Slotting advice usually arrives in one of two forms: a vendor pitch for optimisation software, or a consultant's framework with no numbers attached.
This is neither. It's the actual before and after from a UK pharmaceutical warehouse, with the figures, the sequence, and the mistake we had to go back and fix.
When this work started, the operation was picking around 83,000 lines a day and needed 28 pickers on the stations at all times to hold that line. It was the only way we could survive the volume. Twelve months later, October 2024 to October 2025, the same operation could sustain around 100,000 lines a day with 23 or 24 pickers. Run the productivity maths on those figures and you get roughly 2,960 lines per picker per day at the start and about 4,200 at the end: an improvement of around 40%.
A note on numbers, because credibility matters more than a headline. We have written before about the measurement and coaching side of this same operation, where the published figure was a 21% productivity gain. That number was deliberately conservative, chosen as safe to publish rather than calculated from the raw shift data. The 40% here is the arithmetic of the throughput and headcount figures above. Both describe the same journey; this article covers the layout half of it.
Why lines per picker is the honest lens
Total daily output is a vanity metric on its own. You can grow it by throwing people at the problem, and for years that's exactly what happens in most busy operations. The number that tells you whether the operation is actually improving is lines per picker per day, because it exposes the cost of every extra thousand lines.
At 83,000 lines with 28 pickers, output looked healthy. Productivity was captured in the reports but nobody was monitoring it. The data existed. Nobody was looking at it. That sentence describes more warehouses than most managers would admit.
What we changed
1. We mapped the warehouse in Excel
No slotting software. A spreadsheet with every pick location, a heat map of pick frequency, and a step count for each location from the picking stations. That last part matters: distance is the tax every pick pays, and until you count the steps you're guessing at the tax rate.
The heat map made the problem visible in an afternoon. Fast movers were scattered, some of them sitting in the most expensive locations in the building, while prime locations near the stations held stock that moved once a week.
2. We moved stock gradually, not in a big bang
Around 800 locations moved over the 12 months. Most moves were within the same aisle. Roughly 15% crossed to a completely different station, and those were the balancing moves: spreading workload across stations instead of letting a few busy ones set the pace for the whole floor.
The cadence is the part I'd copy first if I were doing this again. For the first three months we ran dedicated Saturday sessions just for transfers. After that we switched to 15 or 20 moves at the end of a shift whenever there was an opportunity. The weekend model got the backlog down, but little-and-often is what sustained the change for a year. A live operation absorbs small moves without noticing. It fights big ones.
3. We rebalanced the A-frame around fill rates, not just speed
The operation runs an A-frame automated picker. We analysed all 3,500 SKUs and moved the fastest 800 or so onto it. The less obvious change: adding more channels for the fastest products, so the sections ran at speeds people could actually keep up with. Before that, the fastest channels emptied quicker than anyone could refill them, which turns your automation into a bottleneck wearing a high-vis vest.
The A-frame's share of daily picked lines went from about 65% to about 70%. Five points of mix shifted onto automation is a big part of how the manual stations got room to breathe.
4. We balanced the stations
Alongside the cross-station moves, gravity racking went into some of the busier stations. The goal across all of it was the same: stop the operation depending on a handful of overloaded stations and spread the work so the layout, not constant firefighting, sets the pace.
The mistake: optimising for efficiency and forgetting the bottles
The first slotting iteration optimised purely on pick frequency and steps. It ignored handling constraints, and the bottles found the gap. Bottles in this operation have to be bubble wrapped before picking, and the pure-efficiency slotting left them with no suitable locations. We had to revisit the plan and create specific locations for them.
The lesson is blunt: physical and handling constraints are inputs to the optimisation, not exceptions to patch afterwards. Velocity data cannot see that a product needs wrapping, a heavy case needs a waist-height slot, or a fragile line can't sit under the drop zone. Your floor knows these things. Ask before you move stock, not after.
The numbers after 12 months
The operation now holds around 100,000 lines a day as a sustained capability with 23 or 24 pickers. The five heads that used to be locked into picking stations are capacity the operation can point at everything else a warehouse needs doing.
Two things to do this week
- Count the steps to your top 20 SKUs. Not the whole warehouse. Pull your pick frequency data, take the top 20 lines, and physically count the steps from the picking station to each one. If any of your top movers costs more steps than something you pick weekly, you have found free productivity. A spreadsheet and an hour on the floor is the entire toolkit.
- Check what share of lines your automation actually carries. Then check whether its fastest channels run empty during peak. If your fillers can't keep up with a channel, the machine's headline speed is fiction. Rebalancing channels is usually cheaper than any other capacity you can buy.
Neither needs budget or sign-off. The data for both is almost certainly already in your WMS, unread.
That was the whole trick here. Nothing in this article required new software or outside help. It required someone to open the reports the operation was already producing, count some steps, and move 800 locations one shift-end at a time.