Ecommerce fulfilment is not sized by its average week. It is sized by the four weeks a year when volume doubles, and every other decision — how many pickers, how many forklifts, how much floor you dedicate to staging — is really a decision about how much spare capacity you are willing to pay for all year in order to have it in that month.
Using an AMR for ecommerce fulfillment changes the shape of that decision: instead of hiring for a peak and carrying the cost after it, transport capacity arrives as equipment that can be added when volume climbs and redeployed when it falls. This page covers where the hours actually go on a pick-pack-ship floor, which platform suits which load, and the site checks to complete before any machines land.
Why the average week is the wrong planning unit
A fulfilment centre that ships a steady volume every day has a transport problem that grows slowly and predictably. A centre with promotional peaks has a different problem: the same building, the same docks and the same aisles have to move several times the usual volume for a handful of weeks, and the labour market cannot deliver several times the usual headcount for a handful of weeks.
- Temp labour is a peak solution with a peak problem. Recruiting, inducting and supervising seasonal staff is itself a management load, and the first week of a peak is usually the least productive one.
- Congestion replaces shortage. More people in the same aisles does not scale linearly; at some point pickers, trolleys and forklifts queue for the same corridor.
- The wave is the unit of work. Picking runs in waves against cut-off times, so what matters is whether replenishment and transfer keep up between them — not how many bodies are on the floor overall.
Transport automation targets exactly the part of the peak that scales badly with people: the repeated journey between fixed points.
Where the hours go on a pick-pack-ship floor
Four flows account for most internal movement, and each has a different tolerance for waiting:
| Flow | What it is | What automation changes |
|---|---|---|
| Picking walk | Pickers travelling between locations instead of picking | Goods move to the picker, so the walking part of the shift shrinks |
| Replenishment | Restocking pick faces between waves | Runs become scheduled and repeated, the pattern machines handle best |
| Cross-dock transfer | Moving pallets and cages between inbound, staging and outbound | Long, heavy, fixed-route journeys leave the forklift queue |
| Returns | Reverse flow from receiving back to put-away or disposal | Consistent routing for a flow that is otherwise treated as leftover work |
The mistake to avoid is treating these as one project. Replenishment and cross-dock transfer are the usual starting points, because the routes are fixed, the loads are known and the payback is easy to describe — picking redesign is a bigger change that tends to follow once the transport layer is proven.
Matching an AMR for ecommerce fulfillment to the load and the route
Our AMR range is positioned by the load class and the kind of journey, and the right answer for a site is normally a mix rather than a fleet of identical machines:
- AMR F150 — positioned for light-to-medium intralogistics tasks where the constraint is space rather than weight: narrow workcells, tight aisle entries, equipment that leaves little room to manoeuvre.
- AMR F300 — the mid-range platform for mixed workflows, where the same building contains a tight production or pick zone, a staging area with space to spare and a long run connecting them.
- AMR F600 — the high-capacity platform for demanding warehouse transport, where completing a run in one journey instead of three is the saving, not the speed of the machine.
- The open-deck AMR platform — for sites that want to keep their own carriers: the deck takes the trolley, cage or fixture you already use, instead of forcing a change of container with the change of machine.
Payload, dimensions and battery figures for a specific configuration are confirmed with you before ordering — we would rather verify against your loads than quote a number that turns out to be a class above or below what your routes actually carry. The full range is in the logistics robot category.
Adding capacity without adding people
The commercial argument for AMRs during a peak is not that they are cheaper than labour in the average week — it is what happens in week fifty-one.
- Capacity arrives in days, not hiring cycles. An additional machine can be on the floor and mapped within a fraction of the time it takes to recruit, induct and roster additional staff.
- Peak cost ends with the peak. Seasonal labour commitments outlast the volume; equipment can be redeployed to another site or another shift.
- Scaling is additive, not reorganisational. Adding a machine does not require re-planning rest breaks, supervision ratios or aisle discipline.
- The floor stops being the constraint. Once goods move to the picker rather than the reverse, throughput is limited by pick stations rather than by corridor traffic.
One prior decision makes all of this easier: know whether your site is a good fit for dynamic routing at all. If your routes are genuinely fixed and will stay fixed, the trade-offs in the AMR versus AGV comparison are worth reading before you specify anything — the two approaches optimise for different kinds of change.
What to check before you order
- Floor condition and transitions. Slopes, dock levellers, drainage covers and expansion joints along the whole route, not just the main aisle.
- Clear widths and turning space. Measured at the tightest point with existing racking, and with the worst-case load on the deck.
- Charging positions. Where machines park matters for the pick-and-pack layout, and the position has to be reachable without crossing a pick face.
- Doors, lifts and fire routes. Any automatic door a machine must cross, and whether a lift is part of normal operation rather than an exception.
- Data interfaces. Whether the transport layer needs to talk to your WMS or MES, and who owns the integration on your side.
- Carrier interfaces. If the machines must lift the trolleys, carts or cages you already own, confirm those interfaces before ordering rather than after.
- Network coverage. Full-route Wi-Fi, including mezzanines, cold rooms and the far corners of staging areas.
The floor underneath and the markings around it
Two site details are easy to defer and expensive to retrofit. The first is signage and marking: racks, no-go zones and pedestrian walkways need to be legible to people and to machines, and the conventions for what gets marked and how are set out in the floor marking guide, with warning signs and decal sets covering the operational side.
The second is floor cleaning. Wide, hard aisles between racking are exactly the environment a scrubber handles best — a machine on the scale of the SC80 covers long runs in one pass and returns to charge unattended — and the constraints specific to that environment are covered in the warehouse floor cleaning guide. Scheduling cleaning so it does not collide with peak picking is a design decision, not a scheduling problem to solve later.
How to start
Send us the four numbers that decide the platform: the heaviest load on a normal run, the longest route, the narrowest aisle on that route, and how many shifts the machines must cover. From those we will suggest a platform mix, the charging positions and the markings plan — and tell you plainly if the honest answer is that your flows are still better served by a fixed-path solution. Send the four numbers here, or start from the machine overview and the wider industry guide.
- Peak lasts four weeks, the plan must hold fifty-twoHiring for peak leaves the cost behind long after the volume has gone back to normal.
- Temp labour peaks exactly when it is least productiveThe first week of a surge is the week the extra people know the building least well.
- More people in the same aisle becomes congestionBeyond a point, additional staff queue for the same corridor instead of adding throughput.
- Replenishment and transfers compete with pickingThe internal runs that keep waves moving use the same routes the pickers need.
- Manual transport is unbudgetable at peakOvertime, agency rates and error rates all move the wrong way at exactly the wrong time.
- Capacity arrives in days, not hiring cyclesAn additional platform is mapped and working far sooner than a recruitment round can deliver.
- Peak cost ends with the peakTransport capacity can be redeployed to another site or shift instead of being carried all year.
- Replenishment and transfer become scheduled and repeatableThe two flows with fixed routes and known loads are the ones automation handles best.
- Congestion stops scaling with headcountGoods coming to the picker removes the walking that turns extra staff into extra traffic.
- A transport layer sized by routes, not by rotaThe plan is built around flows and cut-off times rather than around how many people can be rostered.
Going deeper on AMR? Read about how AMR connects to your WMS or ERP, traffic and deadlock planning for many robots, and whether to buy or run it as a service.