Case picking of customer orders, one of the more labor-intensive activities in the supply chain, is also the hardest to automate. So in 1993 we introduced innovative technology in another way - by improving the information handling and decision-making. We developed and implemented AutoPalletP3, which is artificial intelligence case-pick planning software. This system has changed the way warehouses pick customer orders, offering major benefits both in productivity and quality, which save millions annually.

Warehouse managers are faced with a myriad of challenges that include:

• Case picking is labor intensive, not very productive, and hard to automate
• Damage caused by material handling, according to a GMA study, runs at 0.4% of sales ($1.4 billion in the grocery industry)
• Labor turnover rates exceed 40%. The warehouse industry considers that training and experience are normally necessary to achieve quality case picking
• The high cost of transportation requires that trailers be cubed out – making tall and well stacked pallets a necessity
• Yet, case picking tall pallets is a significant source of worker’s compensation claims and slows productivity

AutoPalletP3 is now considered indispensable by its users. In seven years there has been virtually no downtime of any kind. During this period, picker productivity has doubled while customer claims for damage and pick errors have dropped by 75%. AutoPalletP3’s proven capabilities to fit more product in a truck have permitted operational changes including re-definition of a ‘full truckload’ in the business. Freight costs are down a massive 10%.

This is not an isolated example. AutoPalletP3 runs in sites across the world creating damage-free pallets from more than a billion cases. In addition to operating in plant distribution operations, it is also working well in two different third-party logistics provider's operations.

A case picker holds the AutoPalletP3 pick
lists while confirming the pick in his terminal.

Without AutoPalletP3: Each order picker is presented with a long multi-page list that is one complete customer order. This may be up to a full truckload made up of 4,000 cases of 4,000 different products, to be selected and stacked as mixed loads in 26 pallet spots on the truck. There are many rules to observe; "do this; don't do that". The goal is perfect quality - no selection or count errors, stable loads, no product damage - but the cases must be picked as fast as possible. It will take 10 - 14 hours to complete this order - up to 20 tons of product to be manually picked up and set down -, and it MUST all fit in the truck. And many other pickers are working the area at the same time, each on a different order.

The order picking process requires continuous
decision-making against conflicting priorities.

Although Quality should be the primary goal of order picking, it is Productivity that will initially provide the primary incentive to implement improvements, as productivity is dollar-measurable, while the other benefits become more apparent after installation. To understand the productivity opportunity at the start of the development, one of our clients contracted with a University Industrial Engineering team to conduct a week-long Work Sampling Study of order pickers, with this interesting conclusion: a quarter of the pickers’ time was being spent planning and trying to make decisions for which they had insufficient data.

With AutoPalletP3, most of the planning time (28%)
becomes available for more case picking.

AutoPalletP3 has the data to make better decisions, and gets it right the first time, which is always the lowest cost way. And it is VERY fast. 20% of the picker’s time is freed up to be plowed back into picking more cases - a productivity gain AS WELL AS providing a higher quality result. And since minimizing travel distance is always a productivity priority, AutoPalletP3 accesses a map of the pick area and computes a minimum total travel distance for picking the order.

With AutoPalletP3, audits are usually unnecessary, so the picker can stay on the job. If audits are required, they can be faster, as the position of every product on each pallet is known from the picklist.

AutoPalletP3 also looks at multiple orders on the same truck, adjusting the plan to make sure that the orders, taken together, will not overflow the truck. If calculated total case cube is within 95% of available truck cube, AutoPalletP3 automatically invokes additional criteria to achieve a tighter fit.

AutoPalletP3 designs stable loads either as one tall unit or, if desired, two units to reduce the height pickers need to "throw" cases in pairs. In either case the product can be built to optimum vehicle height. AutoPalletP3 does not allow overhangs, and creates mixed layers from cases of equal height. The AutoPalletP3 system also assigns a new parameter, "Crush Resistance Index" to every product; it then designs loads in which a risk of crushing simply DOES NOT OCCUR

AutoPalletP3 designs mixed loads that do NOT overhang the pallet unless the original product unit load is designed to overhang. When absence of overhang can be relied on, the truck can be re-planned with 6 extra pallet spaces - an 8% increase in truck utilization.

AutoPalletP3 also designs all mixed loads to an optimum target height set by the user, subject to a not-to-exceed height. This is possible because it works from a database containing the dimensions of every case. When average load height is increased, for example from 80" to 95" by AutoPalletP3’s sophisticated planning, the truck-cube utilization increases by 9%, potentially worth millions each year to cube-constrained shippers.

AutoPalletP3 improves every aspect of these customer needs. AutoPalletP3 creates family groups, separates any products which might contaminate each other, and since it carries a customer-specific database, it can even customize deliveries differently for different customers. Each load is shipped with a contents list matching the AutoPalletP3 picklist, showing not only which products are on each pallet but also WHERE the cases are on the pallet. Check-in is faster and more accurate.

Perhaps the greatest endorsement for AutoPalletP3 was received from a warehouse manager who said, "we can’t live without AutoPalletP3."

AutoPalletP3 analyzes each customer order and considers many ways in which it could be assembled, then chooses the overall best. It does this by consulting a variety of databases, including product type and dimensions, product location, customer preferences, and a 26-point set of rules. It pays particular attention to the route through the warehouse, calculating distances to minimize the total travel distance.

After AutoPalletP3 has generated an initial design for all the loads in an order, it moves into an optimizing routine. Using a penalty scoring system for every less-than perfect element of the solution (no solution of this complexity is ever perfect!), it examines many alternative ways to assemble the whole order, trading solutions in search of the lowest penalty score. AutoPalletP3 will run an order up to 500 times in 90 seconds until it achieves the best possible result. For any picker, even the best, to improve on the AutoPalletP3 result is the equivalent of beating a super-computer at chess - in two minutes or less!

This diagram illustrates how AutoPalletP3
automates case picking of custom orders