The Role of Analytics in Shaping the Future of Warehouse Operations Dr Tony McVeigh
MORE WITH LESS ! 2
ORDER PICKING 3
Order Picking ▪ Half of the order picker’s time is spent on traveling * ▪ The retrieval of items from storage in response to customer demand ▪ Minimizing traveling cost (distance or time) is one ▪ Picking is the most labour-intensive and costly of the most focussed objectives to improve order process that typically comprises 55% of the total picking efficiency warehouse operational expenses* * Tompkins et al (2010) 4
Warehousing and Order Fulfilment ▪ Since order picking is acknowledged as the most labour and capital-intensive activities in warehouse operations, this drives a strong emphasis on picking optimisation; in particular: 1. Storage Allocation Regulates the location of items (to minimise handling costs) 2. Routing Policies Defines the best possible batch and route (for optimal picking) ▪ Order picking research* concentration is 32% and 38% on storage and routing, respectively ▪ With dynamic fluctuations in demand, warehouse picking can be a bottleneck in the order fulfilment cycle, which is an essential service level KPI ▪ The picking strategy should optimise the service level by compressing the order fulfilment cycle * McGinnis et al (2011) 5
ORDER PICKING SYSTEMS 6
Order Picking Interrelationships • ABC Analysis • Slotting • WMS • Product Characteristics • Voice Picking • Demand Profiling • RF Picking • Location System Technology Layout • Pick-to-Light • Space • Put-to-Light • Routing Policies • AS-RS • Handling Units − Totes, Bins, Cartons • Goods-to-Picker − Pallets • Picker-to-Goods − Cages • Walking Distances • MHE Labour Equipment • Batch or Zone • Racking • Training • Carousels • Ergonomics • Conveyors • Health & Safety • Automation 7
Warehouse Design and Simulation 8
Classification of Order Picking System * Who picks the goods? Who / what moves in the pick zone? Conveyance to connect pick zones? Picking policy OPS Automation Level Automation level * Dallari, Marchet and Melacini (2007)
Selection Method: OPS Matrix * Picking Volume Pick-to-Tote Order Lines per Day Sortation Part-to-Picker Picker-to-Part Number of Items ▪ Order sizes less than 0.5m³ * Dallari, Marchet and Melacini (2007)
PICK ROUTES 11
Basic Routing Policies Mid-Point Return S-Shape Optimal Largest Gap Combined 12
Comparison of Routing Methods Most Popular PerformanSC Case Study * ▪ Narrow-aisle configuration 14% travel reduction ▪ Low level picker-to-parts system to pick 11 items ▪ Order size range: 1 to 27 items (average,19) ▪ Pick tours follow S-Shaped routes ▪ Current pick routes approach the optimal solution for a large number of items ▪ For a small number of picks, the current practices are sub-optimal by 25% ▪ For the most frequently picked orders , pick distances may be reduced by 14% solely by re-routing Not for general distribution 13
PRODUCT SLOTTING 14
Slotting ▪ In general, less than 15% of SKUs within a warehouse are assigned to the most efficient location and this results in 10% to 30% cost increase in travel time and under-utilised locations* ▪ Slotting is a procedure used to optimally place individual goods within the warehouse ▪ This process a) Increases the storage density - squeeze more product into the space available b) Reduces picking distance - locate the most popular items close to despatch - place items that frequently ship together next to each other in the pick face c) Enables ergonomic efficiencies - by placing heavy items at waist height d) Accounts for key parameters such as value, cube, weight and crushability e) Determines how many and what size of pick face is required for each product line (e.g. very fast movers will require multiple pick faces to avoid congestion and bottle-necks at a single location) ▪ Drives improvements in space utilisation with possible footprint savings between 35% and 43% ▪ Accounts for seasonality and suggests product transfers * Frazelle (2010) 15
