How to Do Warehouse Slotting Optimization (Method + Worked Example)

Table of contents

• The five inputs you need before you move a single carton
• Step 1: rank every SKU by pick frequency, not by sales value
• Step 2: place the A-movers in the golden zone first
• Step 3: adjust for cube and replenishment, not just frequency
• Step 4: place items that ship together, together
• Worked example: cutting one pick path from 214 m to 96 m
• Step 5: decide how often to reslot
• Keep placement keyed to live pick frequency, not to a spreadsheet
• When slotting optimisation is not the bottleneck
• FAQ

The five inputs you need before you move a single carton

Slotting done from intuition reslots the same SKUs back and forth every quarter. Slotting done from data moves once and holds. You need five things, and every one of them is already in your order history:

• Pick frequency per SKU — how many order lines touched that SKU over a representative period (say 12 weeks). Lines, not units. A SKU picked on 400 orders at one unit each is walked to 400 times; a SKU picked on 3 orders at 500 units each is walked to 3 times. Slotting optimises walking, so it counts visits.
• Units per pick — the average quantity taken per visit, which decides whether a location needs a carton, a shelf or a pallet.
• Cube — the physical size of one selling unit, so a fast, bulky SKU does not get put somewhere that forces a replen every two hours.
• Affinity — which SKUs appear on the same order together, so items that ship as a set sit near each other.
• Current location — where each SKU lives now, so you can measure the move rather than guess at it.

If pulling those five for every SKU is a manual export-and-pivot job that takes a day, that is your first finding: the data exists but is not usable on demand, which is exactly why the slotting decays. Hold that thought — it is the difference between a one-off reshuffle and a layout that stays optimised.

Step 1: rank every SKU by pick frequency, not by sales value

The single most common slotting mistake is ranking by revenue or by units sold. Neither is what the picker experiences. A £2 fast-moving consumable picked on half your orders generates more walking than a £900 item picked twice a month, and it is the walking you are paying for.

So rank by lines picked over your representative window, highest first, and split into velocity bands. The bands mirror an ABC cut, and the full ranking method — the 80/20 logic, where to put the cut lines, how to handle the long tail — is worked through in ABC analysis for inventory. For slotting the practical split is:

• A — the fast face. The top SKUs that account for roughly the first ~70–80% of pick lines. Usually a small fraction of the catalogue.
• B — the middles. The next band, picked regularly but not constantly.
• C — the tail. Everything picked rarely. Often the majority of the SKU count and a small share of the walking.

Do this per pick zone if you have distinct areas (ambient vs chilled, small parts vs pallets), because a SKU that is an A-mover in the small-parts zone is not competing for space with pallet lines.

The output of step 1 is a single sorted list. That list, and nothing about anyone’s opinion of which products are “important”, drives every placement that follows.

Step 2: place the A-movers in the golden zone first

The golden zone (also called the strike zone or power zone) is the band of a pick location between roughly mid-thigh and shoulder height — the space a picker reaches without bending to the floor or stretching overhead. Every bend to a bottom shelf and every reach to a top shelf is time and, over a shift, injury risk. You want your most-visited SKUs in that band and your least-visited SKUs taking the awkward positions, because a slow mover is reached for rarely enough that a bend now and then costs little.

Placement runs on two axes at once:

• Vertical — height within the bay. A-movers at waist-to-shoulder. B-movers above and below that. C-movers on the top shelf and the floor level.
• Horizontal — distance from despatch. A-movers in the bays closest to the pack bench and the start of the pick path. C-movers in the far aisles.

The two axes interact. A prime location is one that is both in the golden height band and close to despatch; there are only so many of them, and they go to the highest-frequency SKUs in strict rank order off your step-1 list. Work down the list, assign the best remaining location to the next SKU, and stop arguing about it — the ranking already made the decision.

One guardrail: do not cluster all your A-movers into one tight block. If your top ten SKUs all sit in the same two bays, you create a congestion point where pickers queue behind each other at peak. Spread the A-movers across the near end of each aisle so several pickers can work fast SKUs at once.

Step 3: adjust for cube and replenishment, not just frequency

Frequency decides priority; cube decides feasibility. A fast SKU that is also bulky will empty its pick location constantly, and every empty pick face triggers a replenishment — a second labour cost that slotting is supposed to reduce, not create. So after the frequency-ranked placement, pass back through the A-movers and check each one’s cube against its location size:

• A fast, bulky SKU needs a larger pick location (a full case flow lane or a pallet position in the golden zone), sized so it holds enough to get through a peak day without a mid-shift replen.
• A fast, tiny SKU (screws, sachets) can sit in a small bin, and you can fit several A-movers into the frontage one bulky SKU would have taken — buying back prime real estate.

