How much does warehouse automation actually cost to pay back?
TL;DR: This page carries one warehouse automation scenario all the way to break-even with every input on the table. A 4,000 sqm two-shift Singapore distribution warehouse, 1,280 internal transport moves per day, six transport AMRs, SGD 828,000 of capital at illustrative market prices, SGD 336,600 of gross annual benefit against SGD 60,000 of annual running cost. Net benefit SGD 23,050 per month. Naive payback 36 months; with a realistic three-month dual-running period and a three-month ramp, cash break-even lands in month 41. The input that moves that number most is not the robot price. It is whether the displaced headcount actually leaves the payroll.
Every warehouse automation article promises an ROI framework and then hands you a list of cost categories. A framework you cannot put numbers into is not a framework. So below is one scenario, worked from first input to break-even month, with the arithmetic visible at each step so you can substitute your own figures and see immediately where the answer moves.
Two things to be clear about before the numbers start. All equipment prices here are publicly searchable market ranges, drawn from vendor list pricing and robot price guides published in 2026. They are not Motionwell quotations and no Motionwell pricing appears anywhere on this page. And the operational figures (move times, moves per shift, utilisation) are model inputs for an illustrative facility, not measured specifications of any delivered machine. Replace them with your own WMS move log before you take anything to a finance committee.
What does the worked scenario assume?
The facility and the work it does
| Input | Value in this model | Where you get your own number |
|---|---|---|
| Floor area | 4,000 sqm (about 43,000 sq ft) | Lease document |
| Operating pattern | 2 shifts x 8 h, 250 days per year | Roster |
| Internal transport moves | 640 per shift, 1,280 per day | WMS move log, peak week |
| Average manual move | 3.6 min door to door, including pick-up and drop | Time study on 30 moves |
| Transport headcount | 12 operators, 6 per shift | Payroll, transport cost centre only |
| Operator productive time | 384 min of a 480 min shift (80%) | Shift observation |
The baseline closes on itself: 6 operators x 384 productive minutes = 2,304 operator-minutes per shift, divided by 3.6 minutes per move = 640 moves. If your equivalent arithmetic does not close, the error is almost always in utilisation, not in headcount. Facilities that skip this step tend to overstate available operator hours by 20 to 30 percent, which inflates every saving downstream. The same discipline applies to throughput baselines generally, which is why OEE is calculated from three separate loss buckets rather than one headline number.
The equipment scope being modelled
Six transport-class AMRs, 500 kg payload, running point-to-point pallet and cage moves between inbound, storage and despatch.
| Input | Value in this model |
|---|---|
| AMR available time | 20 h per day per unit (1,200 min), balance for charging |
| Modelled cycle per move | 6.0 min including docking and traffic queueing |
| Moves per AMR per day | 200 |
| Fleet capacity, 6 units | 1,200 moves per day |
| Demand | 1,280 moves per day |
| Residual manual moves | 80 per day, plus exception handling |
| Retained operators | 4 (2 per shift) |
| Transport roles removed | 8 |
The retained crew is deliberately oversized. Forty residual moves per shift is roughly 144 minutes of one operator’s productive time, so two per shift leaves real headroom for truck loading, damaged pallets, and the days the fleet manager needs a human to unwind a deadlock.
What does the equipment scope cost?
Vendor and aggregator price guides published in 2026 put transport-class AMRs at roughly USD 25,000 to 150,000 per unit depending on payload, docking precision and safety package, which lands in the same band as the SGD 50,000 to 150,000 per unit commonly quoted locally. These are list prices from suppliers rather than an independent industry survey, and they carry two consistent caveats worth building into your own model: integration, network upgrades and support typically add 40 to 60 per cent on top of the vehicle, and per-unit pricing falls materially on fleet orders rather than single units. Fixed-route AGVs sit lower, around SGD 30,000 to 80,000 per unit, because the navigation stack is simpler and the guidance infrastructure moves cost from the vehicle to the floor. This model uses SGD 80,000 per AMR, inside the transport band and deliberately not at the cheap end.
| Line item | Basis | Illustrative cost (SGD) |
|---|---|---|
| 6 transport AMRs, 500 kg class | 6 x 80,000 | 480,000 |
| Fleet management software and WMS interface licence | one-off licence | 60,000 |
| Charging stations, electrical work, Wi-Fi coverage | 3 chargers plus site electrical | 45,000 |
| Integration, interface development, commissioning | project engineering | 120,000 |
| Operator and maintenance training, documentation | initial delivery | 15,000 |
| Subtotal | 720,000 | |
| Contingency at 15% | 108,000 | |
| Total capital, month 0 | 828,000 |
Note that hardware is only 58 percent of the capital. Integration, interface work and site electrical carry the rest, and those are the lines that public price guides never show. Anyone comparing two proposals on robot unit price alone is comparing 58 percent of the project. What sits inside a fixed scope, and what is priced as a variation later, is the question worth putting to a bidder before price is discussed, which is covered in more detail in the guide to scope and responsibility when comparing system integrators.
