday and shipping cut-off. One order line is not necessarily one tote movement: a tote may support several orders, while one order may require several totes. State batching, wave release and sequencing assumptions explicitly.
A first-pass requirement is required moves during a peak window divided by effective operating time in that window. That is demand, not demonstrated system capacity. The system is limited by its slowest relevant component and can lose more capacity through contention and queuing. For illustration only, suppose peak demand is 900 totes/hour, shuttles can move 1,200, lifts 960, conveyors 1,050 and picking ports 800. The ports cap the system at no more than 800, so the design cannot meet demand. These are hypothetical numbers, not manufacturer ratings or actual project results.
Do not apply one average availability percentage to every hour. A planned weekend stop and a breakdown just before cut-off have different business effects. Test normal operation, planned maintenance and one-component failure separately. Examine the demand in 15-minute intervals: an average of 900 against nominal capacity of 1,000 says little if 300 totes arrive in one quarter-hour.
Test lifts, merges and ports independently
Even when a shuttle operates on every level, a shared lift is a common resource. Infeed and outfeed may share it; dual decks and alternative routes may change performance. Swisslog shows a double-deck lift option for high-performance applications, but that feature alone does not prove a particular site will meet its peak. Require simulation outputs for lift utilisation, queue length and longest dwell time using the same order data.
The inbound scanner, load-shape and weight checks, conveyor merge and outbound buffer can also constrain flow. Define where a rejected tote goes and how it is reintroduced, or normal totes may queue behind it. At picking ports, use measured operator speed, staffing, container changeover and empty-container return. A fast machine demonstration is not the same as sustained order throughput.
Evaluate the marginal value of another shuttle
Ask what one additional shuttle contributes to end-to-end dispatch throughput. If a lift is saturated, another shuttle may wait without raising output. If lift capacity remains and one level has concentrated demand, level-specific vehicle allocation may help. Compare assignments by level, travel distance, charging, maintenance and failure recovery rather than total vehicle count alone.
For example, replay a 15-minute period when outbound demand concentrates on the upper levels separately from a 15-minute period when replenishment is spread across all levels. Even with the same total number of moves, the first case may queue at a particular level’s shuttles and the lift entrance, while the second may queue at a conveyor merge. If vehicles are fixed to levels, a vehicle on another level cannot necessarily retrieve stock from the failed vehicle’s level at once. Even if the design allows vehicles to transfer between levels, confirm the transfer equipment and time required. Merely including “one spare shuttle” in a quotation does not show how many minutes later it can work on which level. The buyer should include a demonstration of the recovery procedure for a vehicle failure on a level holding critical SKUs; this reveals what the claimed redundancy actually delivers.
Calculate usable density by subtracting operating constraints
The number of rack openings in a drawing is not the number of slots a business can use. Allow for SKU and lot separation, expiry rules, working space for replenishment, quarantine, cycle counts, empty containers and unavailable areas during maintenance. Multi-deep storage can improve nominal density but may require relocation to reach a rear tote and changes sequencing. Swisslog describes double- to quadruple-deep tote storage for CycloneCarrier; SSI SCHÄFER describes a Cuby arrangement with level shuttles and a maintenance platform. “Shuttle” covers different designs.
Ask for nominal slot count and usable capacity under the buyer’s actual SKU and lot profile as separate figures. Test representative loads for length, width, height, maximum mass, deformation, protrusion, base deflection and uneven weight distribution. For cold storage, make condensation, frost, lubrication, batteries and maintenance exposure explicit. Catalogue compatibility must be confirmed against site conditions in the supplier response.
Overlay columns, clear height, floor loading, sprinklers, fire compartments, escape routes, assembly access and future expansion on the layout. Maximising rack height while eliminating safe maintenance access may increase later downtime. Evaluate retrievable inventory and recoverable design, not volumetric efficiency alone.
Break common failure points into vehicles, lifts, controls and power
“Many shuttles” is not a complete redundancy plan. One vehicle may fail while others work, yet all may depend on the same lift, conveyor, PLC, WCS, network or power supply. For each component record lost output, affected SKUs, recovery time and an alternative procedure. Stock distributed across levels may still be unavailable when the only outfeed lift fails.
