Smart warehouse / Article
How to Choose Between AMR and ASRS (or Both) for Your Floor
An ASRS is storage and an AMR is transport, so the choice is not really between two products. It is a question about which constraint is costing you money. If the building is full, you have a density problem and automated storage is the answer. If the building is fine but your people walk all day, you have a movement problem and mobile robots are the answer. If both are true, sequence them and install the one that unblocks the other first. And if your inventory records do not match the floor, you need neither yet, because automation will simply make the existing error arrive faster. This article gives you the framework we use on real floors.
The numbers at a glance
01 / The framing
These are not alternatives, they answer different questions
The comparison is put to us the wrong way round almost every week. AMR or ASRS, which is better. The honest answer is that the question has no meaning until you say what hurts. An automated storage and retrieval system is a warehousing system for the storage and retrieval of products. It makes a fixed building hold more and hand product back without a person walking to the location. An autonomous mobile robot is a vehicle. It moves a load between two points and plans its own route.
One changes what your building can hold. The other changes what it costs to move across it. Buying the wrong one is not a performance disappointment, it is a category error, and it is expensive because both are long-lived capital.
The reason the confusion persists is that both get sold as warehouse automation, and both produce a labour saving. But they take that saving from different places. Storage automation removes the search and the retrieval. Transport automation removes the walk between points. If your people are not walking much, transport automation has almost nothing to take. If your building is not full, storage automation buys density you will not use for years.
Why picking sits behind both
Both roads usually lead back to the same cost centre. The literature is blunt about where warehouse money goes: order picking has long been identified as the most labour-intensive and costly activity for almost every warehouse, with its cost estimated at as much as 55% of total warehouse operating expense. So when you are choosing between the two, you are really choosing which part of picking to attack: the retrieval, or the travel.
02 / The two questions
Density need against flexibility need
Two measurements decide this, and both come from your own operation rather than from a specification sheet. The first is density need: how many more storage positions you require in the same building over the next three years. The second is flexibility need: how often your layout, routes and product mix actually changed over the last two.
Plot yourself in the table below. Most operations know which row they are in within a minute of reading it honestly, and the honesty is the hard part.
| Your situation | Density need | Flexibility need | What fits | What to avoid |
|---|---|---|---|---|
| Building is full | High | Low | Automated storage, sized to your SKU depth profile | Buying transport robots to move product you have nowhere to put |
| Space is fine, people walk | Low | High | Mobile robots on the routes that repeat most | A rack-supported structure you will outgrow the layout of |
| Both, and growing | High | High | Both, sequenced, under one control layer | Buying both at once before either is proven on your floor |
| Neither is clearly true | Low | Low | Slotting, inventory accuracy and process work first | Any capital purchase justified by a peer's installation |
Scroll the table sideways on a phone
The fourth row is the one nobody wants to land on and the one we hand back most often. It is also the cheapest place to be, because the fixes are analysis and discipline rather than capital.
03 / When it is the AMR
Buy the robot when the route is the cost
Transport automation earns its money where the same movement repeats, the distance is real, and the route may change. Goods travelling from receiving to storage, from storage to a production line, from line to dispatch. Each leg is short on its own and enormous across a year.
What makes an AMR the right instrument rather than a conveyor or an AGV is change. A mobile robot carries its route in software, so re-slotting the floor or adding a delivery point is a remap rather than a construction job. That argument is made in full in AMR against AGV against conveyor, and the short version is that infrastructure buys throughput on a route that never moves, while onboard intelligence buys the ability to change your mind.
Know the ceiling before you plan around it. Individual vehicle throughput is modest: field assessments put AGV and AMR pallet throughput at roughly 10 to 15 pallets per hour per vehicle, against conveyor systems sustaining over 200 pallets per hour. A fleet scales that number, but fleets need charging, aisles and traffic management. If your requirement is hundreds of moves an hour along one fixed line, mobile robots are the wrong tool and a conveyor is the right one.
The Malaysian version of this case
On most Klang Valley floors the transport case is not a greenfield design, it is a retrofit into a building with mixed foot traffic, forklifts and aisles that were never planned for robots. That favours machines that navigate rather than machines that follow a fixed path, and it makes the commissioning work, traffic rules, exception handling, and how the fleet behaves at a blocked aisle, more important than the vehicle specification.
04 / When it is the ASRS
Buy the storage when the building is the cost
Storage automation earns its money when you are paying for space you cannot use. The signals are unmistakable once you look for them: product stored in aisles, a second unit rented next door, pallets double-handled because the location was blocked, or volume turned away because there is nowhere to put it.
The gain is real and large. Moving from conventional racking to a shuttle-based system is reported to deliver typical gains of 30% to over 100% more pallet positions, largely because forklifts need aisles 3.5 metres wide or more to manoeuvre, so more than half the floor plan serves transport rather than storage. Automated storage takes most of that aisle back. CODETRACE tote handling runs in an 850mm aisle for the same reason.
Which type of automated storage you then need is its own decision, driven by whether your building gives you height or footprint and by how deep each SKU goes. That comparison is laid out in ASRS types compared. What matters at this stage is simpler: storage automation is justified by the cost of space and the cost of not having it, and in the Klang Valley that comparison is usually against the rent on a second unit rather than against doing nothing.
One caution. Automated storage rewards a stable product profile and punishes a volatile one, because the rack, the lane depth and the tote envelope are decisions you live with. If your unit sizes and SKU depth are still moving every season, be careful about locking the geometry in this year.
