Warehouse Slotting Optimization: Does AI Halve Picking Time?

"Cut your picking time in half." Nearly every warehouse software site says it. Really? By changing where products sit on the shelves, without buying a single robot, with the same crew, does order picking time actually halve?

Short answer: not from slotting alone, but slotting is the cheapest and fastest place to start. In this article we will first squeeze the vendors' numbers into a realistic gain band, then explain how AI-driven warehouse slotting optimization works and how far a spreadsheet can take you. You can start without a warehouse management system; we will also say when one becomes necessary.

This piece is part of our logistics and distribution series; the wider picture is in our AI by industry map.

What share of picking time is spent walking?

The most-quoted breakdown in warehouse literature goes like this: about 55 percent of order picking time is travel, 15 percent is searching, 10 percent is waiting, 10 percent is the actual pick, and the rest is paperwork. The picker's day is spent largely empty-handed in the aisles. Slotting optimization targets exactly that 55 percent.

Let us question the figure, because questioning is this article's job. It comes from a decades-old facilities planning textbook and is now copied from one vendor blog to the next. Your own share may differ: in a small single-aisle warehouse with a narrow range, travel drops; in a large, many-SKU e-commerce operation it climbs. But the direction holds almost everywhere: the biggest time item is walking.

Now the math. If travel is 55 percent and a good slotting plan cuts travel by 25 to 40 percent, the gain on total picking time lands at 14 to 23 percent. To "halve" it, you either start from a chaotic, never-organized warehouse, or you stack batch picking, zone picking and route optimization on top of slotting. Most vendor figures describe that combined scenario and print only "half" in the headline.

Slotting on its own trims picking time by 15 to 25 percent. Halving is possible, but not with one move; it takes three or four measures added together.

Can you do slotting without a WMS?

Yes, for static slotting a spreadsheet is enough. Count order lines per SKU over the last six months, move the top-selling 20 percent to the bins nearest the packing station, put heavy items low and light items high. That is classic ABC analysis and it needs no warehouse management system. Dynamic, continuously updated slotting is different: that requires a WMS, or at least barcoded location tracking.

Why doesn't everyone do this, you might ask. We see three reasons. First, order history is rarely clean in a small business: marketplace orders in one place, wholesale orders in another, SKU codes inconsistent. Second, there is no bin map; locations known as "that shelf in the corner" have never been written down with coordinates. Third, slotting gets done once and left; three months later the season turns, A items become C items, and the layout decays.

That third reason is where AI enters. Rule-based ABC takes a periodic snapshot. AI-driven dynamic slotting watches order patterns continuously and generates its own suggestions: "this SKU's velocity rose, bring it forward; these two are always ordered together, put them side by side." Affinity slotting, the market-basket logic of co-ordered items, is feasible in a spreadsheet for a few hundred SKUs and practically impossible for a few thousand.

What does AI slotting optimization actually do?

Dynamic slotting software combines four data sets: order history (which SKU, how often, with what), SKU dimensions and weight, the bin map, and the current layout. From these it calculates the best bin for each SKU, produces a move list, and measures picking times as moves are made. The learning part is noticing velocity changes on its own.

The academic side deserves a look too. A review covering 142 peer-reviewed papers from 2010 to 2025 reports AI approaches delivering 15 to 45 percent cycle-time reductions across warehouse functions; that band covers slotting, task assignment and routing together, not slotting alone. We could not confirm in this round whether the review itself is peer-reviewed or a preprint; read the number with that caveat.

Vendor claims are more generous: "25 to 40 percent less travel, 30 to 50 percent higher pick rate." Those figures are unsourced and probably drawn from best cases. The math from the top of the article still applies: cut travel by 30 percent and you have gained about 16 percent on total time. That is a serious number. It is just not "half."

Why "put the best sellers by the door" isn't enough

Because one-dimensional ABC ignores four things: item weight, co-ordered items, aisle congestion and seasonality. Pile every A item into one aisle and pickers queue behind each other; put a heavy A item on a top shelf and picking gets faster while backs get hurt. Good slotting balances speed against ergonomics and flow.

The mistake we see most often in the field: the slotting gets done, the measuring does not. If picking time per order, lines per hour and walking distance per picker are not recorded, nobody knows whether the layout worked. Those three metrics can be started before any software, with a stopwatch and a table.

The same measurement discipline applies out on the road. In our guide to route optimization software we said not to believe fuel savings that were never measured; walking distance inside the warehouse is measured with the same logic as a van's fuel.

