"We want to automate the warehouse" is a sentence that, in engineering terms, can mean at least seven different projects, with CAPEX varying by an order of magnitude, installation times from three weeks to eighteen months and opposite levels of reversibility. The risk is not overpaying for a system: it is choosing the wrong technology family and locking in for ten years an operation whose mix, channels and volumes will change in the meantime.
This guide does not promote a technology: it lays out how the decision is actually reached, which numbers are needed to take it, and how the return is genuinely calculated. It is written for people who have to build a business case that survives a board review, not for people looking for an Industry 4.0 slogan.
The seven solution families (and what they actually solve)
Before talking about suppliers, you need to understand which physical problem you are solving. Every family buys one thing and pays for another.
| Family | Operating principle | What it buys | Structural limit |
|---|---|---|---|
| Static racking + forklifts (man-to-goods) | The operator walks to the goods, on foot or with a cart/order picker | Minimum CAPEX, total flexibility, zero technology risk | Productivity capped by walking; scales only by hiring people |
| Fixed-guidance AGVs | Vehicles on a fixed path: wire guidance, magnetic tape, floor QR codes, laser reflectors | Absolute cycle predictability, robustness on heavy loads | Every layout change is a physical intervention on the building |
| SLAM-navigating AMRs | Robots that map the environment, localize on natural features and re-plan their routes | Flexibility, fast installation, growth one vehicle at a time | Cycle time varies with traffic; does not increase storage density |
| Goods-to-person with under-rack robots | Mobile robots lift movable racks and bring them to ergonomic picking stations | Picking productivity, ergonomics, modularity, density without human aisles | Requires a flat floor and SKUs compatible with the movable rack |
| AS/RS mini-load and multi-level shuttle | Totes and trays moved by stacker cranes or rail shuttles inside a fixed structure | High density on totes, steady throughput, exploitation of height | Fixed installation: changing throughput or layout means rebuilding the structure |
| Grid-based cube storage (stacked totes) | Totes stacked in a grid, with robots operating from above | Maximum density for low and medium turnover SKUs | Non-uniform access: buried SKUs require extra moves (digging) |
| Pallet shuttle and pallet AS/RS | Shuttles inside the channels, or pallet stacker cranes in a served aisle | Maximum pallet positions/m², handling of high volumes per item | Suited to few fast-moving SKUs; little flexibility on the mix |
One clarification removes half the misunderstandings during a tender: transport and storage are two separate problems. AGVs and AMRs solve transport and do not add a single pallet position to the building's capacity. AS/RS, cube storage and pallet shuttles solve storage and density. Goods-to-person sits in between: it solves picking productivity. If the stated problem is "nothing fits any more" and the proposed solution is a fleet of transport robots, there is a misalignment to clear up before signing.
The nine parameters that drive the choice
The choice is made on data, not on impressions. These are the nine numbers to have in hand before talking to any supplier.
- Order lines per hour — average and peak, separating piece, case and pallet picking. It is the single most decisive figure: it sizes stations, robots and shuttles. It must be measured over at least 12 months, not estimated.
- Number of active SKUs — how many items have moved at least once in the last 12 months, as distinct from items merely present in the system.
- Turnover and ABC profile — the classic "20% of SKUs = 80% of movements" concentration has to be verified on your own data, because it determines which portion of the warehouse is worth automating. Often class A deserves goods-to-person while class C stays on static shelving.
- Clear height of the building — under the truss, net of sprinklers and lighting. Practical orientation: up to 5-6 m counterbalance forklifts do the job; at 10-12 m you are in reach-truck territory; VNA and trilateral trucks reach 13-15 m; pallet stacker cranes work up to 40-45 m in rack-clad structures; mini-load shuttles typically sit between 8 and 20 m. If the building is low, an AS/RS loses most of its economic advantage.
- Required storage density — expressed in pallet positions/m² or locations/m². It is the parameter that justifies the CAPEX of dense systems: if covered floor area is cheap and you have plenty of it, density is worth little.
