[0001] This application claims priority to U.S. Provisional Application Ser. No. 60/263,317 filed on Jan. 22, 2001 which is incorporated herein by reference in its entirety.
[0002] This invention relates generally to supply chain management and, more particularly, relates to a method for managing inventory within an integrated supply chain
[0003] In the global economy of today, supply chains are commonly used to deliver goods reliably and at affordable prices. A supply chain typically involves the flow of material, information, and money between customers, suppliers, manufacturers, distributors and, possibly, financial institutions. Material flow includes, among other things, the physical product flows from suppliers to customers through the chain and reverse flows via product returns, servicing, recycling, and disposal. Information flow involves order transmission, order confirmation, and delivery status. Monetary or financial flow includes credit terms, payment schedules, payments, and consignment and title ownership arrangements. These flows cut across multiple functions and areas both within an organization and across organizations. In this regard, supply chains exist in service, retail, and manufacturing organizations, although the complexity of the supply chain may vary greatly from industry to industry and organization to organization. The coordination and integration of the material, information and financial flows within and across organizations is critical to effective supply chain management. Thus, in order for the supply chain to work for its intended purpose, all organizations involved in the supply chain must coordinate their activities with one another so that efficiency throughout the supply chain is achieved.
[0004] To coordinate activities within a supply chain, Manufacturing Resource Planning (“MRP”) and Enterprise Resource Planning (“ERP”) tools have been employed by organizations in an effort to gain control of the flows, plant operations, and to provide management with timely and useful reporting. Some companies use explicit supply chain management (“SCM”) and supply chain execution (“SCE”) systems which are often focused on a specific functional requirement. Generally, MRP systems have been used to translate the schedule for the production of products into time-phased net requirements for the sub-assemblies, components and raw materials, planning and procurement. ERP systems address the technology aspects of MRP such as client/server distributed architecture, RDBMS, object oriented programming etc.
[0005] In addition to MRP, ERP, SCM/SCE systems, a further supply chain management tool is described in U.S. Pat. No. 6,157,915 which relates to an active collaboration technology in an open architectural framework that is used to deliver information and decision support tools to various organizations. In the framework described in the '915 patent, the people across the organizations collaborate through domain task and specific active documents. The active documents contain both the necessary business information and decision support tools. Dynamic decision making is thus made possible through the delivery of active documents to appropriate parties in response to events that are triggered by business processes, the organizations themselves, or other applications. In this manner, the active documents allow organizations to exchange information, use decision tools to act on the shared information, respond to dynamic events that require decision making, and engage the proper role players in accordance with the business process.
[0006] The method and means described in the '915 patent provide the most value in a manufacturing supply chain where people from all the supply chain partners collaborate to create the production and distribution plans (called “business scenarios”). The '915 patent focuses on user access security, workflow routing of the “active documents” (i.e., Lotus Notes documents) and the inclusion in those documents of links to data warehouse information sources and decision support tools which the users can utilize in defining, analyzing, modifying, and approving the business scenarios. While generally response to the needs of the supply chain partners, the system and method described in the '915 does suffer, among other disadvantages, the disadvantage of requiring human involvement in the analysis, planning, and approval stages.
[0007] To overcome this and other disadvantages prevalent in currently implemented supply chain management tools, the subject invention resides in a method for managing inventory within a supply chain. The method is performed by providing forecasts of demand for items distributed within the supply chain, using the forecasts to establish base stocking levels and reorder points within the supply chain, and using the established base stocking levels and reorder points to position items within the supply chain so as to maximize efficiency and profitability when responding to an order for an item. The demand for items may be calculated, in part, using collected historical demand data.
[0008] A better understanding of the objects, advantages, features, properties and relationships of the invention will be obtained from the following detailed description and accompanying drawings which set forth an illustrative embodiment and which is indicative of the various ways in which the principles of the invention may be employed.