Why this matters now for an e-commerce warehouse

Same-day and next-day delivery pressure from marketplaces races every minute in the warehouse against the carrier cut-off. Picking accounts for more than half of warehouse operating cost and labor cost rises every year. The cheapest way to get more orders out with the same crew is to shorten the distance they walk, rather than push them to walk faster.

Cloud and rental WMS models have lowered the entry barrier for smaller operations, but almost no vendor publishes a price; quotes depend on users, warehouse structure and integrations. That is why we give no figure here. The decision rule: if the system's monthly cost is below the labor value of the picker hours it recovers, and it prevents even some late-shipment penalties from marketplaces, the math works.

On the delivery side, whether the time window you promise a customer actually holds depends on when the parcel leaves the warehouse; we covered that chain from the vehicle's end in our delivery ETA prediction guide. The warehouse is the first link.

What data do you need? Three tables

Slotting optimization needs three tables: order history, an SKU master and a bin map. All three usually exist somewhere in a small business, but they cannot see each other; the first week of the project goes to putting them in one format. Software selection only makes sense once these tables are ready.

• Order history: order number, date, SKU, quantity. Six months is enough; twelve is better for seasonal businesses. Merge marketplace and wholesale orders into one file.
• SKU master: dimensions, weight, units per case, fragility. This is the only information that keeps a heavy item off a top shelf.
• Bin map: every location's number and its distance from the packing station. Step counts written on a paper sketch will do.

The place projects most often stall is inconsistent SKU codes: the same product under one code on the marketplace and another in accounting. Any analysis done before that mapping splits your best seller in two and shows both halves as "medium velocity." One week of code cleanup returns more than three months of software trials.

Three months in one warehouse: a realistic scenario

Picture an e-commerce warehouse with 1,800 SKUs, 400 orders a day and four pickers. The numbers are ours.

Month one: measure. Nothing moves. Average picking time per order is measured: 6.5 minutes. Daily walking distance per picker is logged with a step counter: 11 kilometers. Order history is merged into one table; SKU codes are cleaned. This is the most tedious and the most valuable month.

Month two: static slotting. ABC in a spreadsheet: 280 of the 1,800 SKUs account for 72 percent of order lines. Those 280 move to the two aisles nearest packing; heavy items go low. The 40 most common co-ordered pairs go side by side. Result: picking time drops to 5.3 minutes, walking to 8 kilometers. An 18 percent gain. Nobody bought software.

Month three: dynamic slotting and batch picking. A cloud WMS goes live; the handheld guides the picker along the shortest route, several orders are picked in one pass, and the system issues a weekly move list. Picking time falls to 3.9 minutes. A 40 percent gain on the starting point. Close to half; and what got it there was batch picking and routing stacked on top of the slotting.

Four pickers, 400 orders a day; a 40 percent gain means capacity for 650 orders with the same crew. Getting through a marketplace sale week with the existing team instead of hiring temps pays for the software in a single campaign period.

Frequently asked questions

Is slotting the same as warehouse layout?

No. Layout is the physical design of racks and aisles; slotting is how products are distributed across those racks. Slotting can change without moving a single rack.

How many SKUs before it's worth it?

There is no published threshold. Our field observation: with more than one picker and more than a hundred orders a day, slotting optimization produces a measurable difference.

How often should the layout be updated?

Static ABC: at the start of each season, at least twice a year. Dynamic systems issue weekly suggestions; you need not apply every one, but do not let them wait three months.

Does AI slotting need robots?

No. Slotting is entirely a software and data decision; the moving is done by people. Robotics vendors' content puts the robot first, which is their perspective.

What not to do

• Do not re-slot during a sale week; picking slows during the moves and you live your worst day on your busiest day.
• Do not pile every A item into one aisle; spread them over two or three so pickers stop queuing.
• Do not apply every software suggestion immediately; review the weekly list, you know the seasonal swings better.
• Do not start without measuring; "it used to be slower" is not a measurement.

So what should you do?

• Time picking per order with a stopwatch for one week. Without that number you cannot test any vendor's promise.
• Merge six months of order history into one table and clean the SKU codes. Find the top-selling 20 percent.
• Number your bin map; write each location's distance from packing. A paper sketch is fine.
• Do static slotting in a spreadsheet and measure for a month. If the gain is under 15 percent, your data or your map is incomplete.
• Move to a WMS once you have seen the gain; when buying, ask "does it have dynamic slotting suggestions and batch picking?" rather than "does it have AI?"

Back to the opening question: does picking time halve? From slotting alone, no; from slotting plus measurement plus batch picking, it gets close. Starting from doubt is cheaper than believing the number on a vendor's site. If you want to run an ABC trial on your own order history, send us the table and we will read the first analysis together.