A fulfillment warehouse does not gain capacity in proportion to the people hired for peak. By the second and third seasonal operator added to the same zone, the gain per person falls: aisles get congested, packing stations become the bottleneck, and the error rate climbs at exactly the moment an error costs the most. On top of that, the people starting a shift in November are the least trained of the year, working the busiest flow of the year.
Which makes peak season a flow problem, not a headcount problem. And a flow problem is solved with equipment only at the points where the flow genuinely breaks, which are almost never the points where the pain is felt.
Where the flow breaks first
The reflex, when orders pile up, is to put people on picking. That is where the congestion is visible, where the pressure is audible, where the request for help comes from.
Most of the time the real bottleneck sits further downstream, at consolidation and dispatch. Picking is spread across the whole floor and absorbs extra people reasonably well, up to a point. Consolidation and packing are concentrated in a handful of physical stations, and you cannot add a fourth person to a bench designed for two. Parcels accumulate between zones, someone shifts them with a pallet truck, and that is precisely where the time gained upstream is lost.
The diagnosis is straightforward, if you run it before the season: measure how long a parcel sits between being picked and entering the truck. If waiting time between zones is comparable to actual working time, the problem is not picking.
There is a second flow that hits the same stations, a few weeks out of phase. Post-peak returns come back in January, in volumes that are hard to forecast, competing for the same benches, the same people and the same space as dispatch. In a conventional operation, a returned item is back in sellable stock in one to two weeks; where returns have their own flow and stations, in three to five days. The difference is not cosmetic. Every extra day is a day the goods are already paid for, occupy space and cannot be sold, in the exact period when they liquidate worst. Plan peak capacity without putting returns in the same drawing and you have planned half the problem.
Calibration figures
The ranges below are indicative, for calibration during an audit, not quotation promises. They serve one purpose: telling you whether your operation sits at the conventional level or at the level of mature operators.
Order-to-ship cycle time conventionally runs between 24 and 72 hours, while the target for operators promising next-day delivery drops below 4 hours. OTIF, the share of orders leaving complete and on time, sits around 90% conventionally and passes 98% at mature operators. In picking, an operator with a paper list handles between 40 and 70 lines per hour, one with an RF terminal and batch picking reaches 100 to 180, and a conveyor-served zone with pick-to-light climbs to between 150 and 300. At packing, a manually fed station handles 20 to 40 orders per hour, an automatically fed one between 60 and 100.
One figure matters more than all the others: picking consumes between 45 and 55% of a warehouse's labour cost, and roughly half of picking time is travel. Not picking. Walking.
That is where the investment logic comes from. You are not buying speed at the workstation. You are buying the removal of travel between stations.
Peak-to-average ratio, the first decision criterion
Before any conversation about equipment, work out a single ratio: peak throughput divided by average throughput.
If the ratio stays below two or three, a system sized for peak stays reasonably loaded all year and the payback works. If the ratio is higher, which happens often in fashion and electronics, a fixed system sized for November runs underloaded ten months of the year. The right answer then is not a bigger system but a fixed core for base volume plus a component added seasonally: extra packing stations, temporary accumulation zones, people on the zones deliberately left manual by design.
This is the criterion that separates an engineering conversation from a catalogue quote. A supplier who never asks about the peak-to-average ratio will size against the number you give them, and the number you give them is the peak, because the peak is what hurts.
What gets automated first, and at what threshold
The order has a logic to it, and the thresholds are orders of magnitude for qualification, not quotation criteria.
Mechanised transport between zones starts to pay off above roughly 300 to 500 parcels per hour on a stable route. It is almost always the first investment, because it removes the pallet-truck moves between picking, consolidation and dispatch, which is exactly the dead time the diagnosis identified.
Automated sorting for dispatch enters the conversation above roughly 1,500 to 2,000 parcels per hour in the window before cut-off. Here is the nuance most often missed: what counts is throughput in that window, not the daily average. A warehouse with 8,000 orders a day compressed into the three hours before carrier collection has a sorting problem. One with the same volume spread across two shifts usually does not. How the equipment actually works is explained in the article on the cross-belt sorter, and the solution-level view is on the sorting and distribution page.
One detail that matters more at peak than nominal throughput: technical availability. Automated systems are typically contracted at between 98 and 99.5% availability, and the gap between those two percentages is measured in hours of downtime on exactly the day you can afford none. A slower sorter that keeps running gets through the season better than a fast one with two hours of intervention at peak.
Ahead of both, if it has not been done already: slotting. Rearranging items by pick frequency shortens travel without buying anything. It is the cheapest gain available in an operation and, in warehouses that grew organically, often the largest one too.
Where what we deliver stops
Above certain thresholds the conversation is no longer about the same kind of supplier, and it is fair to say so upfront rather than afterwards.
Above 3,000 units per hour you enter hub sorting territory, with installations typically running between 10,000 and 100,000 items per hour. Interroll quotes throughputs of up to 30,000 items per hour for its high-capacity cross-belt sorters. That space belongs to integrators specialised purely in high-capacity sortation, such as Vanderlande or Beumer, on projects of a different scale and budget. An operator with a genuine hub requirement should be routed there, not talked into a modular platform.
Similarly, goods-to-person becomes the right answer above roughly 15,000 to 20,000 lines per day with many small SKUs, where it can reach 300 to 600 lines per hour per operator. That is a system type delivered by integrators specialised in automated storage and picking, with a different project profile from transport and sorting.
ZEDlog integrates Interroll platforms in Romania and works in transport, accumulation and sorting below the hub threshold. What sits above the threshold, we say sits above the threshold.
When the honest answer is process, not equipment
Three situations where the right recommendation is to buy nothing this year.
Volume does not support the investment. Below the thresholds above, a fixed system does not pay back, however good it looks in the quotation. The gain comes from slotting, zone organisation and process discipline.
The floor area is too small for flow lengths that matter. In a dark store or a q-commerce operation, distances are short anyway, so mechanised transport has nothing to save.
The data does not exist. Without an order profile, SKU structure and throughput actually measured at peak, any sizing is a guess, and an expensive guess stays expensive once it is installed.
The real decision window
One thing worth saying plainly, in August: a conveyor and sorting system cannot be designed, manufactured and commissioned before Black Friday. Flow study, concept, quotation, execution and commissioning take months rather than weeks, and phased installation in a working warehouse adds time by design.
The correct window is January through summer, for the season that follows. The strongest signal that the moment has come is the season that has just gone badly: unfulfilled orders, missed cut-off, SLA penalties.
For the current year, what remains are the levers that need no project: slotting redone against the season's demand curve, extra packing stations, temporary accumulation zones, a shift plan built around the window before cut-off. None of it fixes the structural problem. It gets you through the season.
The market is not waiting. Online spending in Romania reached approximately 11.7 billion euro in 2024, of which around 7.7 billion on goods, growing at roughly 10% a year according to GPeC, while analysts cited by Ziarul Financiar estimated around 12.8 billion for 2025. Volumes rise every year. The decision window opens once a year.
For how the steps run from data to concept, the guide to warehouse automation in Romania covers the full route, and the segment-specific flows are on the eCommerce and fulfillment page.