[0009] For a better understanding of the invention, reference may be had to a preferred embodiment shown in the following drawings in which:
[0010]
[0011]
[0012]
[0013]
[0014]
[0015]
[0016]
[0017]
[0018]
[0019]
[0020]
[0021]
[0022]
[0023] Turning now to the figures, wherein like reference numerals refer to like elements, there is illustrated an automated, comprehensive, end-to-end supply chain management system that connects customers, distributors, and suppliers. Within the figures, it will be appreciated that the distributor is sometimes referenced as “Grainger.” In particular, the system allows the supply chain participants to have cognizance of the entire operation of the supply chain as it pertains to their respective business interests. As will become apparent from the description that follows, the system is particularly adapted for use in supply chains where supply chain management and execution is done as a result of specific business transactions between the customer and the distributor and the fulfillment of the customer's requirement is not always done in a routine, repetitive, or pre-defined manner. A specific example of such a supply chain is a maintenance, repair, and operating (“MRO”) supply chain.
[0024] More specifically, the supply chain management system and method of the present invention will allow companies to operate an entire supply chain on a “just in time” basis without requiring those companies to keep an excessive level of product safety stock on hand. The supply chain management system will achieve this goal by being implemented on a comprehensive and reliable data communications network that serves as the foundation for the connectivity between the supply chain participants. The system includes a collection of functions and features implemented in software and/or hardware that make the operation and management of the supply chain as automated as possible. The software includes intelligent software agents that are distributed across the supply chain. The intelligent software agents allow for the monitoring and managing of state changes in the supply chain.
[0025] For determining the amount of inventory needed by a particular supplier, the supply chain management system also includes a database of forecast data. The forecast data may be comprised of the following:
[0026] (1) Expected consumption rates based on historical data. The expected consumption rates can be developed by tracking actual consumption rate data for customers and compiling a database of those actual consumption rates. This can be accomplished by providing a collaborative relationship between the distributor and the customer and continually updating the actual consumption rate data and the expected consumption rate data;
[0027] (2) Deterministic demand data based on scheduled maintenance for customer equipment and/or facilities including an Equipment Knowledge Database comprised of reliability data, maintenance requirements and completed maintenance job data. The deterministic demand data may also include information related to expected repair parts, contingent repair parts, ancillary supplies and necessary tools. The customer may also choose to maintain a small “just in case” inventory or to receive “just in time” delivery from the distributor. Again, a collaborative relationship will be required between the customer, the distributor, and the supplier for developing and maintaining a deterministic demand database that is populated with deterministic demand data; and
[0028] (3) Non-deterministic demand data based on the Equipment Knowledge Base. The non-deterministic demand data is formulated by comparing historical maintenance history for customer equipment to schedule maintenance plans. The difference between the historical maintenance demand level and the preventive maintenance demand level produces the non-deterministic demand data. In contrast to the basing of inventory on aggregate market forces, the non-deterministic demand data is advantageous because it is compiled at a customer-specific level (regarding planned maintenance work, equipment inventories, and the general condition of that equipment as it pertains to potentially needing maintenance) that makes the data more useful and precise.
[0029] The above-described databases are populated by data from the customer, the distributor, the supplier/manufacturer and possibly third party data compilers/publishers. For example, the customer will supply historical maintenance results, advance demand order information, and scheduled maintenance information, which may include scheduled preventive maintenance, condition-based maintenance, planned maintenance projects, etc. The equipment manufacturer will provide equipment maintenance information, replacement information, and data on anticipated reliability. The distributor will supply actual consumption rate information.
[0030] In operation, assuming that the customer has cataloged its equipment and similar inventory into a computerized maintenance management system (“CMMS”), the customer begins by creating a work order
[0031] If the customer should change, reschedule, otherwise modify or cancel the maintenance work order in the maintenance system, the intelligent agent that monitors the maintenance system will detect such a change. Since such changes to the work order may necessitate a change in the advance demand notice, the intelligent agent should respond to any change in a work order to accordingly change the advance demand notice. Furthermore, as will be apparent from the description that follows, changes to an advance demand notice may or may not require changing the actions of the supply chain management and processing system, i.e., by changing the fulfillment plan, the execution of a selected fulfillment plan, etc.
