Put Wall Sortation for E-commerce: Sizing and Throughput

At 14:50 on a promotion day the put wall is full. Every cubby light is lit, the putters are holding totes they cannot place, and the packers on the far side are standing still because almost none of the lit cubbies is complete. Forty orders are waiting for a single unit each, and those units are sitting in a short-pick queue nobody is watching. The wall did not run out of speed. It ran out of empty cubbies.

A put wall is a small buffer with a hard capacity, and its sizing follows from arithmetic most vendor brochures skip: how fast multi-unit orders arrive, how long each one occupies a cubby, how big the orders are, and how fast the pack side drains them. Get those four numbers right and the wall runs at a predictable rate; guess them and the wall becomes the bottleneck it was bought to remove.

• What this guide covers: which orders to route to the wall, the cubby-count formula, cubby dimensions from order cube, put-side and pack-side staffing, put-to-light confirmation and the orphan-cubby failure.
• What it adds: a worked layout for a US site and an Indian fashion DC with cost per order in $ and ₹, a sizing script, an ownership table and a commissioning checklist for the Logistics and Supply Chain team.

Which orders belong on a put wall and which should bypass it

Only multi-unit orders that fit a cubby should touch the wall; single-unit orders and oversize items should bypass it. A put wall exists to reassemble orders after a batch pick has mixed their units in one tote, so an order with one unit gains nothing from consolidation and still consumes a put, a light cycle and a cubby.

The routing decision happens in the WMS or warehouse execution software when work is released, after inventory has been promised to the order. Our article on omnichannel fulfilment architecture and ATP rules covers that earlier promise across channels. This guide starts once the order has been released to the floor and the question is purely physical: where should its units meet?

• Single-unit orders: pick straight to a pack station or a single-line pack lane. Sending them to the wall wastes roughly one cubby cycle per order.
• Multi-unit orders within cube: route to the wall when the predicted order cube is at or below the cubby design limit.
• Oversize or heavy orders: anything above the cubby cube limit, or with a unit over the shelf weight rating, goes to a floor consolidation lane.
• Orders with gift wrap or kitting: send to a value-added station with its own staging, because their pack time breaks the pack-side rate assumed for the wall.

Cubby count from order arrival rate and dwell time

The number of cubbies a wall needs equals the arrival rate of multi-unit orders times the average time each order occupies a cubby, plus headroom for variation and orphans. That is Little's Law, L = λW, which John Little proved for stationary queues in Operations Research in 1961. Applied to a wall, L is occupied cubbies, λ is orders per hour and W is cubby dwell time in hours.

Dwell time is the number teams underestimate. It starts when the first unit of an order is put and ends when the packer empties the cubby. It has three parts: the fill spread while the remaining units travel from pick to wall, the wait after the order is complete until a packer reaches it, and the few seconds to clear the cubby.

# Put wall sizing (peak hour). All inputs from WMS history, not vendor defaults.
# lam      = multi-unit orders routed to the wall per peak hour
# U        = units routed to the wall per peak hour  (= lam * units_per_order)
# T_fill   = mean minutes from first put to last put of an order
# T_wait   = mean minutes from order complete to packer pull
# T_clear  = mean minutes to empty and reset the cubby
# k        = variability factor (1.2 to 1.4; use the p90/mean ratio of hourly arrivals)
# s        = share of wall orders that go short at pick
# T_orphan = orphan timeout in hours before a cubby is released to the exception rack
# P        = confirmed puts per putter-hour;  u = planned utilisation (0.80 to 0.85)
# t_pack   = seconds to pack one wall order

W_hours  = (T_fill + T_wait + T_clear) / 60
cubbies  = ceil(k * lam * W_hours) + ceil(s * lam * T_orphan)
walls    = ceil(cubbies / cubbies_per_wall)
putters  = ceil(U / (P * u))
packers  = ceil(lam * t_pack / 3600 / u)

# US site example: lam=1080, U=3456, T_fill=18, T_wait=4, T_clear=1,
# k=1.25, s=0.015, T_orphan=0.333, P=600, u=0.85, t_pack=55, 72 cubbies per wall
# W_hours=0.383 -> 518 + 6 = 524 cubbies -> 8 walls; putters=7; packers=20

The variability factor matters because orders do not arrive evenly inside the peak hour. Pull hourly arrival counts for the busiest four weeks from the WMS and divide the 90th-percentile hour by the mean hour; that ratio, usually between 1.2 and 1.4, is a defensible k. The orphan term reserves cubbies for orders that will sit until the timeout releases them.

