Inside an e-commerce warehouse, two coupled decisions: which SKU is stored on which shelf (slotting), and along which aisle sequence and route a picker walks for a given pick list (picking). The classical warehouse OR problem of minimising total walking distance.
In plain words
Sound familiar?
- We stock 20K-150K SKUs; we pick 5K-30K order lines a day. Pickers walk 15-25 km/day; fatigue complaints rise and the error rate climbs at end-of-shift.
- When new SKUs arrive the warehouse lead says 'put it on any empty slot'; pick velocity, co-order pattern and ergonomics are not considered. Fast-mover SKUs can land on top shelves, slow movers at eye level.
- A pick list has 8-25 lines; the picker eyeballs the list and walks 'roughly in this order' โ some end up crossing the same aisle twice. A single order pick varies between 200 and 800 m of walking.
- In seasonal peaks (year-end, sales periods) order line volume jumps 3-5x; the current slotting jams up because fast movers sit far from the door, creating a bottleneck.
- Order lines per picker per hour is 20-35% below the sector benchmark; we do not know where to start โ slotting, routing or batching first?
- Search time inside pick-pack-ship is eating 20-30% of total pick time; SKU locations keep changing, knowledge is lost between shifts.
- Our warehouse management system (WMS) produces a pick list but does not solve the route โ it just sorts SKUs by aisle number. No batch picking, no correlated slotting.
Why it matters
How it's solved
Technical depth
How it's solved
Technical depthIn one sentence: First, place SKUs on shelves smartly (fast movers near the door, heavy items at waist height, co-ordered pairs side by side), then solve the pick route with a math rule rather than “in aisle order” (S-shape, return, midpoint, etc.). Thinking about the two decisions together wipes out 25-40% of the walking.
Operations Research (the discipline that uses math and computers to solve business decisions) treats this as two coupled decisions: Slot Assignment (Slotting) is SKU-to-shelf mapping, Order Picking Routing is the picker route for a given order. Solution in three stages:
1. Modeling. Inputs: SKU list (weekly pick frequency, cube, weight, ergonomic class, seasonality, co-order pairs โ which SKU pairs come together often), shelf layout (aisle count, aisle length, shelf height levels, door-to-aisle distances), order history (at least 6-12 months of line-level data), workforce (shifts, walking speed, carry capacity โ cart, low pallet, basket). Constraints: shelf capacity (volume and weight), ergonomic zone (heavy SKU at 60-120 cm waist height), cold-chain / dry segregation, hazardous-material segregation, rack-type compatibility (pallet rack vs flow rack vs case-flow), seasonal SKU surge buffer. Objective: minimum average pick walking distance, or minimum total walking minutes per picker, or maximum lines per hour.
2. Solver-driven decision.
- For slotting: Heuristic ABC class-based assignment (sort SKUs into A/B/C by pick velocity, place A near the door) is a fast start. Correlated assignment (place often-co-ordered SKU pairs side by side) opens a 25-40% gap over the heuristic. For better results an SKU-shelf MIP (Mixed-Integer Linear Programming โ optimisation with some 0/1 variables and some continuous; assignment variables plus affinity weight plus ergonomic constraints) is solvable in small-mid warehouses (15K-50K SKUs); at large scale (100K+ SKUs) apply simulated annealing, tabu search (intelligent search methods โ produce near-optimal solutions) or clustered MIP.
- For routing: Single-order routing heuristics (simple rules for walking-route choice): S-shape (traverse each aisle end-to-end), return (enter, pick the last SKU, return), midpoint (leave from the middle), largest-gap (skip the largest gap), combined. For optimum, a TSP / Steiner-tree (Travelling Salesman Problem โ shortest closed tour) formulation is feasible on small SKU counts. Combined batch picking (multiple orders in one tour) uses savings algorithm, column generation or clustering plus TSP.
- Joint decision: Slotting and routing are not two separate problems โ the true optimum surfaces only when they are solved together. In practice a two-stage approach: slotting MIP first (long horizon, monthly or seasonal), routing next (real-time per pick list).
3. Field integration. Output is three-layered: (a) slotting plan โ SKU-shelf table loaded into the WMS, physical relocation done on night shift or weekend; (b) batch and route plan โ module called by the WMS at pick-list generation time; the picker’s hand-terminal shows an ordered SKU list with the aisle sequence; (c) performance dashboard โ lines/hour, walking distance, error rate, ergonomic violation counter. Monthly or seasonal re-slotting โ recompute when order pattern shifts (new SKU, campaign, season). Quarterly operations committee: lines/hour curve, most-walked aisle analysis, mispick rate, ergonomic load balance.