[0032] The distributor connected to the supply chain management system will receive the advance demand notice from the intelligent agent and a further team of intelligent agents will collaborate to determine the probability that the product(s) listed in the work order will be needed
[0033] In an alternative embodiment, a probability of need can be provided by the customer in an advance demand notice order. The customer may have knowledge of the probability of need from internal condition monitoring systems, equipment status assessment systems or other internal system/sources. As noted, if the customer does not provide the probability of need data, the distributor can estimate a probability of need based on data contained in an equipment knowledge base, described hereinafter. Such estimates by the distributor can utilize historical data from multiple customer's experiences with like or similar instances of equipment or parts being maintained.
[0034] At the distributor, an intelligent software agent may accept the advance demand notice and determine a fulfillment level for the product
[0035] By way of example, if the fulfillment level is the speed and convenience model, the buy-hold-sell (“BHS”) procedures and practices will be implemented. Moreover, if the probability of demand is 100% in the speed and convenience model, then the product will be shipped from the distributor to the customer. If the probability is less than 100%, the product will be reserved for this demand, but kept at is current location.
[0036] Based on a plurality of sourcing factors, intelligent agents determine sourcing alternatives for the advance demand notice. The sourcing factors may include factors, such as, the required fulfillment level for a particular customer, the inventory supply category for the product, the lead time required for the maintenance project, and the number of sourcing suppliers capable of providing the product to the customer according to the customer's time and delivery constraints. If an automated sourcing alternative cannot be generated, an intelligent agent should contact a designated customer contact to initiate the offline selection of a sourcing option. The customer contact may be charged with coordinating the approval procedure at the customer site. The designated customer contact can be contacted via telephone, e-mail, or similar communication means. All off-line decisions by the designated customer contact should be entered into the customer system and synchronized with the supply chain management system including the distributor computer system, the supplier computer system and any other computer system attached to the supply chain management system that tracks customer purchase orders.
[0037] For automatically generating sourcing options, an intelligent agent may first determine the distributor's ability to provide the product. For example, the product may be held on an available to promise (“ATP”) or capable to deliver (“CTD”) status. The various sourcing options are established supply chain methods and practices. For example, if the product is being sourced from the distributor, then a fulfillment level may be provided with the advance demand notice. The available fulfillment levels may start back, i.e., “upstream,” as far as possible in the distribution network. Once the supply chain management system validates a particular sourcing option, the supply chain management system may then issue a purchase order for the product in accordance with the level of service requirement provided within the advance demand notice constraints.
[0038] It will be appreciated that a fulfillment plan may reach “upstream” to involve a supplier if a product must be ordered from the supplier for the subject advance demand notice either to be staged at some point in the logistics network or to be shipped directly from the supplier to the customer. The supplier will also be involved in the fulfillment plan if the plan has the advance demand notice being sourced from the distributor's stock, and the distributor's inventory will have to be immediately replenished if the customer does really need the product that has been reserved. Depending on the supplier's level of business process sophistication, technology sophistication, electronic communications capabilities, and manufacturing methods utilized for the product (e.g., make-to-order or make-to-stock), the supplier may become collaboratively involved in the fulfillment planning process by providing the available to promise (“ATP”) and capable to deliver (“CTD”) information to the distributor.
[0039] A fulfillment plan will be attached to the advance demand notice and any product retained within the inventory of the distributor's logistic network will be committed as part of the fulfillment plan. The reservation and commitment of product to the advance demand notice may require the distributor to move and reposition the product to different stocking points in order to comply with the customer's level of service specifications. The planning should also account for the distributor's safety stock requirements at the various stocking points. The supply chain management system may monitor the progress of the fulfillment plan and may make this information available to suppliers and customers. The fulfillment status may also be monitored by authorized parties to allow the parties to track delivery of the products. If the movement is not proceeding according to the plan, corrective and responsive actions can be initiated. The actions may include the determination and execution of a new, alternative fulfillment plan which takes into account the current state and status of the logistics network that caused the initial fulfillment plan to fail.