- Weight and format of the load unit — EPAL pallets, totes, special containers, oversized packages. Standardizing the load unit is a technical precondition, not a detail: many projects get complicated because 5% of the packages are not compatible with automation.
- Seasonality and peaks — the ratio between peak day and average day. A ratio of 3:1 or above shifts the project logic: you size on a high percentile, not on the absolute maximum, and cover the tail with flexible capacity.
- Brownfield or greenfield — in an existing building the dominant constraint is reversibility: what can be done without stopping operations. In a greenfield the building is designed around the system and dense options become far more attractive.
- Modular scalability and stability horizon — how much volumes must grow over five years and with what uncertainty. A system that scales by adding robots and racks has a completely different risk profile from one that scales only by rebuilding the structure.
| If your dominant figure is... | Indicative value | Technology direction |
|---|---|---|
| Low lines/hour, many SKUs, limited budget | < 150 lines/h | Optimized man-to-goods; automate transport only (AMR) |
| Medium-high lines/hour on many piece-picked SKUs | 300-1,500 lines/h | Goods-to-person with mobile under-rack robots |
| High lines/hour on totes, scarce space | > 1,000 lines/h | Multi-level shuttle or mini-load |
| Few SKUs, very high volumes per item | < 300 SKUs, full pallets | Pallet shuttle or pallet AS/RS |
| Repetitive internal transport between departments | > 40 missions/h | AMRs or robotic forklifts |
| Limited clear height | < 7 m | Goods-to-person or AMR: an AS/RS has no height to exploit |
| High, seasonal peak-to-average ratio | > 3:1 | Modular solutions, rentable or step-scalable capacity |
Technical and economic comparison
The useful comparison is not "how much does it cost", but where the cost concentrates and how much it costs to change your mind.
| Criterion | AMR (transport) | Goods-to-person | Fixed AS/RS (shuttle, mini-load, pallet) |
|---|---|---|---|
| CAPEX profile | Low, incremental per vehicle | Medium, modular per robot, rack and station | High to very high, in aisle steps |
| Annual OPEX (% of CAPEX) | 5-8% | 6-9% | 4-8% |
| Installation time | Days-weeks | 6-12 weeks | 4-9 months (shuttle) · 9-18 months (pallet) |
| Invasiveness on operations | Minimal | Low: can start zone by zone | High: clear areas and civil works |
| Flexibility to layout changes | Re-mapping in software | Adding or moving racks and stations | Low: the structure is fixed |
| Density gain | None | Medium: human aisles removed | High to very high |
| Obsolescence risk | Low: vehicles are replaced | Low-medium | Medium-high over a 15-20 year horizon |
| Degradation on failure | Graceful: the fleet keeps running | Graceful | Aisle down: goods unreachable |
| Typical payback | 1.5-3 years | 2.5-4 years | 3-6 years (shuttle) · 5-8 years (pallet) |
The "degradation on failure" row is the one business cases underestimate most. In an AS/RS with one stacker crane per aisle, a failure of that machine makes all the stock in the aisle unreachable: the declared availability (typically 97-99%, computed as A = MTBF / (MTBF + MTTR)) must be read together with the contractual restoration time and the availability of spare parts in Italy. A fleet of mobile robots, by contrast, degrades gracefully: losing one robot out of twenty means losing 5% of capacity, not shutting down a zone.
How payback is really calculated: a worked example
The correct method has five steps, and the third is what separates a serious business case from an optimistic one.
- Measure the man-hours of the current process on the flow you intend to automate.
- Estimate the man-hours of the automated process using the system's contractual performance, not brochure figures.
- Apply a conversion factor: not every hour saved becomes a cost removed. Only hours that translate into fewer hires, less agency labour, fewer overtime hours or redeployment to value-adding tasks actually count. A prudent factor of 70-85% is realistic.
- Value it at the fully loaded hourly cost, not at gross pay: in Italy, for a warehouse operator, the company cost sits indicatively between €24 and €32/h depending on the collective agreement, grade and shift premiums.
- Subtract the annual OPEX of the system and compute at least two sensitivities on volumes.