Put wall cubby count for e-commerce consolidation follows Little's Law (Operations Research, 1961): occupied cubbies equal multi-unit orders per hour times dwell hours. A US site routing 1,080 orders per peak hour with 23-minute dwell needs about 414 occupied cubbies, or 524 after a 1.25 variability factor and orphan reserve.

Sizing each cubby from the order cube distribution

Size cubbies from the 95th-percentile order cube divided by a fill factor, and route the largest few percent of orders elsewhere. Sizing to the largest order makes every cubby bigger, the wall wider and each put walk longer, while sizing to the average leaves a fifth of orders overflowing into a second cubby, which breaks the one-order-one-light rule.

Calculate order cube from item master dimensions multiplied by ordered quantity, summed per order, for at least eight weeks of wall-eligible orders. Item masters are often wrong for new SKUs, so compare a sample against a dimensioner or tape measure before trusting the percentile. Then divide by a fill factor of about 0.6 to 0.7, because soft goods, boxes and polybags never pack a cubby solid.

• US example cubby: 95th-percentile order cube 0.62 cubic feet, fill factor 0.65, so usable volume of 0.95 cubic feet; a 12 by 11 by 13 inch opening gives 0.99 cubic feet.
• India fashion example: folded apparel orders average smaller but are soft and bulky; in the illustrative sample used here, a 300 by 260 by 350 mm cubby held the 95th-percentile order with room to spare.
• Through-depth and push-through: the cubby must be open on both faces so the putter pushes in from one side and the packer pulls from the other.
• Shelf weight rating: confirm the per-cubby load rating against the heaviest 1% of wall orders, not the average.

Height placement is the other dimension. The NIOSH Applications Manual for the Revised Lifting Equation treats 75 cm, roughly knuckle height, as the optimum hand height and 175 cm as the upper limit of vertical reach for lifting. Puts are light, so the equation does not strictly apply, but it is a sound guide: keep the busiest rows near 75 to 130 cm and avoid putting fast movers in the top or bottom row.

Put-side and pack-side labour: where throughput actually caps

On most walls the pack side, not the put side, sets the throughput ceiling. A putter scans a unit and places it in a lit cubby in a few seconds, while a packer must pull every unit, verify the order, choose a carton, add documents and label it. In the US example, seven putters feed a wall that needs twenty packers.

Put rate depends on walk distance across the wall face, scan time and the confirmation method. A 72-cubby face about 13 feet wide keeps the walk short; wider faces add steps to every put. Measured confirmed puts per hour, taken from WMS put-confirmation timestamps rather than a vendor demonstration, should drive staffing.

• Put cycle components: take unit from tote, scan barcode, find lit cubby, place unit, press or break the light beam, return to tote.
• Pack cycle components: see completion light, pull all units, scan each against the order, select carton, insert invoice or packing slip, seal and label.
• Balancing rule: pack labour hours per hour equal orders per hour times pack seconds divided by 3,600, then divided by planned utilisation.
• Two-sided staffing: plan packers per wall face, typically two to three, so each packer covers 24 to 36 cubbies without crossing another packer's lane.

Put wall throughput in the US example is capped by pack-out, not putting: 1,080 orders per hour at 55 seconds each need 16.5 pack hours per hour, or 20 packers at 85% utilisation, against seven putters confirming 600 puts per hour from WMS put-confirmation timestamps on 3,456 units.

When pack-out falls behind, completed orders wait longer, dwell time rises and the wall fills from the pack side. Staffing the put side harder in that situation makes the problem worse. Supervisors should watch the count of completed but unpacked cubbies, and move a putter to pack-out when it exceeds about a tenth of the wall.

Put-to-light signals, scan confirmation and mis-put detection

Every put must be confirmed against the cubby it was meant for, either by a light-button press, a beam break or a cubby barcode scan. A put wall without confirmation turns every slip of the hand into a mis-shipped order, and the error surfaces only when the customer opens the parcel.

Put-to-light uses an indicator on each cubby. After the unit scan, the controller lights the assigned cubby; the putter places the unit and presses the button to confirm. Beam-break walls detect the hand entering the opening and flag a put into the wrong cubby. Scan-to-cubby walls with no lights ask the putter to scan a cubby label, which is slower per put but cheaper to install.

• Light latency: measure time from unit scan to cubby light; anything above about 300 milliseconds makes putters look around and slows the wall.
• Completion signal: use a distinct colour or blink pattern on the pack face when an order is complete, so packers never pull a partial order.
• Multiple putters on one face: assign each putter a colour so two lit cubbies are never confused.
• Controller failure mode: the wall must fall back to scan-to-cubby with printed labels if the light controller loses its connection.

