The racking was specified for a business that received pallets and shipped pallets. It still does that well. But somewhere in the last two years the orders changed — more lines, smaller quantities, cartons and singles instead of full pallets — and the same steel is now being asked to do a different job.

Nothing has broken. The symptoms are softer than that: pickers walking further than they used to, a forklift constantly interrupting picking to let a pallet down, part-pallets accumulating at ground level, and a warehouse that feels full while the racking above head height is half air.

This is one of the most common reasons a warehouse stops performing without anything visibly going wrong, and it is a layout problem rather than a capacity problem.

Selective racking is optimised for pallet moves, not for hands

Selective racking earns its place because every pallet is directly accessible. That is exactly what you want when a truck is putting away and retrieving whole pallets.

Piece picking asks something different. A person on foot needs to reach the goods without a machine, which in practice means the lowest one or two levels. Everything above that is storage, not picking. So the number of positions your operation can pick from is not the number of pallet positions you own — it is roughly the number of bays in the run, once or twice over.

That is the whole constraint. Storage capacity scales with height. Pick faces scale with floor length. As soon as your SKU count grows faster than your floor, the two stop matching, and the operation starts paying for the gap in walking time and forklift interruptions.

The four symptoms, and what each one is telling you

Pickers waiting for a truck. The clearest sign that a fast-moving SKU has no permanent ground-level face. Every let-down is a pallet move that produces no throughput, and it blocks the aisle while it happens.

Long walks per order. If the items that appear on most orders are scattered down the run rather than grouped near despatch, the operation is paying for that on every single order, all day. Velocity grouping is free to implement and is almost always the first thing worth fixing.

Part-pallets on the floor. Broken pallets parked in the aisle or at the end of a run are usually a signal that there is nowhere designed for them to live. It is also a safety problem, because ground-level obstruction is exactly where forklift impacts concentrate.

Replenishment happening during picking. If the only way to keep faces stocked is to interrupt the pick, the reserve is in the wrong relationship to the face. On a well-set-up run, replenishment is a separate task done in a separate window from above the face it feeds.

The options, cheapest first

1. Dedicate the lower levels and re-zone by velocity. Cost: almost nothing. Decide formally that levels one and two are pick faces and everything above is reserve, then move your fastest-moving SKUs to the bays nearest despatch. Most warehouses running into this problem get a real improvement here before spending anything, and it tells you how much of the problem is genuinely structural. Our note on SKU velocity as a warehouse outgrows its racks covers the thinking.

2. Put slow and small items on boltless shelving. A pallet position given to a SKU that ships four units a month is expensive real estate. Boltless shelving gives many more small faces per metre of floor, and it is the right home for the long tail. We set out the trade-off in boltless shelving versus pallet racking. The honest limit is weight and pallet handling — boltless is for hands, not forks.

3. Flow racking for the genuinely fast movers. Loading from the back and picking from the front separates replenishment from picking completely, which is the specific problem in symptom four. It suits a small number of high-velocity lines rather than a whole warehouse, and the depth and pitch have to match the carton or pallet, so it is designed rather than bought off a list. Our piece on pallet flow racks explains where the economics work.

4. Add floor area with a mezzanine. If the constraint really is floor length rather than layout, more pick faces means more floor, and going up is usually cheaper than moving. That is a bigger decision with submission implications, and it is worth running the numbers first — we built a break-even comparison against expansion for exactly this decision.

What you give up, stated plainly

Every pick face you create takes something from bulk storage. Reserving two levels for picking across a run removes those positions from your pallet count, and if you were already tight, that shortfall has to go somewhere — higher levels, denser storage elsewhere, or off site.

This is why the question should be answered with numbers rather than instinct. Counting how many pallet positions a warehouse actually holds before and after is a half-day of work and it prevents the common outcome, which is solving a picking problem and creating a storage one four months later.

Three things that catch people out

Decking. Racking configured for pallets often has no decking at all, because a pallet spans the beams. The moment a person is placing cartons and loose items by hand, the level needs a deck — and the choice between wire mesh, steel panel and chipboard is not cosmetic. We covered the decking options and where each belongs separately.

Load notices. If beam levels are moved or a level is repurposed, the safe working load for that configuration changes, and the load notice on the run has to reflect what is actually installed. A rack signed for a configuration it no longer has is worse than an unsigned one, because people trust it.

People and trucks in the same aisle. Picking on foot in aisles designed for machine access changes the traffic pattern the layout was drawn around. Aisle width, one-way rules and where pedestrians cross all deserve a fresh look, alongside the escape route and exit access that the racking layout has to keep clear.

What to measure before changing anything

Five numbers, and none of them require a consultant.

1. Active SKUs versus ground-level faces. If SKUs materially exceed faces, that gap is your problem in one line.
2. Lines per order. The higher this is, the more walking dominates, and the more velocity grouping is worth.
3. Order share by SKU. The familiar pattern is that a small proportion of SKUs appears on most orders. Those are the ones that deserve the best faces.
4. Let-downs per day. Count them for a week. This is the cost of missing pick faces, expressed in forklift moves.
5. Walking distance for a typical order. Pace one out. It is usually a longer number than anybody expects.

What it costs

We do not publish a price list, and the reason is the same here as anywhere else on the site: the same "selective racking per bay" can run from S$380 to S$1,200 depending on load case, height, beam length and seismic class. Reconfiguration is even more variable, because a large part of it is labour and sequencing around a warehouse that has to keep operating while the work happens.

What we can say is the order of spend. Re-zoning by velocity is free. Decking and load notices on levels you are repurposing are modest. Shelving for the long tail is moderate. Flow racking and mezzanines are the real capital decisions, and they are the two worth measuring properly before committing.

The short version

Selective racking scales storage upward and pick faces sideways. When the order profile shifts from pallets to pieces, the pick faces run out first, and the warehouse pays for it in walking and forklift interruptions rather than in anything that looks like a fault. Fix the free things first — dedicate the lower levels, group by velocity — then spend on shelving for the tail and flow racking for the fast movers, and count the pallet positions you are giving up before you give them up.

If your orders have shifted and the racking has not, send us your SKU count, your lines per order and roughly how many ground-level faces you have today. Message us on WhatsApp, or request a site survey and we will walk the run with you and tell you which of these is worth doing.