At minimum, request responses to one-shuttle failure, loss of communication on a level, one-lift failure, one outfeed conveyor failure, one-port failure, WCS outage, WMS disconnection, power restoration and fire-alarm interaction. For each, state safe-stop time, visible inventory, manual-release rules, duplicate-move prevention and stock reconciliation after restart. One MTBF or uptime number cannot explain the event that stops a production line.
Define permissible interruption from the operation. A warehouse that can recover an order the next morning has a different need from a factory whose line stops after 30 minutes without parts. Decide whether to invest in another lift, bypass conveyor, line-side stock or manual withdrawal using the loss caused by interruption and the time to recover. Distinguish equipment redundancy from a service the business can actually tolerate.
Build safety and maintenance access into the first layout
If a shuttle stops inside a rack, who enters, from where, after isolating which equipment, and how is it retrieved? Show access on the layout together with rescue, falls, crushing, stored energy and unintended start of other levels. SSI SCHÄFER lists a maintenance platform as part of one Cuby configuration. Maintenance space is a design input, not an afterthought.
ISO 12100:2010 sets out principles for identifying hazards, estimating and evaluating risk, reducing it and documenting verification in machinery design. It is not a certificate for a product or a declaration of compliance with Thai law. The manufacturer, installer and factory safety team must determine applicable standards and local requirements, then assess access doors, interlocks, emergency-stop zones, lockout, fire response and escape routes for this project. Do not automatically apply a driverless-truck standard to a rack shuttle merely because both move without a driver.
In the service agreement specify the first responder in Thailand, support language, night and holiday coverage, spare-part location, diagnostic-log access, backup ownership and recovery drills. A spare shuttle does little if the failed unit cannot be extracted safely or recommissioned. Check lead times for proprietary tools, batteries, drive wheels and sensors before signing.
Assign WMS, WCS and PLC responsibilities by event
A common architecture uses the WMS for inventory, lots, allocation and priorities, the WCS for movement sequencing, and PLCs for machine control and safety interfaces. Actual product boundaries vary. Define which system is the stock record of truth, when an inbound move is complete, and who handles a misread, shortage, disconnection, retry and timeout. Agreeing an API format without event meaning can leave persistent stock differences.
Interface tests should cover duplicate admission of the same tote ID, unreadable barcode, order cancellation, priority change, communication loss during retrieval, retry after recovery, partial dispatch, count discrepancy and fire or emergency-stop interactions. Logs should connect order ID, tote ID, location, command ID, timestamp and state transition. Agree retention and access rights. Include language, time zone, time synchronisation and user permissions so local IT staff can maintain the installation.
Where a new machine joins an unchanged ERP or WMS, each boundary creates an exception path. If the WCS confirms a putaway before the WMS records it and a retrieval order arrives meanwhile, which state wins? Keep test and production master data separate. If mechanical and software work are under separate contracts, name the incident coordinator and final owner of interface defects.
Put Thai implementation conditions into the RFP
Thailand’s Board of Investment describes an industrial-upgrade measure under Smart and Sustainable Industry. Its current public page states conditions for eligible Group B activities, including a THB 1 million minimum efficiency-improvement investment, excluding land and working capital. Buying an AS/RS does not automatically qualify a project. Confirm eligible activity, timing, expenditure and procedure with BOI. If an investment case includes a potential incentive, show both approval and no-approval scenarios; the base cash plan should work without an unapproved benefit.
Budget for import and customs, local installation, power quality, floor reinforcement, building and fire-protection changes, network, Thai-language training and a local maintenance response. Comparing machine prices alone hides differences in construction, controls, software integration and training scope. Our automated warehouse price and ROI guide explains cost categories; map the performance and acceptance conditions in this article to those cost lines.
For a live-warehouse retrofit, plan incoming equipment routes, temporary storage, stock migration, weekend shutdowns and a temporary shipping process. In a new build, coordinate columns, fire design, egress and expansion space with the building concept. Put the site’s allowable shutdown dates into the RFP before suppliers schedule installation.
