Eleven Seconds per Tote

The single most informative number in any distribution centre is not headcount, floor area or annual throughput. It is the cost per pick — the fully loaded expense of moving one item from storage into a customer's order. Every decision about automation resolves, eventually, into whether it moves that number and by how much.

In a conventional manually operated facility, the figure typically falls between twenty and forty cents. Roughly sixty per cent of that cost is walking.

The travel problem

A picker working a traditional aisle layout spends the majority of the shift in transit. Industry time-and-motion studies consistently attribute between fifty and seventy per cent of order-picking labour to travel between locations rather than to the act of picking itself. The worker is paid for the whole shift. Value is created only during the small fraction in which a hand closes on an item.

Three structural responses exist, and they are worth distinguishing carefully because they are frequently conflated.

Better routing. Batch multiple orders into a single trip and sequence the locations to minimise distance. This requires no capital equipment beyond software and typically reduces travel by twenty to thirty per cent. It remains, in most facilities, the highest-return intervention available and is routinely skipped in favour of more impressive-looking projects.

Better slotting. Place fast-moving items nearest the despatch point. Demand in most catalogues follows a steep distribution in which a small minority of stock-keeping units accounts for the majority of movement. Re-slotting quarterly against actual velocity data yields further double-digit reductions. Again: software, not steel.

Eliminate travel altogether. This is the goods-to-person architecture, and it is where the capital goes.

Goods to person

In a goods-to-person system, inventory is brought to a stationary operator rather than the operator being sent to the inventory. Two implementations dominate.

Mobile robots lift entire storage pods from the floor and transport them to a pick station. The pods are not assigned fixed homes; a returning pod is stored wherever a slot is nearest, and the system tracks its location. Storage density improves because human aisle width is no longer required, and the floor reorganises itself continuously so that popular inventory drifts towards the stations without anyone planning it.

Alternatively, a dense grid of stacked bins is served by robots travelling across the top surface, lifting totes vertically through the stack. Cube utilisation is exceptional — there are no aisles at all — at the cost of retrieval time for items buried deep in a stack.

The operational effect in either case is the same. The operator stops walking. Reported pick rates rise from approximately 100–150 units per hour in manual aisles to 300–600 at a well-configured station, with the variance driven mainly by item handling characteristics rather than by the technology.

Metric	Manual aisle	Goods-to-person
Units picked per hour	100–150	300–600
Travel share of labour	50–70%	~0%
Storage density (index)	100	160–220
Capital cost per unit of throughput	Low	High
Practical payback period	—	3–6 years

The unsolved step

Notice what has been automated and what has not. Transport is solved. Storage is solved. Sequencing is solved.

The grasp is not.

An operator at a pick station reaches into a tote containing an unknown arrangement of items, identifies the correct one, and removes it without disturbing the rest. This takes roughly two seconds and is performed without conscious attention. It remains, after twenty-five years of serious investment, the binding constraint on full automation.

The difficulty is catalogue variety. A general merchandise operation may carry hundreds of thousands of distinct items spanning rigid boxes, sealed pouches, transparent packaging that defeats depth sensing, loose apparel, items that deform under suction, and items whose only stable grasp surface is obscured by their neighbours. A robot that handles ninety-five per cent of a catalogue is not ninety-five per cent of a solution, because the remaining five per cent still requires the full human station, the full staffing, and the exception-handling process — which means the labour has not been removed, merely made less predictable.

Current systems combine suction and mechanical grasping, select strategies from learned models, and achieve creditable rates on constrained catalogues. Performance degrades sharply outside them. The honest operational position is that robotic picking is presently viable for narrow, well-characterised product ranges and is not viable as a general replacement.

Where the economics bind

Three constraints determine whether a project proceeds.

Peak ratio. Fulfilment demand is seasonal. A facility sized for peak runs substantially below capacity for most of the year. Automated capacity is a fixed cost that must be paid during quiet months; manual capacity can be scaled with temporary staffing. The higher the ratio of peak to average, the weaker the automation case — a point routinely underweighted in business cases prepared during a busy quarter.

Rigidity. Fixed automation encodes assumptions about item dimensions, order profiles and throughput. Assortments change faster than depreciation schedules. Several large installations have been decommissioned well before end of life not because they failed, but because the business they were designed for no longer existed in that form.

Labour market. The case strengthens as wages rise, and strengthens further where turnover is high. Annual turnover exceeding one hundred per cent is common in the sector, and the cost of continuously recruiting and training a workforce that does not stay is frequently larger than the wage line itself. Automation is often justified less by displacing labour than by removing the recruitment problem.

Conclusion

The industry's trajectory is not towards the unattended facility, which remains a marketing artefact. It is towards a stable division in which machines perform transport, storage and sequencing — tasks characterised by repetition, distance and weight — while humans perform grasping, inspection and exception handling.

Eleven seconds per tote is the current benchmark cycle time at a good station: the tote arrives, the operator picks, the tote departs. Every remaining cost improvement in this industry is a fight over fractions of that eleven seconds, and most of the fractions that are easy to win have already been won.