Storage density vs selectivity: why more pallet places can hold fewer pallets

A freezer 60 × 50 m with 12 m of clear height takes 5,400 pallet places in selective racking. Drive-in racking six pallets deep fits 7,440, 38% more. Fill both with the stock of a distributor whose products have about 7 pallets each, and the drive-in room holds 1,608 pallets, 30% of what the selective room holds. The extra places exist, but most of them stand empty, because a lane holds one product at a time. That gap between places and pallets is the trade-off between storage density and selectivity, and one number decides it more than any other: how many pallets each product has. Every figure below comes from the Cold store layout tool, which runs the same calculation for any room.

Two words for one trade-off

Storage density is pallet places per square metre of floor. Selectivity is the share of pallets you can take without moving another one first. Selective racking (adjustable pallet racking) puts every pallet on an aisle, so its selectivity is 100%, and its density is the lowest of all, because the aisles take a large part of the floor. Every denser system gets its extra places from the aisles, in one of two ways: it closes them (mobile racking, where the racks roll together and only one aisle is open at a time), or it puts pallets one behind the other in lanes (double-deep, push-back, drive-in, pallet flow, pallet shuttle).

Closing aisles costs time: the racks need tens of seconds to roll an aisle open (the tool assumes about 40), and only one forklift can work in each open aisle. Pallets behind pallets cost access, and something less obvious: places that only one product can use.

Places are not pallets

A lane holds one product. When it is restocked, the lane fills; then the stock runs down a pallet at a time, and the places it frees stay empty until the lane is empty, because no other product can go behind the last pallet. Warehouse people call the empty places honeycombing. Bartholdi and Hackman work it out exactly: a lane k pallets deep that is emptied evenly is, on average, (k + 1)/2k full, so it loses (k − 1)/2 places.John J. Bartholdi III and Steven T. Hackman, Warehouse & Distribution Science, release 0.98, § 6.1, pp. 56–58. The book is free to download from Georgia Tech. A product that holds fewer pallets than its lane loses more: two pallets in a lane six deep never fill more than a third of it.

Drive-in racking adds a second loss. The forklift drives into the lane at floor level, so a bay holds one product over all its levels: six deep and 5 high is 30 places for one product. A product with 7 pallets takes the whole bay. That is why drive-in, the densest-looking option on the drawing, holds the least here.

Three things in that table are worth a second look.

• Mobile racking has density without the queue. 8,100 places, 50% more than selective, and every one of them holds a pallet, because every pallet still faces an aisle once the aisle is open. It is also the system seen most often in published European freezer projects.Counted over rack makers’ and builders’ case studies. No survey of rack types in cold stores exists; the market-share percentages that circulate contradict each other and cite no method.
• Double-deep loses little. 7,200 places, 6,632 pallets held: each lane is only two deep, so it wastes at most one place. It needs a reach truck with telescopic forks, and half the pallets sit behind another.
• Deep lanes only pay with deep stock. At 7 pallets per product the shuttle room, with 69% more places than selective, holds 4,061. At 80 pallets per product it holds 8,485: same racks, different stock.

Where the lines cross

The table has two stocks. The chart has all of them: the same freezer with fewer or more pallets per product, the best rack type alone, and the best split of the room into two rack types.

Mobile racking holds the most on its own up to about 35 pallets per product. From about 50, a pallet shuttle does, and its best lane gets deeper as products get bigger: at 80 pallets per product it is 26 deep and holds 9,006, 11% more than mobile racking. From about 2 to about 100 pallets per product the orange line is above the ink one, by up to 10%: splitting the room into two rack types beats every single one. Below that, every product is small and mobile racking alone is as good as any split. Above that, products are so big that one deep rack type does it all.

Most stores should not choose one

Real stock is uneven. A few products have many pallets and a long tail has a few each; in these examples the biggest 20% of products hold 70% of the pallets. The split that follows from it is old warehouse practice: deep lanes for the few big products, open racks for the tail. For the 7-pallet stock the tool puts mobile racking on 65% of the floor and a shuttle 10 deep on the rest, and holds 8,570, 6% more than mobile racking alone. For the 80-pallet stock the split is 20% mobile and a shuttle 26 deep: 9,808, 9% more than the shuttle alone.

Operators who publish their layouts do the same. Coloma Frozen Foods in Michigan built its drive-in freezer 2 to 5 pallets deep “for both storage efficiency and an ability to accommodate a range of SKUs”; Zero Mountain in Arkansas runs 2 to 9 deep by area, with push-back at the docks. Research says the same: for block-stacked warehouses, Derhami, Smith and Gue show that the classic lane-depth model understates the space lost to access, and that bays of several depths, each product in a bay that suits it, beat any single depth.Steel King case study, “Cold storage”, 6 Nov 2023 (Coloma); Refrigerated & Frozen Foods, 6 Oct 2016 (Zero Mountain); both are vendor-sourced. Derhami, Smith and Gue, IISE Transactions 51(9), 2019.