RFP input data and required supplier response
Roughly a year of handling history is useful; when less exists, disclose the observation period and seasonal uncertainty. Give every bidder the same input and require supplier assumptions in a separate field. SKU identities can be anonymised, but the distributions of dimensions, weights and order lines must remain usable.
| RFP input | Buyer supplies | Supplier returns |
|---|---|---|
| Loads | Tote/case dimensions, mass, base and deformation | Accepted range, reject logic and load trial |
| Demand | 15-minute and hourly moves, order lines, SKU velocity, growth cases | Normal and peak output, waiting-time distribution |
| Stock | SKUs, lots, expiry, stock targets, empty containers | Nominal and usable slots, allocation assumptions |
| Building | Columns, floor, clear height, fire, power and access | Layout, structural conditions, excluded work |
| Interruptions | Allowable downtime, critical SKUs, manual alternative | Remaining capacity and recovery for each failure |
| IT | ERP/WMS events, IDs and network | WCS/PLC roles, interface, logs and access |
| Contract | Start date, training language, support hours | FAT/SAT, warranty, spares and change prices |
Require a common table showing rack levels and depth, shuttles by level, lifts, ports, component capacities, whole-system output, output under failure, maintenance access, energy assumptions and installation plan. One “maximum per hour” figure is inadequate. Show peak 15-minute, continuous-hour and whole-shift results with mean, 95th-percentile and worst delay. A 95th percentile is a location in the delay distribution; it is not the absolute maximum guarantee.
Mark rack foundations, sprinklers, fire compartments, electrical mains, network, WMS changes, test stock, training, Thai and Japanese documents and initial spares as included, excluded or buyer-supplied. Link payment milestones to design approval, FAT, installation, SAT and stable running, not just delivery of equipment.
Compare alternatives under identical data and constraints
A shuttle bid and a crane bid cannot be compared from advertised cycle rates. Give both the same SKU profile, stock, order history, picking method, ports, building geometry and operating hours. Test current demand, a justified future base case and a severe peak, documenting the origin of each forecast. These are comparison scenarios, not universal planning horizons.
Deliverables from simulation should include input files, assumptions, batching and release rules, utilisation by component, queues, order-completion distribution, usable slots and failure results. The buyer needs to be able to rerun or challenge the assumptions. A demonstration using another customer’s SKU distribution cannot prove this site’s performance. Contractual access to inputs and model assumptions makes changes easier to manage.
Score capacity, density, interruption tolerance, maintainability, IT fit, construction time and total cost with weights agreed before bids arrive. A factory exposed to production-line stoppage may weight failure-mode capacity heavily; a leased warehouse with scarce floor area may weight usable density. Changing weights after viewing prices biases the outcome.
Use FAT and SAT to prove sustained performance with real loads
Factory acceptance testing checks controls, safety, interfaces, representative loads and exceptions at the maker’s facility or a test environment. Site acceptance testing checks the installed machines, electrical and network connections, fire interaction, real loads and orders, maintenance and training in the Thai building. A start-up demonstration alone proves neither peak output nor recovery.
For each test, record purpose, input data, assumptions, pass criterion, measurement method, responsible person, evidence and retest rule. If 900 totes/hour is the promised rate, specify duration, infeed/outfeed mix, tote weights, SKU mix, port staffing, exception rate and work-in-process at the start and finish. Preloading the rack and conveyor then counting a short dispatch burst does not establish sustained capacity.
Induce safe, planned failures of one shuttle, one lift, network, emergency stop, power and barcode reading. Verify safe stop, the operating area that remains, accurate stock locations and no duplicate dispatch after recovery in the logs. Run hazardous tests under an approved safety plan; agree alternative evidence and residual risk where a live test is unsafe. Passing a performance run does not replace safety assessment, and passing safety checks does not prove output.
