Slotting Optimization
Slotting is the process of determining where items should be placed in the picking area so that the popular items are stored close together and in close proximity to the starting point of picking so as to minimize walking time. Slotting is often painful since it requires substantial labor to relocate items. Furthermore, picking often must stop while items are being relocated.
A naïve slotting approach would be to simply sort the items by popularity and place them on contiguous picking bays close to the start of picking. The problem with this approach is that the popular zone of the picking area may get congested and pickers will be tripping over each other. A better approach would be to “stripe” the popular items on shelves located at ergonomically beneficial heights. This approach would lengthen the popular picking area but would also reduce congestion, improve picking speed and reduce worker fatigue.
Another important consideration in slotting relates to item size, weight and durability. Fragile items should be picked last and placed on top of durable, heavier items already picked. Similarly, better packing density is generally achieved when larger items are picked before smaller items that can be packed around the larger items. A good slotting algorithm will take all of these factors into consideration when assigning slots to products.
Box Selection Optimization
Picking an optimal (fewest) set of boxes from a set of available boxes, and then picking, packing and shipping an order are challenging procedures and greatly influence shipping costs. Before a solution can be derived, the cube (dimensions) of the picked items must be known. Devices that automatically measure the weight and X, Y and Z lengths of a rectangular prism that will exactly contain an item can be purchased from several sources. Using this data can help determine the smallest set of boxes needed to minimize wasted space in those boxes.
A good algorithm to determine this set of boxes will include these parameters:
- Maximum fill level of each box to allow for insertion of padding materials.
- Maximum weight constraint on each box.
-
Placement order of picked item:
- Best Fit – place largest items first in box and place smaller items around the larger items
- Pick Sequence – place items in box in the order picked, regardless of size.
- Orientation of picked items in the box to comply with “this side up” specifications.
- Nestability: items can “nest” one inside the other (like tapered trash cans). Maximum nesting level controls how many items can be nested together.
- Containability: items have hollow insides that can contain other items.
- Inclusion (or exclusion) lists to control which box types must be used (or cannot be used) to pack specified item types. For example, temperature sensitive item types may require insulated box types.
Some algorithms simply ensure the available volume in the box is large enough to contain the volume of the items to be packed (water fill placement). Using these parameters, the good algorithm will computationally “pack” (3-D placement) each item, one-at-a-time, in candidate boxes to ensure each item will fit in the remaining space.
Batch Picking Optimization
Batch picking is the process of picking multiple order boxes with a single trip through the warehouse – usually using a cart. As a simple example, suppose a wave of 400 orders is to be picked with carts, each with a capacity of 10 boxes. Picking this wave will require 40 cart-trips. The question is “which 10 boxes should go on each cart to minimize the workload, subject to a set of constraints like total order weight?”
Since walking distance usually dominates total pick time, the optimization problem is how to allocate the 400 orders across the 40 carts so as to minimize the total walking distance to pick all the orders. An exhaustive evaluation technique would require the walking distance calculation on all combinations of 400 order taken 10 at a time, or 400!/[(400-10)!*(10!)] = 2.58x1019. Assuming a computer capable of evaluating (calculating the walking distance for) each combination at a rate of one billion evaluations per second (using a super computer) this calculation would take about 817 years! Such an exhaustive technique is clearly not feasible.
A “greedy” algorithm (heuristic) would simply put the best 10 boxes in close proximity on the 1st cart, put the next 10 boxes in close proximity on the 2nd cart, etc. Although this might produce good walking distances for the first few carts, the last many carts might have terrible walking distances.
A better heuristic would minimize TOTAL walking time for ALL carts.
Biologically inspired heuristics, such as genetic algorithms, can produce “near optimal” solutions in only a few seconds using a PC. The advantage of genetic algorithms is that they always have a “best so far” result during computations so the longer the heuristic runs, the better the results.
Other variables sometimes used in batch selection optimization are maximizing cart shelf space utilization (when multiple box sizes are used), minimizing stops (to maximum picking at each bay), maximizing cluster picking (where one product in a bay is picked and distributed to several order boxes or several products in a single bay are picked and placed in a single order box).
Summary
Although working faster may help to improve DC performance and reduce costs, the use of optimization methods in advance to plan for and organize the picking process can yield far greater benefits.
Making the right decision for your warehouse starts by asking the right questions, such as:
· Is your business growth being stunted by limitations to order fulfillment operations?
· Are most of your overhead costs related to order picking labor?
· Are you wasting precious time having your order pickers travel most of your warehouse to pick orders?
· Do you need to increase picking productivity to provide better customer service?
· Would accurate order picking improve your customer service?
· Would you gain a competitive advantage by increasing pick speeds, having error-free picking and reducing labor costs?
Dr. Peck is a Professor Emeritus in the Computer Science Department at Clemson University. He retired from Clemson as Chairman of the Department. At Clemson he served in a variety of capacities including Director of the Division of Information Systems Development, an organization he founded to assist South Carolina State Government in the design and development of very large database/data communication systems such as Medicaid and Online Public Healh Medical Records. In addition to authoring numerous scholarly research papers and serving as a principal or co-principal investigator on several Department of Defense and National Science Foundation contracts, Dr. Peck served as an accreditation visitor on several teams for the Southern Association of Colleges and Schools and worked for many years as a consultant in computer applications for manufacturing and distribution through Foxfire Technologies Corporation, a software company he co-founded in 1987 and sold in 2007. Dr. Peck currently serves as President and CEO of FastFetch Corporation, a company he co-founded in 2006.