[0040] As noted, an intelligent agent may be used to verify that the fulfillment plan is being executed according to the level of service requirements provided by the customer
[0041] If the probability of demand is 100% for a particular product the customer may use the advance demand notice as the purchase order or issue an independent purchase order. If the advance demand notice serves as the purchase order, the product may be automatically delivered to the customer. If the probability of demand is less than 100% the product will remain at a delivery site in accordance with the level of service agreement. When the customer determines that the product is actually needed
[0042] When the probability of demand is less than 100%, the advance demand notice may also indicate how long the product should be held at the final staging point before a purchase order must be received from the customer. For example, if the probability of demand is only 50% and a purchase order is not received within a designated time, then the supply chain management system will assume that the product is not needed in the maintenance task and determine what should be done with the product 209. Depending on the contractual agreement with the customer, there may be a restocking fee for unneeded products.
[0043] When a maintenance task is complete, an intelligent agent can capture pertinent maintenance data which includes, among other things, lists and quantities of products used during the maintenance work and statements of equipment condition before and/or after the maintenance work
[0044] Turning now to the comprehensive and reliable data communications network that serves as the foundation for the connectivity between the supply chain participants, the supply chain management system comprises a network of general purpose computing devices having computer executable instructions. An exemplary network is illustrated in
[0045] For performing the tasks in accordance with the computer executable instructions, the general purpose computing devices preferably include one or more of a video adapter, a processing unit, a system memory, and a system bus that couples the video adapter and the system memory to the processing unit. The video adapter allows the computing devices to support a display, such as a cathode ray tube (“CRT”), a liquid crystal display (“LCD”), a flat screen monitor, a touch screen monitor or similar means for displaying textual and graphical data to a user. The display allows a user to view information related to the operation of the supply chain management system.
[0046] The system memory in the general purpose computing devices may include read only memory (“ROM”) and/or random access memory (“RAM”). The general purpose computing devices may also include a hard disk drive for reading from and writing to a hard disk, a magnetic disk drive for reading from and writing to a magnetic disk, and/or an optical disk drive for reading from and writing to a removable optical disk. The hard disk drive, magnetic disk drive, and optical disk drive may be connected to the system bus by a hard disk drive interface, a magnetic disk drive interface, and an optical disk drive interface, respectively. The drives and their associated computer-readable media provide non-volatile storage of computer readable instructions, data structures, program modules and other data for the general purpose computing devices.
[0047] To connect the general purpose computing devices within the network, the general purpose computing devices may include a network interface or adapter. When used in a wide area network, such as the Internet, the general purpose computing devices typically include a modem or similar device which functions in the same manner as the network interface. The modem, which may be internal or external, may be connected to the system bus via an external port interface. It will be appreciated that the described network connections are exemplary and that other means of establishing a communications link between the general computing devices may be used. For example, the supply chain management system may also include a wireless access interface that receives and processes information from the general purpose computing devices via a wireless communications medium, such as, cellular communication technology, satellite communication technology, blue tooth technology, WAP technology, or similar means of wireless communication.
[0048] While the preferred network is the Internet because of its ubiquitous accessibility and cost advantages, there may situation where specific supply chain partners may choose to implement the network between them as a private network. Examples of private networks include, but are not limited to, virtual private networks (“VPNs”) and electronic data interchange (“EDI”) networks. In this regard, the description of the system which follows will support the use of both public and private networks.
[0049] To provide network security, the network may also utilize security techniques that have become customary when conducting electronic business. These security techniques include, but are not limited to, firewalls, encryption, authentication certificates, directory-based user registration and security management, etc. Because the capabilities and best practices of network communication security are constantly evolving and improving, this document does not espouse the use of any particular technique, technology or product. Rather, it simply specifies that the network architecture should support the use of security practices necessary to protect the business interests of the participants and to insure the overall integrity and confidentiality of the supply chain.