Alternatives
Manual assignment + WMS pick list
FreeZero extra license (existing WMS suffices)
Who it fits: Small warehouse (<20K SKUs, <3K lines/day), 4-15 pickers
- + No extra software cost
- + Warehouse lead's experience leads
- + Easy response to fast SKU changes
- โ Above 20K SKUs the human mind cannot keep up with assignment
- โ No correlated assignment, no batch picking
- โ Jams under seasonal surge
- โ Walking distance stays 30-50% above the optimum
Local warehouse management system (WMS) standard module
Enterprise150K-800K TRY license + 50K-200K TRY/year maintenance (TR market observation)
Who it fits: Mid-size warehouse (20K-80K SKUs, 3K-15K lines/day)
- + Close to TR regulation (e-invoice, customs, e-commerce marketplace connectors)
- + Standard pick-list and barcode flow
- + Turkish UI, local support
- โ Slotting optimisation module is usually heuristic (ABC class) โ no MIP
- โ Routing is typically 'sort by aisle' โ Petersen heuristics absent
- โ Batch picking and correlated assignment limited
International warehouse management system + optimisation module
Enterprise1-4M EUR license + 300K-1M EUR/year maintenance (large solution providers)
Who it fits: Large warehouse (80K+ SKUs, 15K+ lines/day), multi-site
- + Mature slotting MIP plus correlated-assignment module
- + Petersen routing heuristics plus batch picking
- + Automated seasonal re-slotting
- + Integration with conveyor and AS/RS automation
- โ High license cost plus long (10-20 month) implementation
- โ TR e-commerce integration is custom work
- โ Operations training programme is wide
Open-source solver + custom slotting/routing module
Open SourceZero license; in-house build 16-32 weeks or 600K-1.5M TRY consulting
Who it fits: E-commerce warehouse with a tech team, integration with an existing WMS
- + Zero license cost
- + Slotting MIP and routing TSP are well defined for open-source solvers
- + S-shape, return, midpoint, largest-gap and combined route heuristics available in open-source libraries
- + Re-slotting cadence is free; intervention easy
- โ Internal OR expert + WMS integration team required
- โ Academic prototype to field system takes 6-12 months
- โ Maintenance falls on the operator; 24/7 shifts demand a clear support plan
Recommendation
Ask in the meeting
- Which slot-assignment approach does the module support โ ABC class only, or also correlated assignment (placing co-ordered items side by side) and a formal optimisation model?
- Which single-order routing heuristics are supported โ S-shape (traverse the aisle end-to-end), return (enter, pick, return), midpoint, largest-gap (skip the largest gap), combined? Which conditions trigger which?
- Is batch picking (multiple orders in one tour) supported? Which clustering method and what typical batch size?
- How often is re-slotting run โ weekly, monthly, seasonal? Is the physical relocation scheduled automatically?
- How are ergonomic constraints (heavy SKU at waist height, top shelf for light only) and segregation rules (cold-chain, hazardous material) expressed in the model?
- Is seasonal SKU surge (campaign, sale period) supported via temporary buffer-zone allocation and dynamic re-slotting?
- In a pilot (8-12 weeks) on real operational data, how is the walking-distance and lines/hour saving versus the prior manual slotting reported?
- If the contract ends, in which standard format can we export SKU-shelf mapping history, order line data, pick performance data and routing algorithm parameters?
Technical details
Editor’s note
This problem is known in plain language as “warehouse order picking”, “shelf layout” or “slot plan”. In academic literature it has two coupled names: the SKU-shelf decision is Slot Assignment (also Slotting / Storage Assignment), the picker’s route decision is Order Picking Routing. If you walk into a WMS contract without separating these two names, you cannot test which decision layer the module you are buying actually optimises โ “pick-list generation” is one thing, “Petersen-heuristic route optimisation” is another, “correlated-assignment MIP” is a third.
This problem lives inside the warehouse โ picker walking route and SKU-shelf mapping. VRP (problem #002) is route between depots or depot-to-customer, at the truck and vehicle layer. Shelf assortment (#017) is which SKU appears to the consumer in a retail store โ store shelf-and-window plan. Facility location (#010) is where to site the warehouse in the city. #041 is which SKU appears to the picker first, how many steps it takes to fetch โ operational, hour-by-hour decision.