What density costs besides access

Moves. A forklift in a deep lane is slower than one at a rack face: rack makers quote 8 to 10 pallets an hour for drive-in against about 25 for a reach truck in selective racking.Drive-in: Frazier, vendor figure. Reach truck: derived from VDI 2198 data sheets (Yale MR14–25, Linde R14–R18); field figures run from 15 to 35. In a freezer the driver works less of each hour: below −18 °C the German guidance is 90 minutes in the cold, then 30 minutes to warm up, about 75% of the time.DIN 33403-5 recommendations, as summarised by BG BAU, Arbeiten bei Kälte, Infoblatt 4/2022. Mobile racking feels it most: one moving aisle would give this room 9,450 places, but at 400 moves a day the busy hour has about 38, and an open aisle moves about 20 an hour before the cold cuts it to 15. Three aisles have to open at once, so the room gets 8,100 places, not 9,450.

Date order. Drive-in, push-back and a shuttle loaded from one side are last in, first out: the newest pallet is in front. Pallet flow (and a shuttle loaded from the far end) is first in, first out. Products that must leave strictly by expiry date belong where the oldest pallet is always reachable: open racks or flow lanes. The tool keeps them there, which takes floor from the deep zone.

Energy, the reason density matters more in the cold. A cold room costs about the same to cool whatever racks stand in it. At the European survey mean for frozen stores, 69 kWh per m³ a year, this room uses about 2.5 GWh, about €500,000 a year at €0.20 per kWh: about €92 per pallet held in selective racking, €61 in mobile racking.Evans et al., survey of more than 300 European cold stores (ICE-E project): mean 69 kWh/m³ a year frozen, 56 chilled, with a wide spread (6–240 frozen). Volume was the strongest driver of energy use. The same holds for the building: the insulated shell is paid per square metre, so every pallet that fits lowers the cost of each pallet stored. That is what pushes cold stores towards dense racks, and why honeycombing is expensive there.

Check it for your own store

1. Get pallets per product from a stock report. Count the products in stock and the pallets; divide. Then sort the products by pallets and add up the biggest 20%: what share of the pallets do they hold? Those two numbers decide more than the room does.
2. Ask the rack supplier for pallets, not places. A supplier’s drawing counts places. Ask how many pallets of your stock it holds on an average day, and how they counted the lanes that are partly empty.
3. Count the busy hour. Moves a day times about 1.5, divided by the working hours, gives the peak hour. Check it against the moving aisles in mobile racking and against deep-lane trucks in a freezer.
4. Count the products that must leave strictly by date. They need reachable places, whatever the density.
5. Plan for the peak, not the average. German cold stores averaged 70.6% pallet occupancy in 2024 and 73.6% in 2025, 81.4% in their record year 2020. Rack makers put “practically full” at about 85%.VDKL monthly occupancy survey, reported by VerkehrsRundschau, 6 Feb 2026 and 29 Jan 2021. The 85% is an operating target repeated by rack makers and consultants, not a measured value.

Steps 1, 3 and 4 give the tool’s inputs, and it does the rest for any rectangular room: open this freezer in the tool, then set your own room, stock and moves.

Where this breaks

• The spread of pallets over products is assumed. The tool draws it from the average and the share of the biggest 20%. No public data on real spreads exists, and it moves the answer more than any other input; a real stock list is the fix.
• One product per lane. Some sites put several lots of one product in a lane, or mix products and accept moving pallets out of the way. Both raise the fill above these figures, at the cost of more moves.
• Stock runs down from a full delivery to the last pallet. Safety stock that never leaves the lane raises the fill of deep lanes.
• An average day in one rectangular room. No columns, docks, offices or cold lock, no seasonal peak, and no fire, structure or floor-load check: those are a rack supplier’s job.
• Throughput figures are estimates. Moves an hour per rack type come from data sheets and vendor field figures, not from a published standard; none exists for manual trucks.

Sources

• Bartholdi, J. J. and Hackman, S. T., Warehouse & Distribution Science, release 0.98, § 6.1 (lane depth and honeycombing). PDF
• Derhami, S., Smith, J. S. and Gue, K. R., “Space-efficient layouts for block stacking warehouses”, IISE Transactions 51(9), 2019. Abstract
• Evans, J. et al., energy use in European food cold stores (ICE-E survey). LSBU Open Research
• BG BAU, Arbeiten bei Kälte, Infoblatt 4/2022, citing DIN 33403-5. PDF
• VDKL cold-store occupancy, VerkehrsRundschau, 6 Feb 2026 and 29 Jan 2021.
• Steel King, “Cold storage” case study, 6 Nov 2023 (vendor). steelking.com
• Refrigerated & Frozen Foods, Zero Mountain, 6 Oct 2016 (vendor-sourced). Article
• Frazier, drive-in throughput (vendor). frazier.com
• Yale MR14–25 and Linde R14–R18 data sheets (VDI 2198). Yale, Linde

A planning explanation, not a rack design. The numbers are the tool’s model of one example room; rack suppliers check structure, fire rules and floor load. Vendor figures are marked as such.