[0050] For exchanging information between the supply chain partners, the network preferably utilizes TCP/IP as the foundation communications protocol. Generally, HTTP and HTTPS are utilized on top of TCP/IP as the message transport envelope. These two protocols are able to deal with firewall technology better than other message management techniques. However, supply chain partners may choose to use a message-queuing system instead of HTTP and HTTPS if greater communications reliability is needed. An example of a message queuing system is IBM's MQ-Series or the Microsoft message queue (MSMQ). The system described hereinafter is suited for both HTTP/HTTPS, message-queuing systems, and other communications transport protocol technologies. Furthermore, depending on the differing business and technical requirements of the various supply chain partners, the physical network may embrace and utilize multiple communication protocol technologies.
[0051] For the purpose of providing a better understanding of the electronic data communications that preferably take place over the network, the network can be divided into two logical groupings, namely, the network for agent interaction and transaction passing and the network used for human-based intellectual and cognitive collaboration activities. Since collaboration involves different kinds of communication services that are frequently more synchronous in character than those of the transaction domain, the collaboration network has been differentiated from the agent/transaction network illustrated in
[0052] Focusing now on the intelligent agent network, the intelligent agent network is provided to connect the supply chain partners so that business transaction and agent messages can be exchanged throughout the supply chain. The intelligent agent network is controlled and managed by an Agent Network Manager and Broker
[0053] To perform the various pieces of the functionality that comprise the intelligent agent network, the system includes a collection of stationary agents. The stationary agents are hosted on agent servers
[0054] At the customer spoke of the hub, illustrated in
[0055] At the distributor spoke of the hub, illustrated in
[0056] At the supplier spoke of the hub, illustrated in
[0057] As further illustrated in
[0058] Turning now to the broker
[0059] The broker
[0060] Still further, the broker preferably operates a publish and subscribe service
[0061] The broker
[0062] Still further, the broker
[0063] To perform the functions noted above, the broker
[0064] The agents and roles directory
[0065] The brokerage status database
[0066] It is to be noted that, while not illustrated, the broker
[0067] As described above, agent servers
[0068] The communications layer
[0069] Above the protocol services
[0070] Above the XML parser
[0071] The communicator services
[0072] Turning now to the agent standards layer
[0073] By way of example, if one of the application task agents at the spoke needs to initiate communications with some other spoke in the network, then the particular application task agent will send an initiation request together with the appropriate data to the domain manager
[0074] The domain manager
[0075] The agent interaction protocol or AIP manager
[0076] To perform according to the tasks that are prescribed in the AIP specifications for the conversation type, the AIP manager
[0077] The AIP manager
[0078] Turning to the task manager layer
[0079] As illustrated in
[0080] information agents which locate information sources, extract information from those sources, provide necessary filtering of the information for relevance, and prepare the resulting information for return to the requester.
[0081] integration agents which work in the other direction from information agents in that they receive information and data from outside the spoke and add it to the appropriate legacy system or database at the spoke.
[0082] cooperation agents which take plans created by planning agents and then direct the task manager
[0083] planning agents which develop plans and strategies. The planning agents plan out complex tasks that do not have a pre-defined execution path or AIP defined. An example of a planning agent would be at the customer site. If a shipment or a critical repair part is delayed significantly to completely ruin a maintenance schedule, the planning agent will first determine how to define an alternative sourcing for the part and what corrective actions need to be done with the maintenance scheduling and execution work. To this end, the planning agent may engage the cooperation agent to initiate the corrective work as this may involve the cooperation between several spokes on the network.
[0084] transaction agents which handle business transaction data, usually acting as interfaces to the site's legacy application systems.
[0085] believability agents which are basically simulators that will test certain suggested plans or hypotheses. Within an MRO supply chain environment, these agents are likely to be hosted by the distributor's site/spoke, but can be accessed and utilized by other partner spokes.
[0086] assistance agents which serve as personal assistants to humans. These agents help to locate and retrieve information, provide performance support assistance, and provide other human interface services such as the maintenance application services to maintain ontology files.
[0087] anticipation agents which serve to monitor situations and anticipate developing problems.
[0088] Most application task agents will be developed and supported by the company or organization that is sponsoring and operating the supply chain to insure that there is as much consistency as possible across the entire network. In many cases, this will be the distributor. However, there may be situations where those agents that interface to legacy systems will be written by either the spoke company or by the software vendor supporting the legacy system.