The most-skipped point in the field: Slotting cannot be solved alone โ it must be considered jointly with picking routing. Most heuristic slotting approaches (ABC class-based, sort by velocity) put fast-movers near the door; yet real-time pick routing optimisation (S-shape vs largest-gap vs combined heuristics compared) can beat that heuristic layout. Correlated assignment โ placing co-ordered SKUs side by side โ opens a 25-40% gap over plain ABC. Second-most-skipped point: ergonomic load. Heavy or oversized SKUs on top shelves drive injury and end-of-shift fatigue, raising the error rate; ergonomics must be a hard constraint in the MIP.
Step-by-step path โ for SMBs
Stage 1 โ Measure first, plan later. At least 6-12 months of line-level order data: lines per order, SKU list, picker name, walking time, total time. Per-SKU weekly pick frequency, cube, weight, ergonomic class. Co-order pair matrix (which SKUs are co-ordered often). Shelf inventory: aisle count, aisle length, height levels, door-to-aisle distances. Picker walking speed (empirical, average 0.8-1.2 m/s loaded).
Stage 2 โ Extract knowledge capital. Which SKUs are high-velocity (A class), co-order dense (correlated cluster), heavy/ergonomic (lower shelf), seasonal-surge sensitive. Plot the walking distribution of the current layout โ the 10 most-walked aisles and the 10 least-walked. Record error (mispick) rate and ergonomic violations.
Stage 3 โ Pilot. 8-12 weeks. In one zone (the aisle set with the highest line volume), apply a new slotting via MIP; flip the routing algorithm from S-shape to combined Petersen. Decision rights stay with the operations lead; the system advises. Success criteria set in advance: walking distance -15% minimum, lines/hour +10% minimum, mispick rate flat or better.
Stage 4 โ Roll-out. Within 9-15 months, full warehouse + batch picking + zone picking + re-slotting automation. Seasonal re-slot calendar (monthly check, quarterly full re-slot, pre-campaign special re-slot). Quarterly committee: walking distance curve, lines/hour, mispick rate, ergonomic violations, top-walked aisle analysis.
Risks โ what can go wrong
- SKU promotion surge. On a campaign day one SKU can take 10-30x its normal velocity; heuristic slotting will not bring it near the door, creating a jam. Dynamic re-slotting and temporary buffer-zone allocation are essential.
- Seasonal order disruption. The summer-winter mix shift breaks slotting โ a winter SKU on a summer-month far-aisle slot is wrong. Without 2-4 re-slottings a year, walking distance creeps up.
- Ergonomic violation and injury. Heavy SKUs on top shelves cause injury, compensation, productivity loss. Ergonomics must be a hard constraint; as a soft constraint it gets ignored.
- Single-supplier WMS lock-in. If the WMS contract lacks a ’export SKU-shelf mapping, line history, pick performance data in a standard format’ clause, leaving the system means losing the warehouse’s operational memory.
- Batch picking fatigue curve. Picking 8-12 orders in one tour raises the load per picker; if the fatigue curve is not in the MIP, the end-of-shift error rate climbs.
Technical view of solution methods
| Approach | Typical scale | Solve time | Optimum guarantee? |
|---|---|---|---|
| Heuristic (ABC class) | Small warehouse, <20K SKUs | instant | No, 50-65% optimum |
| Correlated assignment (Frazelle 2002) | Mid warehouse, 20K-80K SKUs | minutes | No, 70-80% optimum |
| Slot-assignment MIP (affinity-weighted) | Mid-large warehouse, โค80K SKUs | minutes-hours | Yes (within bound) |
| Simulated annealing / tabu (slotting) | Large warehouse, 100K+ SKUs | hours | No, good practical quality |
| Petersen routing heuristics (single-order) | Any scale | instant | No, 85-95% optimum |
| TSP / Steiner tree (single-order optimum) | Bounded SKU per order (โค25) | minutes | Yes |
| Batch picking MIP / column generation | Mid-large, 5K-30K orders/day | hours | Yes (within bound) |
| Joint slotting + routing MIP | Academic prototypes | hours-ongoing | Yes, limited scale |
Objective function choice:
- Objective 1 โ Minimum total walking distance: Classical formulation, most common.
- Objective 2 โ Maximum lines per hour: Throughput-focused, for peak season.