Simple Slotting Technique ▪ Slotting software is integrated within many of today’s WMSs but can also be sourced separately ▪ Alternatively, simple slotting can be accomplished by order profiling using spreadsheet models ▪ Orders may be analysed in a number of different ways - the following grid presents one possible scheme Frequency High Moderate Low Aa Ab Ac High Value Ba Bb Bc Moderate Ca Cb Cc Low ▪ Aa products are those SKUs that generate the most value and are sold most frequently, while Cc products, on the other hand, are considered to be slow-movers; selling the least and moving less frequently ▪ Each SKU is assigned one of nine possible classifications ▪ Items having high sales and pick frequencies may be located in close proximity to despatch ( Golden Zone ) ▪ The data can then be used to generate a warehouse heat map in which with colour codes are used to depict the positions and classifications of each SKU 16
Slotting Case Study ▪ Warehouse footprint: 200m x 55m (2 mins to briskly walk from wall-to-wall) ▪ Order profiling analyses performed Walking Speed [ms -1 ] Walking Speed Level ▪ Mean Std Dev The top 50 fast movers identified and relocated closer to despatch ( Golden Zone ) Slow 0.65 0.21 Normal 1.11 0.17 ▪ Fast 1.47 0.2 Consequently, aggregate pick travel reduced by 17km per shift Faster 1.83 0.25 Running 2.7 0.46 ▪ This process is deployed across all SKUs and re-slotting occurs each quarter Mobile Gold Standard (2011) ▪ The order picking cycle time reduced by 20% resulting in improved warehouse throughput and material handling savings Not for general distribution 17
INVENTORY REPLENISHMENT 18
Conventionally normal! 19
Demand Partitions Smooth ▪ Demand is largely static; does not vary significantly over time ▪ Call-off remains relatively close to the average values ▪ Mainstream products High Erratic ▪ Sales may be substantial ▪ Volumes may vary dramatically ▪ Low-value items usually fulfilled immediately ex-stock Intermittent Variability in ▪ Sporadic demand demand quantity ▪ Products approaching end-of-life ▪ May account for half of stocked goods Lumpy ▪ Slow-moving / expensive SKUs ▪ High value maintenance components ▪ Sold on occasional basis Low ▪ Safety stock requirements ▪ SKU cost / lead-time according to service level criticality Variability in demand timing Low High * Syntetos, Boylan, Croston (2005)
Demand Distributions CV² 0.49 p 0 1.32
Inventory Replenishment Models CV² Gamma Negative Binomial Continuous Discrete 0.49 Functions Functions Normal Poisson p 0 1.32
Demand Distribution: Retail Sector Case Study 1 ▪ 12 week timescale: 2.8 million units across 17 thousand SKUs 30% 33% 37% ▪ There are 3 essential SKU classes: Lumpy , Erratic and Smooth ▪ However, volume-wise, most items are found to be Smooth and the balance, Erratic ▪ Stock management is improved by the adoption of 2 replenishment models; Normal and Poisson ▪ The impact on service and working capital is both positive and measurable ▪ Storage requirements (pallet spaces) reduced by 8 – 11% Not for general distribution 23
Demand Distribution: Hardware Sector Case Study 2 ▪ Typically, inventory replenishment was driven by Normal rules ▪ Demand partitioning provided insights into SKU classifications ▪ Introducing a more appropriate replenishment model to Lumpy SKUs enabled a reduction in component stock levels ▪ For a 98% service protection target, the annual impact is a) 14% less stock in the system, and b) 9% less material value Normal NBD Not for general distribution 24
SUMMARY 25
In Conclusion ▪ The application of analytics is key for the advancement of warehouse operations performance ▪ Three (of many possible) paths to performance improvements were presented - − Order Picking − Item Slotting − Inventory Replenishment - each requiring demand and master data and each alone, capable of delivering rapid economic benefits ▪ Logistics and warehouse operations needs analytics and, in turn, the needs that arise in these activities drives new problems that analyses will help solve ▪ In reality, the implementation of meaningful solutions for warehouse and supply chain processes requires a) the use of technology to-hand for the management of data and knowledge extraction and, b) the resolution of optimisation models to assist decision making and so, making it possible to achieve …. More with less ! 26
Thank you for your attention
Recommend
More recommend