[0089] By way of further example,
[0090] The user interface agents
[0091] The algorithm executor agents
[0092] The rules inference engine
[0093] The knowledge manager
[0094] The collaboratory interface
[0095] Another purpose of the collaboratory interface
[0096] At the supplier spokes, shown in
[0097] At the transportation spoke, legacy interface agents
[0098] At the distributor spoke, shown in
[0099] As discussed above, the preferred knowledge exchange protocol used is the Knowledge and Query Manipulation Language or KQML. KQML is the result of a research project sponsored by DARPA. A KQML web site is maintained by the University of Maryland Baltimore County (http://www.cs.umbc.edu/kqml/) and contains detailed descriptive information as well as case studies of its use. In particular, the web site includes the current detailed specification of KQML.
[0100] KQML defines both a message format and a message handling system for multi-agent systems. The message transmission definitions are virtual. Instead of defining a particular technical standard for implementing message transport, KQML defines a model of the services that the message transport system performs. The KQML message transmission definitions are incorporated and modified as needed in the communications layer
[0101] KQML is built around a set of performatives. A performative is a fundamental action that one agent can request another agent to perform. The definition of performatives is open-ended in KQML. Implementers can add specialized new performatives as long as the parameters and syntax comply with the KQML specifications. When used in connection with a MRO supply chain an additional group of performatives be added that can be called “Business Transaction Performatives.” This new group allows business transactions to be handled over the agent network as well as agent message traffic and also allows the agents to interface with and become involved with the business transactions. Examples of Business Transaction Performatives are:
[0102] Trigger-Event—the recipient is receiving a business transaction that it should process but is not subject to transaction semantics (two-phase commit)
[0103] Start-Transaction—the recipient is receiving a business transaction that it should process and the transaction is subject to transaction semantics
[0104] Add-Event—the recipient is receiving another transaction in the set or series embraced in the currently active Start-Transaction
[0105] Commit-Transaction—All agents have successfully processed the transaction events in the logical Start-Transaction set and can now commit and all agents can now commit their results
[0106] Abort-Transaction—the sender is unable to successfully process a transaction event in the current Start-Transaction series and therefore all agents participating in the currently active Start-Transaction set should abort their efforts and back-out any results already completed
[0107] Each performative has a series of parameters identified by keywords. A new performative parameter may be added to the standard parameters defined in KQML to specify the name of the AIP that specifies the steps to follow in executing the performative. This is the “:aip<word>” parameter.
[0108] Turning to the ontology
[0109] The ontology
[0110] The frame construct also allows operations
[0111] The operations
[0112] Since frames
[0113] As noted, frames
[0114] Likewise, if a frame
[0115] The semantic network in the ontology
[0116] Knowledge and understanding are achieved by the agents by traversing the ontology
[0117] Like any other database, the ontology
[0118] The supply chain partner should also provide a core set of AIPs which, among other execution directives, define what kind of relationship names to use in different performative actions and what kinds of relationship names to create when creating new relationships. Therefore, the real semantic definition of relationships is inherent in the AIPs since the AIPs are determining how the concept frames and relationships are being used and interpreted by the various application task agents. In a preferred embodiment, the agent interactive protocol or AIP is patterned after the RosettaNet Partner Interface Process or PIP. The PIP depicts the activities, decisions and partner role interactions that fulfill a business transaction between two partners in the supply chain. PIPs are explained in detail on the RosettaNet web site: http://www.rosettanet.org/.
[0119] The AIP builds on the PIP implementation and is a more detailed and expanded concept specifically designed to coordinate the execution of various agent tasks within a single agent server
[0120] More specifically, the AIP specifies overall sets of activities that each agent server/spoke
[0121] The definition of the task activities in the AIP can, at times, be very granular and detailed. The various task agents are provided with instructions and processes on how to complete their assigned responsibilities. This also includes directives regarding how to use and update the ontology
[0122] To insure that all participating agent servers
[0123] Another feature of the AIP is that there are circumstances when it is not possible to predetermine exactly how many exchanges of intermediate conversations between the various agents will take place. Sometimes there are many iterations of intermediate communications until some conclusion is reached between cooperating agents. For this situation, the AIP has an activity definition construct that specifies an iterative loop to take place until some condition is met (equivalent to the programming constructs of “while <condition>do <activity>” or “do <activity>until <condition>”).