Annual running cost:
| Cost component | Basis | Annual (SGD) |
|---|---|---|
| Maintenance and spares | 5% of hardware value | 24,000 |
| Software support subscription | vendor | 12,000 |
| Electricity, fleet and chargers | metered | 6,000 |
| Internal support | 0.2 FTE technician share | 18,000 |
| Total annual operating cost | 60,000 |
What is the annual benefit, line by line?
First, the loaded cost of one displaced operator. This is where most business cases are quietly wrong, because they use basic wage instead of loaded cost.
| Component | Monthly (SGD) | Source of the number |
|---|---|---|
| Gross monthly wage | 2,600 | Mid of the SGD 2,200 to 2,800 band Singapore job boards show for warehouse assistant and forklift roles |
| Employer CPF at 17% | 442 | Statutory rate for employees aged 55 and below |
| Other employer-borne cost at 12% | 312 | SDL, work injury insurance, medical, PPE, supervision share |
| Loaded monthly cost | 3,350 | |
| Loaded annual cost | 40,200 |
If the roles are Work Permit holders, add the applicable foreign worker levy. That raises loaded cost, raises the benefit, and shortens payback, so leaving it out keeps this model conservative.
| Benefit line | Basis | Annual (SGD) |
|---|---|---|
| 8 transport roles not backfilled | 8 x 40,200 | 321,600 |
| Transport-related overtime removed | about 640 h at 1.5x an SGD 15.60 base hour | 15,000 |
| Error, mis-pick and damage reduction | not booked | 0 |
| Recovered floor space | not booked | 0 |
| Gross annual benefit | 336,600 | |
| Less annual operating cost | (60,000) | |
| Net annual benefit | 276,600 | |
| Net monthly benefit | 276,600 / 12 | 23,050 |
Error reduction and space recovery are both real and both booked at zero here on purpose. Error cost is defensible only if you already track mis-pick and damage cost per event; if you do not, a finance reviewer will discount the line to zero anyway. Floor space only converts to cash if you actually give the space back or fill it with revenue-generating stock, and at the SGD 1.50 to 3.50 per sq ft per month that Singapore industrial space commonly rents for, an unused recovered aisle is worth exactly nothing. Booking both at zero means every number above the line survives challenge.
When does the investment actually break even?
The naive answer is 828,000 / 276,600 = 3.0 years, or 36 months. That figure is wrong, and it is the figure most vendors quote, because it assumes benefit starts the day the equipment is paid for.
It does not. There is a commissioning period where you run manual and automated flows in parallel and carry both cost bases, then a ramp while traffic rules, docking positions and exception handling settle. This model books zero benefit in months 1 to 3 and half benefit in months 4 to 6.
| Period | Benefit booked | Cumulative cash (SGD) |
|---|---|---|
| Month 0 | (828,000) capital | (828,000) |
| Months 1-3, dual running | 0 per month | (828,000) |
| Months 4-6, 50% ramp | 11,525 per month | (793,425) |
| Months 7-12 | 23,050 per month | (655,125) |
| Months 13-18 | 23,050 per month | (516,825) |
| Months 19-24 | 23,050 per month | (378,525) |
| Months 25-30 | 23,050 per month | (240,225) |
| Months 31-36 | 23,050 per month | (101,925) |
| Months 37-40 | 23,050 per month | (9,725) |
| Month 41 | 23,050 | +13,325 break-even |
Six months of ramp costs five months of payback. Over five years the cumulative net position reaches about SGD 451,000, a 54 percent return on the SGD 828,000 committed. Discounting the annual cash flows at a 10 percent cost of capital gives a net present value of roughly SGD 126,000 over the same five years. Positive, worth doing, and nothing like the returns implied by a 12-month payback headline.
Which input breaks the business case first?