During the stabilisation period measure exceptions, downtime, manual interventions, stock differences and late orders each day. Keep incomplete items on a punch list with owners, due dates and temporary procedures. Final handover should cover layout, mechanical and electrical drawings, contracted PLC/WCS source or usage rights, backups, spare lists, maintenance instructions, training records, test evidence and software licence terms.
Compare TCO and interruption exposure over the same period
Vehicle price alone misses racks, lifts, conveyors, controls, building works, fire protection, installation, software, power, maintenance and spares. Align the comparison period, operating days, growth, replacement parts, service contracts, labour, stoppage loss and residual value. Software support, version upgrades, WMS changes and spare-shuttle battery replacement may matter over the contract life. Keep unapproved tax benefits in a separate scenario.
Benefits can include less picking travel, fewer line shortages, better stock accuracy, fewer mis-shipments, better space use and resilience to seasonal demand. Use measured baselines and consistent definitions. Do not count the same freed labour time both as a wage saving and as a separate productivity gain. Where a baseline is missing, treat the benefit as a measurement task rather than a bankable return.
First reject options that fail the required service. Then compare total cost and operating risk among the qualifying designs. A cheap bid that misses the final 15-minute shipping wave is unsuitable. A more expensive bid with meaningful recovery routes may merit another look when interruption losses are included.
A practical sequence from data to stable operation
First, measure orders, movements, load form, stock, exceptions, interruption cost and building constraints, noting gaps and seasonality. Second, define the service: peak rate, cut-off, capacity after one failure, usable slots and safety access. Third, issue a common RFP and collect simulation results and itemised prices.
Fourth, coordinate rack, lifts, ports, fire, maintenance, WMS/WCS/PLC and network in one responsibility matrix, with change control. Fifth, complete FAT, installation and SAT, including stock migration and operator training. Sixth, confirm stable operation and accept the service and handover package. The project ends when the site can process real orders reliably, not when hardware arrives.
Frequently asked questions
Is shuttle or stacker-crane AS/RS faster?
The technology name does not decide. Loads, order waves, lifts, aisles, ports and sequencing determine whole-system output. Compare peak 15-minute, sustained one-hour and failure-mode performance using identical data. Shuttle vehicle count is not a dispatch guarantee.
How should shuttle AS/RS throughput be designed?
Convert order lines to tote moves; include replenishment, returns, empty containers and exceptions. Model shuttles, lifts, conveyors and ports with queues. Separate normal peaks, planned maintenance and one-component failure. Require simulation on representative order history.
What belongs in a shuttle system RFP, FAT and SAT?
The RFP needs load form, SKU and lot profile, time-based demand, building conditions, interruption tolerance, IT ownership and maintenance terms. FAT covers control, safety, exceptions and interfaces; SAT covers installed performance with real loads and orders plus fault recovery. Agree pass thresholds and log evidence before contract award.
Can a Thai AS/RS project use BOI incentives?
Eligibility depends on the specific activity and application. BOI publishes conditions and procedures, but purchasing a warehouse system alone does not confer approval. Confirm the case with BOI and present the investment case without a benefit until approved.
Summary
Shuttle AS/RS selection is about delivering actual orders within the required time. Measure density as usable slots, redundancy as service left after failure, and installation quality through FAT and SAT evidence. Include Thai site, building, IT, fire and maintenance conditions in the RFP, and compare bidders with the same demand data.
If you are considering a shuttle system for a Thai factory or distribution site, TOMAS TECH can help turn load and order data into comparable RFP conditions. You can contact us while the technology choice is still open.
Primary references
- Dematic, Multishuttle — product components and applications; claimed benefits are not a project guarantee.
- Swisslog, CycloneCarrier — small-load storage, GTP supply, buffer, depth and lift options.
- SSI SCHÄFER, Shuttle Systems brochure — example of level shuttles, lifts and a maintenance platform.
- Thailand BOI, Smart and Sustainable Industry — published incentive conditions; check project eligibility with BOI.
- ISO 12100:2010, Safety of machinery — principles for machinery risk assessment and risk reduction.
Sources checked 8 November 2026. Recheck product configurations, incentive rules and standards at procurement.