Current situation (man-to-goods): measured productivity 65 lines/h per operator → 3,600 / 65 = 55.4 man-hours per day.
Goods-to-person assumption: 200 lines/h per station, prudent within the 150-300 range typical of mobile under-rack robot systems → 3,600 / 200 = 18.0 man-hours per day.
Gross saving: 37.4 hours/day × 240 days = 8,976 hours/year.
Conversion factor 80%: 7,181 hours/year actually removable.
Valued at €28/h: €201,100/year.
Assumed CAPEX: €620,000 (mobile robots, movable racks, 4 picking stations, WCS, integration).
Annual OPEX at 7% of CAPEX: €43,400/year.
Net benefit: 201,100 − 43,400 = €157,700/year → simple payback ≈ 3.9 years.
Sensitivity: with volumes at −25% the payback rises to about 5.4 years; with a 100% conversion factor and unchanged volumes it falls to about 3.1 years. The business case should be presented with the range, not with the best number.
Two items are left out of this calculation because they vary too much case by case, but they often improve the picture significantly: the reduction in picking errors (every return caused by a wrong line has a fully loaded handling cost that must be measured internally) and the value of the space released, which becomes dominant when the real alternative is renting or building a new building. The opposite sign must be considered just as honestly: electricity consumption, software subscriptions and training. The full economic analysis for the mobile-robot side — TCO, cost items and the variables that move the payback — is in the dedicated article ROI, TCO and payback of AMRs in intralogistics.
WMS, WCS and integration: where projects get complicated
The separation of roles must be clarified at quotation stage, because it is the first source of dispute. The WMS manages master data, stock, locations, waves and orders. The WCS (Warehouse Control System) — or WES, when it includes execution logic — orchestrates the physical devices: it assigns missions to robots, sequences the totes arriving at the stations, manages queues and recoveries.
The recurring mistake is assuming the existing WMS can perform real-time sequencing. Almost no general-purpose WMS does that well. Three concrete points must be verified before choosing: whether the WMS exposes documented APIs or message interfaces; whether it manages dynamic locations (indispensable with movable racks and cube storage); and whether it can work both in waves and in continuous flow. Integration typically accounts for 15-30% of project engineering: budgeting it from the start avoids the mid-project surprise.
On the fleet side, if vehicles from different manufacturers are present the issue becomes interoperability. The VDA 5050 interface standard exists precisely to let AGVs and AMRs from different suppliers talk to a single master control system: requiring support for it during the tender is an insurance policy against lock-in, even if today you buy from one manufacturer only.
Energy and availability
In mobile-robot systems the energy strategy determines how many vehicles you actually need. Two approaches:
- Opportunity charging — the robot docks at stations during natural idle windows, typically 5-15 minutes at currents in the order of 0.5-2C. No battery handling is required, but it needs well-distributed stations and a fleet manager that plans charging as a mission in its own right.
- Battery swap — maximizes machine time across three full shifts, but introduces a manual activity (or an automatic swap station), a spare battery pool and a dedicated room.
The prevailing chemistry in modern intralogistics is LiFePO4 (lithium iron phosphate): compared with NMC it offers higher thermal stability and long cycle life — indicatively 2,000-4,000 cycles at 80% depth of discharge according to manufacturer specifications — against lower energy density, which matters little in industrial vehicles because the added mass is often useful as ballast. The figure to ask for in a quotation is not the nominal capacity in kWh but the usable availability of the vehicle in the real mission profile, charging included.
On the network side, many anomalies blamed on the robots are in fact coverage problems. The practical conditions to specify in the tender documents: 5 GHz coverage with RSSI no lower than roughly −65 dBm across all operating areas, roaming handover under 100 ms, a dedicated SSID separate from the guest network, and verification of behaviour with the racks full, because goods attenuate the signal far more than the empty structure surveyed during the site survey.
Safety and compliance
Driverless vehicles — AGVs and AMRs — fall under ISO 3691-4 (industrial trucks, safety requirements and verification, part 4: driverless industrial trucks and their systems), which governs person detection, protective fields of the safety laser scanners, speed management, protective stops and operating-area requirements. Rail-dependent storage and retrieval equipment — stacker cranes — has its reference in EN 528. Risk assessment follows the general methodology of ISO 12100 and, on the European market, the whole system remains machinery subject to CE marking, with the legislative framework transitioning to Regulation (EU) 2023/1230, applicable from January 2027.