Orphan cubbies: the short-pick failure that freezes a wall

An orphan cubby holds part of an order whose remaining unit will not arrive, usually because of a short pick. Without a timeout, orphans accumulate until the wall cannot accept new orders. The fix is a fixed orphan timer that releases the cubby to an exception rack and sends the order to a recovery path.

The following trace from an illustrative US wall shows how quickly orphans spread on a peak afternoon when nobody owns them.

1. 13:42, short pick: picker scans location A-14-03 for SKU 448812 and finds it empty; the WMS logs short-pick SP-22931 against batch B-7710, but the order lines for order 9001457 and 31 others in that batch stay assigned to cubbies on wall W3.
2. 14:05, wall pressure: 46 cubbies on W3 and W4 are partial and aged over 20 minutes; the WES keeps assigning new orders to remaining empty cubbies.
3. 14:38, put block: W3 reports zero free cubbies; putters hold 11 totes; the put-confirmation rate drops from 590 to 140 per putter-hour.
4. 14:50, pack starvation: packers on W3 have no completed cubbies; carrier cutoff for the 16:00 trailer is 70 minutes away with 312 orders still in batches.
5. 15:10, manual recovery: supervisor pulls orphan units into labelled bags by hand, losing cubby-to-order traceability for nine orders, three of which later ship incomplete.

In the controlled version, the WES applies a 20-minute orphan timer to any cubby whose missing unit has an open short-pick record. At timeout it lights the cubby amber, the packer moves the partial units to an exception rack slot with a scanned label, and the order is re-released for replenishment or a partial-ship decision. W3 never fills, and the trace ends at 14:05.

Worked layout for a US site and an Indian fashion DC

The same formula produces very different walls in the two markets because labour is priced so differently. In the US example, labour dominates cost per order; in the Indian example, the lit wall's capital cost is almost a third of cost per order, which often favours scan-to-cubby walls without lights.

All inputs below are illustrative. The US site ships general merchandise; the Indian site ships apparel and accessories from a mezzanine pick area, so its fill spread is longer. Equipment figures are planning assumptions to be replaced with supplier quotations, not prices from any vendor.

The Indian figure shows why the light hardware deserves scrutiny there. Replacing lights with cubby-label scanning might add a second or two per put and require a fifth putter at ₹115 per hour, about ₹0.21 per order, while removing a large part of the ₹1.58 capital charge. The US site faces the opposite trade, where each extra labour hour costs far more than the lights.

Peak hour, not the daily average, drives the wall count. The US site's average hour carries about 750 orders, so the wall runs at 70% occupancy most of the day. That spare capacity is the price of meeting carrier cutoffs on the busiest afternoon, and it should be agreed with finance before the purchase rather than discovered after it.

Where fixed WES rules decide and where software only advises

Cubby assignment, put confirmation and orphan release should be fixed rules in the WES; forecasting and staffing moves can use predictive tools that advise a supervisor. The test is whether a wrong answer ships a wrong order. If it does, the decision needs a rule with a hard stop, not a probability.

Put wall orphan control works best as a fixed WES timer: a cubby whose missing unit has an open short-pick record is released to an exception rack after 20 minutes. Keeping orphans below 1.5% of wall orders protects roughly 6 of 524 cubbies at a 1,080-order peak hour and prevents wall lockout.

Who owns the cubby map, the assignment logic and the sign-off

A put wall crosses engineering, IT and operations, so each record needs a named owner or the wall drifts from its design within months. The cubby map is physical; the assignment logic lives in software; the exception rack belongs to inventory control. Write the owners down before go-live.

The batch release that feeds the wall is a separate decision with its own owner. Our comparison of batch and wave picking covers how to group picks, and the guide to wave release against carrier cutoffs covers when to release them. Both affect the fill spread, which is the largest part of dwell time.

Commissioning test and buying checklist before go-live

Prove the wall with a recorded peak-hour replay before it carries live orders. A replay uses real order files from a busy week, a planted set of short picks and the planned staffing, and it must show the occupancy, put rate and pack rate that the sizing formula predicted.

Ask every supplier to run this replay with your order file rather than a demonstration dataset, and to state the put rate and light latency they will guarantee. Compare the full cost per order, as in the worked table, rather than the price per cubby. Labour on the pack side and capital per order usually decide the choice.

If you are planning a wall and want the sizing, replay and WES rules reviewed before you commit capital, describe your order profile and site through our request for proposal form. We will return a cubby count, staffing plan and test script built from your own data.