The number that governs this is how long a pick face lasts: units in the location ÷ units picked per day = days of cover. Aim for a pick face that survives at least one full pick day for A-movers, so replenishment happens on a predictable off-peak schedule rather than interrupting picking. If a golden-zone location cannot physically hold a day of a bulky A-mover, that SKU earns a bigger location even at the cost of prime frontage, because the replen labour it otherwise generates outweighs the walking you saved.

Step 4: place items that ship together, together

Affinity is the second-order gain and the one most reshuffles miss. If two SKUs appear on the same order most of the time — a product and its consumable, a base unit and its most-common accessory — putting them in adjacent locations means one stop instead of two. Across thousands of orders that compounds.

You find affinity in the same order history: for each pair of SKUs, count the orders that contain both. The pairs with high co-occurrence are candidates to slot side by side, provided doing so does not drag a fast mover out of its golden-zone rank. Affinity is a tie-breaker and a fine adjustment, not an override — you never demote an A-mover to sit next to a C-mover it happens to ship with. Where it earns its keep is among SKUs of similar velocity: two B-movers that ship together should be neighbours.

For a wholesale operation the strongest affinity signal is often the standing order — the same customer taking the same basket every week. If 40% of your volume is repeat baskets, slotting for those baskets specifically shortens the path on the orders you can predict. Keeping those baskets and their pick paths visible is squarely a job for the warehouse management software sitting over the pick face, which we come back to at the end.

Worked example: cutting one pick path from 214 m to 96 m

Take one real order profile through the method. The distributor’s most common order type is a 12-line trade order. Before slotting, the SKUs on that order are scattered by the accident of when each was first put away — new lines went wherever there was an empty slot.

Before — the path the picker actually walks:

Measured with a trundle wheel, that path runs 214 metres, with the picker doubling back to aisle 4 twice and bending to floor level for two of the most-picked lines on the order.

Now apply the method. Rank the 12 SKUs by pick frequency across all orders (step 1). Nine of them turn out to be A-movers — this order type is common precisely because it is made of the fast SKUs. Slot those nine into golden-zone locations in the two aisles nearest despatch (steps 2–3), spread across both aisles’ near ends to avoid congestion, with the two bulky lines given case-flow lanes. The three tail SKUs on the order move to the far aisle, where they are reached for rarely enough that the distance costs little.

After — the same order, reslotted:

The path is now a near-straight run down two adjacent aisles and back to the bench: 96 metres, with no doubling back and every fast line at reach height instead of on the floor. The walking on this order type falls by more than half, and the two floor-level bends on the busiest lines are gone — the ergonomic win that does not show up in a distance measurement but shows up in the injury log.

The honest caveat: you optimised for this order profile. A rare 30-line order that reaches into the tail will walk further than before, because you deliberately pushed the tail to the far aisle. That is the correct trade — you shortened the many common orders at the cost of the few rare ones. Slotting is optimisation against your actual demand mix, not against every conceivable order.

Step 5: decide how often to reslot

A layout slotted perfectly today is wrong within months, because demand moves. A seasonal SKU that was a C-mover in January is an A-mover in November; a discontinued line is holding a golden-zone slot it no longer earns. So reslotting is a cadence, not an event. Three layers:

• Continuous, small — the drift correction. A handful of SKUs whose velocity has clearly shifted, moved as you spot them, ideally flagged automatically. This is cheap and stops the layout decaying between the bigger passes.
• Periodic, medium — the seasonal reslot. Ahead of a known peak, pull the last period’s pick frequency, re-rank, and move the SKUs that changed band. Time it before the peak, not during it.
• Occasional, large — the full reslot. A ground-up re-optimisation, warranted after a range overhaul, a layout change, or when the small corrections stop keeping up. Disruptive; do it rarely and deliberately.

The trap is doing only the large one, once a year or “when it gets bad”. By the time it is bad enough to notice, you have paid for months of excess walking. The cheaper regime is frequent small corrections plus a seasonal pass, affordable only if re-ranking is quick — again, a data-availability problem, not a warehouse one.

Two operational notes. First, never move a SKU without moving its stock record in the same transaction — an optimised layout with a wrong location file is worse than a bad layout with a right one, because pickers walk to empty slots and short orders. Reslotting and location accuracy are the same discipline; the count that confirms a moved SKU is where the file says it is belongs to your inventory cycle count routine, which is how a reslot stays trustworthy. Second, stage moves to avoid picking against a slot mid-move — reslot in off-peak windows and update the system as each SKU lands, not in a batch afterwards.