[0124] To maintain the various pieces of the AIP, including the specification of task activities, technicians can use the user interface Web server
[0125] For managing order fulfillment within the supply chain network, the system includes agents for intelligently selecting a fulfillment plan. Determination of the fulfillment plan balances the requirements of staging the product according to the customer's level of service specification against the possibility that the customer will end up not needing the product. If the product is not needed, the distributor may end up with the product which has very little overall demand being inventoried at a point where handling and transportation costs to reposition the product to a location which has more overall demand will totally consume the distributor's gross margin on that product. Therefore, the planning method should seek to economically stage the product to comply with the level of service requirements while selecting a staging point that is economically feasible for the distributor given the probabilities of need and the expectation of future, near term demand.
[0126] To this end, as illustrated in
[0127] delivery to the customer's maintenance site;
[0128] delivery to the customer's storeroom/tool crib;
[0129] staging at a distributor branch for short-notice will-call pickup;
[0130] staging at a distributor distribution center for overnight shipment to the customer;
[0131] staging at a distributor distribution center for lower priority shipment to the customer to replenish customer safety stock inventory if the customer had to use inventory in the maintenance task covered by an advance demand notice; and
[0132] staging for direct shipment from a manufacturer/supplier for lower priority shipment to customer to replenish customer safety stock inventory if the customer had to use inventory in the maintenance task covered by an advance demand notice.
[0133] It is to be understood that these levels of service are exemplary only and are not meant to be limiting. It is also to be understood that the steps of the fulfillment planning process can be applied to all orders received by the distributor including speed and convenience orders for unplanned purchases having a 100% probability of need, planned advance demand notice (level of service) purchases with both certain (100% probability of) need and uncertain (less than 100% probability of) need. In this way, the same inventory base and logistics capabilities throughout the supply chain can be used for all orders and there is no need for the distributors to have separately managed inventories for each type of order or business model.
[0134] For use in determining candidate sourcing points for each line item of the order, the agents employ a branch and bound technique. This approach is desirable as it reduces the number of alternatives that might otherwise be considered using every combination and permutation of sourcing points throughout the logistics network. The branch and bound approach is implemented by selecting tiers of sourcing points in a n-tier logistics network generally illustrated in
[0135] The criteria for defining each tier depends on the specific logistics network topology that is being used. Therefore a table and parameter driven approach is used to define these tiers and how to assign sourcing points to each tier. By maximizing the use of parameter tables and rules, the basic intelligent fulfillment planning process can be used in any kind of logistics network topology. This is desirable as the system can be adapted quickly without major programming activity as a company changes its logistics network topology to respond to changes in business activity and sales patterns.
[0136] To proceed with the intelligent order fulfillment planning procedure, the system utilizes some or all of the data contained in an electronically transmitted order, i.e., an order generated at the customer spoke/distributor spoke and issued to the distributor spoke for fulfillment. The data contained in the electronically transmitted order preferably includes, but is not limited to, the following:
[0137] type of order (e.g., walk-in, will-call, ship, advance demand notice reserve);
[0138] point of delivery;
[0139] code for the supplier branch that normally services this customer's site;
[0140] customer's end-delivery site ZIP code;
[0141] delivery date;
[0142] whether or not the customer requires a single consolidated shipment or whether the customer will accept multiple split shipments;
[0143] supplier stock number for each line item in the order;
[0144] quantity ordered for each line item in the order; and
[0145] any supplier diversity requirements for the line items.
