What actually fits in a cold warehouse

Most cold stores keep everything that is not storage outside the cold room: a chilled loading dock along one long side, which also does the airlock's job, and offices, battery charging and the plant room at one end or above the dock (batteries do not charge below zero). Given the whole building, the tool takes a dock strip off one long side (20 m deep by default, 10 m under 2,500 m²) and a strip across one end for the rest (8 %, or 10 % in small buildings), and lays out the racks in what is left. In eight US cold stores with a published floor split, the cold rooms took 68–83 % of the building. So check which area a quoted pallet count stands on: for one US cold store, the same count is 1.83 places per m² of building and 2.30 per m² of cold room.

Each rack type is laid out in the room both ways round, with EN 15620 clearances, a 0.5 m gap under the sprinklers and coolers, a cross aisle where there is more than one aisle, and the aisle your forklift needs. Levels stop where the forklift's lift ends: 12 m for a reach truck (the tallest lift 13 m, most 8.5–9 m), 6.5 m for a counterbalance truck, 5 m for a pedestrian stacker. Double-deep needs a reach truck. The drawing is the same calculation as the count. Checked against published stores: selective within −6 to +10 %, pallet shuttle within 5–11 %, mobile racking within 3 % of a Spanish freezer (and too high where a vendor quotes the whole site, docks included).

Your products are spread from few to many pallets each: every product at least one pallet, the rest a lognormal spread with the share you set for the biggest 20 % (when the average is low, the share is capped at what it allows, and the tool says so). Stock is counted in whole pallets and runs down a pallet at a time between deliveries. One-deep racks waste nothing; a lane holds one product, so a lane k deep loses on average (k − 1)/2 places per product, the result in Bartholdi and Hackman's Warehouse & Distribution Science (§ 6.1), computed exactly per product. Products go to the deep zone in the order that wastes least. The figure assumes the racks in full use; real stores average 64–82 % over the year, so plan for your peak.

Every split of the room (open racks plus one deep system, in 5 % steps, both directions) and every single rack type is scored on pallets held, share reachable, cost per pallet, forklifts at the peak hour, and how it fares if your products change: in each of four changed mixes (half or double the pallets per product, a flatter or steeper spread) its pallets as a share of the best layout in that mix, taking the worst. Layouts beaten on every measure are dropped. The tabs show the best at each measure; the safest layout gets its own tab only when it is at least 2 points safer than the one that holds the most. Near the top, layouts often hold within 1 % of each other: "most pallets" then shows the simplest of them (one rack type before two, shallower lanes, more moving aisles), so a small change to an input does not redraw the store for a fraction of a percent; the very most is listed beside it.

About 25 moves an hour per reach truck (manufacturers' data sheets, field figures 15–35), fewer in deep lanes (drive-in 10, a vendor field figure of 8–10), and only 75 % of the time in a freezer: below −18 °C a driver works 90 minutes, then warms up for 30 (DIN 33403-5). The peak hour is 1.5 times an even spread over 16 hours. An aisle takes one truck per 20 m of its length, at least one; a layout whose aisles cannot take the trucks it needs is dropped. Mobile racking needs one open aisle per truck and moves about 20 pallets an hour through it, as the racks take some 40 s to open an aisle; the tool picks the fewest moving aisles that keep up.

Columns stand at every crossing of the grid inside the room, 0.4 m square unless you give another size, with 0.1 m clear on each side. The first column lines are one spacing from the walls unless you set them nearer. The tool tries both ways rack designers place them. Either each column line is the gap where two racks meet back to back, so the racks between two lines are laid out like a room of their own; or, for selective and double-deep racking, a line that falls on the racks stays there and its columns take the bays they stand in, while a line that would fall in an aisle becomes such a gap. Both orientations and a few starting offsets are tried, and the drawing is the count. Mobile racks cannot roll past a column, so every bay between column lines gets its own moving aisle; deep lanes have to fit between the lines. "Why this layout" compares with the same room without columns. Real designs set aisles row by row and can win back part of the loss: published figures are 1 to 4 % when the grid fits the racks and 10 to 30 % when it does not.

A picker takes cases from a pallet on the floor all day, so pick faces need aisles that are always open: selective racking. Mobile racking opens one aisle at a time, and a deep lane's front place holds the next pallet of its lane. A picked product keeps one pallet in its face and its other pallets wherever they fit best. With picking the tool also tries a selective pick strip 6, 9 or 12 m deep along one long wall, with mobile and deep racks behind it, and a selective zone in front of mobile racking. The pick faces never outnumber the floor places of the selective racks, and "Why this layout" says what they cost against the same room with whole pallets only. The "when each rack type wins" chart leaves case picking out: it compares rack types for storage.

Rack prices per place from German and US price sources disagree by half, so treat them as ±50 %. Forklifts (€45,000 each) and shuttle vehicles (€30,000) are counted; no list price was found for either, so both are assumptions. Drivers are not counted. The room itself (€2,000 per m² by default) is most of the investment, so cost per pallet mostly follows pallets held. Energy per pallet uses the European survey mean (69 kWh per m³ a year frozen, 56 chilled).