Keep placement keyed to live pick frequency, not to a spreadsheet

Here is where most slotting projects quietly fail. The reslot is run as a project: someone exports six months of order lines, builds a slotting spreadsheet over a fortnight, the warehouse gets rearranged over a weekend. Then the spreadsheet is never opened again, demand drifts, new SKUs get put away wherever there is a gap, and eighteen months later the aisles are as scattered as before. The project delivered a snapshot, not a system.

The alternative is to keep the slotting logic inside the system you already run the warehouse from, keyed to live pick frequency, so the layout is re-scored continuously against real demand rather than against a stale export. Concretely, an owned operations system that already holds your orders and stock can:

• Rank SKUs by pick frequency on demand, from live order lines, so a reslot is a report you run in minutes rather than a fortnight’s export-and-pivot.
• Flag drift automatically — a C-mover that has become an A-mover, an A-mover slot held by a dead SKU — so continuous correction is a worklist, not a discovery exercise.
• Guide put-away for new SKUs into a location that fits their expected velocity band, so the layout stops degrading between reslots at the source.
• Enforce the location-file update as part of the move transaction, so an optimised layout and an accurate stock record never diverge.

None of that requires a standalone enterprise slotting engine. Dedicated slotting and warehouse software exists and earns its price at real scale — that whole category, and when the maths of buying it works out, is covered in warehouse slotting software. But for a business too messy for spreadsheets and not ready for a full warehouse-management platform, the pragmatic answer is a right-sized operations system that already knows your orders and stock, doing the ranking and the drift-flagging as a native function. The slotting stays optimised not because someone remembers to reshuffle, but because the placement logic lives next to the demand data that should be driving it.

When slotting optimisation is not the bottleneck

Slotting shortens the walk between picks. If your lost time is somewhere else, reslotting polishes the wrong surface. Before committing to a full reslot, rule out:

• The location file is wrong. If pickers routinely walk to slots that are empty or hold the wrong SKU, fix location accuracy first — no slotting helps when the map is lying.
• Replenishment is the constraint. If pickers wait at full-but-not-replenished faces, the problem is replen scheduling, and better slotting of a starved face changes nothing.
• The pick method is the ceiling. Single-order picking one order at a time up and down the aisles has a floor that no slotting beats; batch or zone picking may be the larger lever, and slotting then optimises within the better method.
• The catalogue is the problem. If a third of your SKUs were picked zero times in the last quarter, the question is whether they should be stocked at all, not where to put them.

The honest test: measure walking time as a share of total pick time. If pickers spend most of a pick cycle walking, slotting is your lever. If they spend it waiting, searching or handling, fix that first — then slot.

FAQ

What data do I need to start warehouse slotting optimization?

Five fields per SKU, all already in your order history: pick frequency (order lines that touched it over a representative period), average units per pick, physical cube of one unit, affinity (which SKUs share orders), and current location. If pulling those on demand is a day’s export-and-pivot rather than a quick report, that difficulty is itself the finding — it is why the layout decays between reslots.

Should I rank SKUs for slotting by sales value or by pick frequency?

By pick frequency — order lines, not units or revenue. Slotting reduces walking, and walking is driven by how often a picker visits a location, not by what the item costs. A cheap consumable picked on half your orders generates far more walking than an expensive item picked twice a month, and it is the walking you are paying to remove.

What is the golden zone in slotting?

The band of a pick location between roughly mid-thigh and shoulder height, reachable without bending to the floor or stretching overhead. Your most-frequently-picked SKUs go there, and your rarely-picked SKUs take the top shelf and floor level, because a slow mover is reached for seldom enough that the awkward position costs little.

How often should I reslot the warehouse?

On three cadences at once: continuous small corrections for individual SKUs whose velocity has shifted, a periodic reslot ahead of each known peak, and an occasional full re-optimisation after a range or layout overhaul. Doing only the annual full reslot means paying for months of excess walking before anyone notices — frequent small corrections plus a seasonal pass is cheaper, provided re-ranking is quick.

Do I need dedicated slotting software?

Not at small or mid scale. Dedicated slotting engines earn their price at real volume, but for most growing distributors the practical answer is to keep the ranking and drift-flagging inside the operations system that already holds your orders and stock, so a reslot is a report rather than a project. The dedicated-software category, and when its maths works out, is a separate question from the method itself.