- Objective 3 โ Minimum picker fatigue load (weighted walking + ergonomics): Where ergonomics is priority.
- Objective 4 โ Minimum mispick rate (walking + search time): Quality-focused.
Multi-objective: weighted sum or hierarchical (ergonomics as hard constraint first, then walking, then batch size).
Academic references
Listed in the page frontmatter sources field.
Sources
- de Koster, R., Le-Duc, T. and Roodbergen, K. J. (2007). Design and control of warehouse order picking: A literature review. European Journal of Operational Research, 182(2), 481-501. Canonical survey of order picking OR.
- Petersen, C. G. (1997). An evaluation of order picking routeing policies. International Journal of Operations & Production Management, 17(11), 1098-1111. Comparison of S-shape, return, midpoint, largest-gap and combined heuristics.
- Roodbergen, K. J. and Vis, I. F. A. (2009). A survey of literature on automated storage and retrieval systems. European Journal of Operational Research, 194(2), 343-362. AS/RS and automated warehouse systems survey.
- Frazelle, E. H. (2002). World-Class Warehousing and Material Handling. McGraw-Hill. Practical reference for correlated assignment and slotting.
- Tompkins, J. A., White, J. A., Bozer, Y. A. and Tanchoco, J. M. A. (2010). Facilities Planning (4th ed.). Wiley. Foundational text on facilities and warehouse layout.
- YรK Thesis Centre โ keywords: ‘depo yรถnetimi’, ’toplama rotasฤฑ’, ‘raf yerleลimi’ or ‘WMS’ โ 30+ Turkish theses. tez.yok.gov.tr
Glossary
- Order Picking
- Warehouse operation of retrieving SKUs from storage locations to fulfil customer orders, typically performed by walking pickers.
- Slot Assignment
- OR problem of deciding which storage location (shelf, bin, slot) each SKU occupies in a warehouse, optimising expected walking distance, ergonomic load and order-affinity grouping.
- VRP
- The decision of which vehicles, leaving from one or more depots, visit which customers in which order.
- MIP
- An optimization model where some decision variables are forced to be whole numbers (e.g. number of trucks, number of shifts).
Related problems
Dozens of SKUs from the Same Supplier โ At What Frequency Do I Order Each So Trucks and Stock Cost Are Minimum Together?
This page is for an SMB wholesaler, importer or manufacturer pulling 50-500 SKUs from the same supplier โ typically across 3-15 main suppliers. The weekly question is the same: from this supplier, which products should ship today and which can wait until next week? If every product triggers its own order, the same supplier ends up sending three trucks and three customs filings a week; one truck, one customs filing and one setup cost can cover all of it if the groups are right. Manual planning falls apart past 50 SKUs: some weeks a half-empty truck, others three back-to-back orders โ fixed costs come back as 5-15% of unit product cost.
How Much of Each Product Should I Order So I Neither Run Out Nor Pile Up?
Written for retail, e-commerce, or manufacturing SMBs managing 500-5,000 SKUs. If your purchasing team still places orders every month from a spreadsheet using last-three-months average plus ten percent, and yet you see empty shelves and clogged warehouses at the same time, you are in this problem. When the same company carries both stockouts (5-15% lost sales) and dead inventory (10-25% capital tied up), intuition has hit its limit: seasonality, lead-time uncertainty, and slow-moving SKUs cannot be corrected by hand item-by-item. Systematic forecasting paired with an inventory policy typically frees 1.8-5.4M TRY/year of working capital on 50M TRY revenue and lifts the service level from roughly 85% to 95%.
Stock Is Not Clearing as the Season Ends โ When and How Deep Should I Mark Down So That Margin Holds and Dead Stock Stays Low?
This page is for you if you run a seasonal-collection fast-fashion retailer, a 30-100 store apparel/shoes/accessories chain, or a fresh-food chain (greengrocer, bakery, butcher). The classic pain: a 200-1,000 SKU collection sells for 4-12 weeks and then the unsold stock becomes a problem โ discount too early and you erode the gross margin, discount too late and you sit on a pile of dead stock at the end of the season. The decision: which item, when, and by what percentage to mark down? In a mid-size fashion chain run on gut-feel rules, 20-35% of end-of-season stock ends up as salvage; a proper markdown schedule lifts the gross margin by 5-15%, cuts dead stock by 20-30%, and adds 15-60 million TRY of operating margin a year on a typical revenue base.