Three points deserve attention at design stage:
- Mixed human-robot areas. They are the real test. The configuration of the protective fields of the safety laser scanners must be tuned to the actual speed and stopping distance of the loaded vehicle, not the empty one: the difference in mass changes the stopping distance appreciably.
- System boundaries. In goods-to-person and AS/RS installations the robotic zones are generally segregated: the interface points (stations, maintenance entries) and the safe access procedures must be defined precisely.
- Storage structures. Installation tolerances and periodic racking inspections follow the European standards of the EN 15620 and EN 15635 series: in an automated warehouse these are not a formality, because a deformed upright can push an entire shuttle-served aisle out of tolerance.
On the ergonomic side, the benefit of goods-to-person is documentable: removing the walking and bringing the pick into the ergonomic zone reduces exposure to risk factors for musculoskeletal disorders, which remain the most widespread work-related health problem in Europe according to EU-OSHA and are covered methodologically by the ISO 11228 series.
Five mistakes that make projects fail
- Sizing on the absolute peak. Sizing on the worst day of the last three years means buying capacity that sits unused 95% of the time. The correct approach is to size on a high percentile (typically the 95th day) and cover the tail with flexible levers: overtime, back-up manual stations, pulling work forward into the preceding days.
- Ignoring receiving and shipping. Automating picking and leaving the docks untouched simply moves the bottleneck downstream. A system that prepares 3,600 lines/day but has two shipping docks saturated by 16:00 has solved nothing: the flow must be analysed from the inbound truck to the outbound truck.
- Designing on outdated turnover data. An ABC analysis built on an extraction from two years ago, or on an anomalous year, produces systematically wrong sizing. You need at least 12 rolling months of real order lines, with seasonality visible and discontinued items excluded.
- Underestimating flooring and flatness. It is the most banal and most frequent cause of on-site delay. Movable-rack systems and stacker cranes require tight flatness tolerances; even AMRs, far more tolerant, suffer from joints, level differences and steps beyond the order of 10 mm. An instrumented floor survey should be done before signing, not during installation.
- Treating Wi-Fi as an IT detail. A site survey carried out in an empty warehouse does not describe operating conditions. Coverage, access point density and roaming times must be verified with the racks full and the fleet moving.
The recommended path: assessment, simulation, pilot, scale-up
The most effective way to reduce risk is to proceed in phases, deciding at each gate whether to continue.
- 1. Data assessment (2-4 weeks). Extraction of 12 months of order lines, ABC analysis, load-unit master data, mapping of layout and flows, measurement of current man-hours per phase. Without this step any quotation is a hypothesis.
- 2. Simulation (2-6 weeks). A discrete-event model of the flow, verifying behaviour on the peak day and during transients. It sizes stations and fleet and, above all, it uncovers upstream and downstream bottlenecks before they become a physical problem.
- 3. Pilot (8-12 weeks). One zone or one product family, with measurable targets defined in advance: actual lines/hour, error rate, availability, ramp-up time. The pilot is there to validate data and integration, not to run a demonstration.
- 4. Scale-up. Phased extension, with the lessons from the pilot written into the specification. Modular technologies — AMRs and goods-to-person — allow scaling by adding units; fixed technologies must instead be sized correctly at the first installation, which is exactly why simulation is not optional for them.
How PITECH supports the choice
PITECH Industrial Solution operates as a technology and business-development partner for the Italian market, representing selected foreign manufacturers in the field of intralogistics automation. The contribution concentrates where decisions most often go wrong: reading the operational data (lines/hour, ABC, load units, building constraints), translating the data into verifiable specification requirements, selecting the technology and the vendor consistent with those requirements, and building a business case with declared sensitivities.