Rerunning the model with one input changed at a time, everything else held:
| Change from base case | Net monthly benefit (SGD) | Break-even |
|---|---|---|
| Base case | 23,050 | Month 41 |
| Loaded labour cost 4,000 instead of 3,350 | 28,250 | Month 34 |
| AMRs at 55,000 per unit, low end of the public band | 23,675 | Month 33 |
| Fleet covers 900 moves per day, not 1,200, so only 6 roles go | 16,100 | Month 56 |
| Only 5 of 12 roles actually leave the payroll | 13,000 | Month 69 |
Halving the robot price improves payback by eight months. Failing to convert the headcount assumption into an actual payroll change costs you twenty-eight. The equipment quotation is the number everyone negotiates and the least sensitive input in the model.
That has a practical consequence. Before signing capital, get written agreement on what happens to the 8 roles: attrition without backfill, redeployment to work that would otherwise have been hired for, or contract non-renewal. “Staff will be redeployed to higher-value tasks” is not a financial benefit unless a req somewhere else closes as a result. If nobody will commit to that in writing, model the project at five roles and see whether it still clears your hurdle rate.
Which equipment type should you model?
Short version, because this belongs elsewhere on the site rather than here. AGVs follow fixed guidance and suit stable high-volume lanes at a lower unit price with higher floor-infrastructure cost. AMRs navigate dynamically, cost more per unit, and avoid floor work, which is what makes them viable in leased space and in facilities that re-rack every year. ASRS is a different category of decision, changing storage density rather than transport labour, and belongs in a capital plan with a much longer horizon. The full navigation, payload, deployment and traffic-management comparison sits in the AMR vs AGV comparison guide, and the system-level view of storage, transport and end-of-line equipment is on the warehouse and intralogistics automation page. If the end-of-line is your constraint rather than internal transport, the arithmetic above transfers directly to robotic palletizing systems by substituting cases per hour for moves per shift.
Whichever platform you model, the capital structure holds: hardware around 55 to 65 percent, integration and site work the balance.
What does this model deliberately leave out?
Costs that appear during implementation and rarely in the business case:
- Safety assessment and zone design. Mobile robots sharing aisles with pedestrians and manual forklifts need a risk assessment, scanner field configuration and traffic rules. Budget engineering time for it. The standards framework is set out in the notes on collaborative and mobile robot safety standards and the machine safety and compliance work that goes with it.
- WMS and ERP interface ownership. Someone has to own work order handover, status feedback and exception codes on both sides of the interface, through go-live and every subsequent WMS upgrade.
- Network segmentation. A fleet on the same flat network as office IT is a finding waiting to happen; segmentation and access control are described in the IEC 62443 industrial cybersecurity overview.
- Map and layout maintenance. Every rack move is a map update. Small cost, permanent cost.
- Process redesign and SOPs. The workflow around the transport step changes, and the training that goes with it is not a one-off.
Adding these back typically consumes most of the 15 percent contingency in the capital table, which is why the contingency is there rather than in a footnote.
How do you run this on your own numbers?
Six steps, in this order:
- Pull one peak week of move data from your WMS. Moves per shift, distance bands, and how many are exceptions. Peak week, not average week, because the fleet gets sized on peak.
- Time thirty manual moves. Door to door, including waiting. Do not use a standard time.
- Compute utilisation and close the loop. Operators x productive minutes / minutes per move should reproduce your observed move count. If it does not, fix that before going further.
- Size the fleet against peak demand, then subtract. Fleet capacity below demand is fine and normal; what matters is what the residual manual work requires in headcount.
- Build loaded labour cost from payroll, not from wage. CPF, levy where applicable, insurance, supervision.
- Book only benefits someone will sign for. Then apply the ramp and read the break-even month off the cumulative column.
Motionwell has delivered AGV installation and AMR-based transport work in Singapore, and is the authorised SIASUN distributor for Southeast Asia. A QA lab automation programme, re-ordered across four consecutive years, combines AMR transport, a collaborative robot, vision and server-based task scheduling in a QA laboratory; the QA lab automation case study describes how that scope was phased. Phasing matters for exactly the reason the sensitivity table shows: a pilot zone produces real move times and real headcount outcomes, and those two numbers decide the business case for the full rollout.
Next steps
To have this calculation run against your facility, contact the Motionwell engineering team with four things: one peak week of WMS move data, your current transport headcount by shift, your loaded labour cost per operator, and your facility layout. That is enough to size a fleet and return a payback model in the format above, with the assumptions listed so your finance team can challenge each one.
Figures on this page are illustrative market ranges for modelling purposes and do not represent Motionwell pricing, quotations, or a commitment on delivered performance.