05 / When it is both
Both is common. Both at once is not
Growing operations hit both constraints, usually within a year of each other. Storage fills up, and the movement between storage and the line becomes the next bottleneck. So the answer is often both, and the useful question becomes sequence rather than selection.
Phasing is not caution for its own sake. It is how you keep the second decision informed by real data instead of by the assumptions that produced the first one. It also spreads capital, which for most Malaysian manufacturers matters more than the theoretical optimum of a single big installation.
06 / When it is neither
Sometimes the right answer is not to buy anything yet
We do not sell hardware on day one, and some floors should not be buying hardware this year at all. Four situations where automation is premature, and where we will say so:
The first is the one we see most. Accurate stock is the precondition for every other layer, which is why fixing the count is usually the highest-return project available to a warehouse that has never done it. It is also the cheapest, and it makes any automation you buy afterwards work better on day one instead of month six.
None of this means waiting forever. It means the sequence is data, then slotting, then movement or storage. Skip a step and you will pay for it in a business case that quietly underperforms while everyone blames the machine.
07 / The framework
Six steps, in this order
Run these before you take a vendor meeting. They cost time rather than money, and they change what you buy.
Step five is the one most often skipped and the one that most often decides the outcome. Industry practice agrees: fleets should be sized using throughput simulations before procurement. Simulation is also where the cheap discoveries live, that the handover point rather than the robot count is your ceiling, or that a two-hour evening wave needs a buffer rather than a bigger fleet.
08 / On your floor
We study the floor first, then design and simulate
Our sequence is fixed because the expensive mistakes all happen before installation. A floor study first: SKU profile and movement history, real clear height and column grid, inbound and outbound patterns by hour, where the queue forms today, and how far your inventory records sit from a physical count. Then a design. Then a simulation against your own order data. Only then a build.
That order protects you from the two failure modes we see most. The first is buying density into a building that is not actually full, which converts capital into unused positions. The second is buying transport into a floor whose real problem is that nobody trusts the location data, which converts capital into faster errors. Both look like hardware problems afterwards and are design problems in fact.
CODETRACE integrates on site from Shah Alam in Selangor and Batu Kawan in Penang, so the team that models your floor is the team that commissions the system and the team you call in year three. CODETRACE is a member of the NVIDIA Inception program. Send us your drawings and a month of movement data, and we will tell you which of these your floor needs, in what order, and when the answer is not yet.
Full building, buy storage. Too much walking, buy transport. Wrong data, buy neither yet.
FAQ / AMR or ASRS
Questions, answered.
01What is the difference between an AMR and an ASRS?
An ASRS is storage. It holds product in a structure and retrieves it automatically, so the building holds more and a person does not walk to the location. An AMR is transport. It moves a load between two points on a route it plans itself, and it changes that route when the floor changes. One makes your building denser, the other makes movement across it cheaper. They are not alternatives to each other in any technical sense. They are answers to two different constraints that often exist on the same floor.
02Should I choose an AMR or an ASRS first?
Choose by which constraint is costing you more this year. If you are renting extra space, turning away volume, or storing product in the aisles, storage is the binding constraint and the ASRS comes first. If the building has room but your people spend most of the shift walking between points, transport is the constraint and the AMR comes first. If you install storage automation in a building that is not full, you have bought density you cannot use.
03Do I need both an AMR and an ASRS?
Many operations eventually do, because the two constraints appear together as volume grows. The usual pattern is storage automation in the reserve area and mobile robots on the variable legs between storage, production and dispatch, with one control layer scheduling both. The order matters more than the combination: install the system that unblocks the other one first, then measure again before committing to the second.
04When do I need neither yet?
When your inventory records do not match the floor, when slotting has never been reviewed, when demand is still moving too much to size a system against, or when the real bottleneck is a process step rather than a physical one. Automation multiplies whatever discipline you already have. If the count is wrong, automation industrialises the error. Fixing data and slotting costs analysis rather than capital, and it often removes enough of the pain to change what you buy later, or whether you buy at all.
05How do I know how much density I actually need?
Calculate storage positions per square metre today, then model what your three-year volume needs in the same footprint. If the gap is small, better racking, slotting and disciplined putaway may close it without automation. If the gap is large and the building cannot grow, that is the case for automated storage, because density in a fixed envelope is exactly what an ASRS sells. Rent and land cost in the Klang Valley usually make the comparison against a second unit, not against doing nothing.
06What does a CODETRACE floor study involve?
We look at your real floor before we discuss hardware: SKU profile and movement history, clear height and column grid, inbound and outbound patterns by hour, where the queue forms today, and how accurate your inventory records are against a physical check. From that we design a system and simulate it against your own order data. The output is a recommendation with numbers behind it, including the recommendation to do nothing yet when that is the honest answer.
07How long before automation pays back?
It depends on labour cost, volume stability and how much space or throughput the system unlocks, which is why we will not publish a single figure. What we can say is that the payback is decided during design, not during operation. A system sized against an average rather than a peak, or installed on top of inaccurate inventory data, will underperform its business case no matter how good the hardware is. Simulation against your own order history is the cheapest way to protect the number.
Sources / Every figure in this article
Where the numbers come from
Sources are listed by what they are rather than by brand name. Vendor and trade figures are field observations from companies that sell these systems, so read every range as a well-tuned installation rather than a planning number. The decision framework itself is CODETRACE engineering judgement, written as judgement.
Related / Smart warehouse