The portfolio includes the goods-to-person solutions of the Scallog automated warehouse and Multiway robotic forklifts and AMRs, within the broader scope of industrial automation, robotics and material handling solutions. If the comparison to settle sits upstream — fixed-path transport or autonomous navigation — the full picture is in AGV vs AMR: real differences and which one to choose. For an assessment on your real flow, simply describe volumes, SKUs, clear height and constraints via the contact page.
Frequently asked questions about warehouse automation
How much does it cost to automate a warehouse?
There is no price per square metre, because the cost follows the required throughput (order lines/hour) and density, not the floor area. As an indicative order of magnitude on the Italian market: an AMR fleet for internal transport alone starts from a few hundred thousand euro and grows per vehicle; a goods-to-person installation with mobile under-rack robots typically sits between 300,000 and 1,500,000 euro depending on locations and picking stations; a mini-load shuttle or a pallet AS/RS with stacker cranes easily exceeds one million and can reach several million. On top of these figures comes the annual OPEX, typically 5-10% of CAPEX across maintenance, spare parts, software licences and energy. The useful number is not the absolute price but the cost per order line handled over the life of the system.
AS/RS or AMR: which is better?
They solve different problems. An AS/RS (mini-load, multi-level shuttle, pallet stacker cranes) buys storage density and steady throughput by exploiting height: it makes sense when space is the constraint, volumes are stable and the amortization horizon is long. AMRs buy flexibility: they install in weeks, require no building works, are reconfigured in software and grow one vehicle at a time. Rule of thumb: if the bottleneck is space and the flow is predictable for at least 8-10 years, the AS/RS is defensible; if the bottleneck is man-hours of transport and picking, the layout changes or the volumes are uncertain, start with AMRs and goods-to-person. Many mature installations combine both: an AS/RS as a dense buffer, AMRs for variable distribution.
Do I need to change the WMS to automate the warehouse?
In most cases no, but it has to be integrated. The WMS remains the owner of master data, stock and orders; automation brings a WCS/WES that orchestrates the devices, sequences the missions and manages the stations. The typical mistake is expecting the existing WMS to also perform real-time sequencing: almost no general-purpose WMS does that well. Three things must be verified before deciding: whether the WMS exposes documented APIs or message interfaces, whether it manages dynamic locations, and whether it can work both in waves and in continuous flow. Integration typically accounts for 15-30% of project engineering: budget it from the start rather than discovering it mid-project.
How long does it take to install an automated warehouse?
It depends on the technology family, and the difference is an order of magnitude. An AMR transport fleet is mapped and commissioned in days to weeks. A goods-to-person system with mobile under-rack robots typically requires 6-12 weeks from material delivery, most of it for racking assembly, stations and software integration. A mini-load shuttle typically requires 4-9 months. A pallet AS/RS with stacker cranes, especially a rack-clad building, reaches 9-18 months including civil works. Ramp-up must always be added to these figures: contractual performance is normally reached after several weeks of tuning and data cleaning.
Can an existing warehouse be automated without stopping operations?
Yes, and it is the norm in brownfield projects, provided you choose low-invasiveness technologies and proceed zone by zone. AMRs and goods-to-person with mobile robots are the natural candidates: they require no floor works, coexist with people and forklifts and can start on a single pilot area while the rest of the warehouse works as before. Fixed AS/RS installations are far more invasive and generally require clear areas, civil works and downtime windows or temporary external storage. The boundary conditions to verify in every case are floor flatness and condition, aisle widths, network coverage and the electrical power available for charging stations.
What payback should I expect from an intralogistics automation project?
With current labour costs in Italy, the defensible ranges are: 1.5-3 years for AMR fleets dedicated to internal transport, 2.5-4 years for goods-to-person, 3-6 years for shuttle and mini-load systems, 5-8 years for pallet AS/RS, where the value of the land saved weighs heavily. The correct calculation starts from the man-hours that can actually be removed — not from every hour saved on paper — multiplied by the fully loaded hourly cost and by the working days, minus the annual OPEX of the system. Two sensitivities should always be computed: what happens if volumes drop by 20-30%, and what happens if only 70% of the hours saved genuinely translates into lower cost.