[0146] For advance demand notice orders the point of delivery (i.e., level of service) should be specified as the customer's delivery site designated by the ZIP code or the code of the distributor facility where the advanced demand notice is to be staged per the level of service agreement. For walk-in and will-call orders, the point of delivery should be specified by the code for the distributor branch that normally services this customer's site. For ship orders, the point of delivery should be specified by the code for the distributor distribution center that supports the customer's normal branch. Furthermore, for advance demand notice orders, the date should specify the date on which the product(s) is/are to be staged at the level of service or “LOS” point. For ship orders and will-call orders the date should specify the date the customer wants the product either shipped or at the will-call counter and, for walk-in orders, the date should specify the current date.
[0147] To further facilitate the intelligent order fulfillment planning procedure, for each line item within an order the agents may determine if the distributor has access to any equivalent products
[0148] In accordance with the procedure for determining fulfillment plan alternatives, illustrated in
[0149] For determining the primary sourcing point for an item, the information in the ontology
[0150] To determine the tier one secondary sourcing points, the information in the ontology
[0151] When assigning tier one secondary sourcing points to line items, it is further preferred that sourcing points be eliminated if the sourcing point is not equipped to handle direct shipment to customer's sites for ship-to orders and advance demand notice orders staged at the customer's site. Similarly, tier one secondary sourcing points may be eliminated if they are not capable of shipping Hazmat items. In the case of will-call orders, the tier one secondary sourcing points must be capable of having the customer come to the location to procure the requested item. As will be appreciated, walk-in orders are converted to will-call orders if the customer is referred to a tier one secondary sourcing point for order fulfillment.
[0152] Once the primary and tier one secondary sourcing points are determined for each line item, the intelligent order fulfillment planning procedure determines the on-hand inventory at the assigned sourcing points
[0153] To determine the fulfillment plan alternatives, the primary and tier one secondary sourcing points and their respective on-hand inventory for each line item is considered. If the entire order (i.e., each line item) can be filled using unreserved, on-hand inventory at the primary sourcing point, this inventory and sourcing point should be used to create the fulfillment plan. If, however, the primary sourcing point does not have enough unreserved, on-hand inventory to satisfy the entire order, additional fulfillment plan alternatives may be considered.
[0154] For creating additional fulfillment plan alternatives, the system may first determine if any currently reserved on-hand inventory at the primary sourcing point can be feasibly reallocated to the subject order. The reallocation may be deemed to be feasible if the item(s) to be replenished can be received, put-away, and packed/delivered in time to meet the order for which the item(s) are reserved. The costs associated with replenishing items in this manner should be assigned to the order that necessitated the action and not to the order associated with the reservation. A fulfillment plan alternative created in this manner will be added to a list of fulfillment plan alternatives to be considered for possible implementation.
[0155] A further fulfillment plan alternative can be created if the entire order can be supplied from the unreserved, on-hand inventory at the determined tier one secondary sourcing point. If the tier one secondary sourcing point has the unreserved, on-hand inventory to meet the order, this fulfillment plan alternative will be added to the list of fulfillment plan alternatives to be considered. If, however, the tier one secondary sourcing point does not have enough unreserved, on-hand inventory to meet the order, the system may determine if any currently reserved on-hand inventory at the tier one secondary sourcing point can be feasibly reallocated to the subject order using the approach described previously. A fulfillment plan alternative created using reallocated items would also be added to the list of fulfillment plan alternatives to be considered.
[0156] Candidate fulfillment plan alternatives will have a sourcing point for each line item in the quantities required by the order. The line item quantities can, however, be split between multiple sourcing points. Accordingly, the system should create fulfillment plan alternatives that use the various permutations/combinations of sourcing points capable of fulfilling either entire or partial line items at the primary and tier one secondary sourcing points.
[0157] Once all fulfillment plan alternatives have been created and added to the list for consideration, a fulfillment plan is selected for possible implementation. If multiple fulfillment plans have been created, selection should be based upon an evaluation of the costs associated with each fulfillment plan alternative. If no alternative fulfillment plans have been created using the methodology described above, a further iteration may be performed to devise additional fulfillment plans for consideration.
[0158] To select the fulfillment plan for the order, a cost analysis may be performed
[0159] To evaluate the routing of line items the shipping requirements of the customer are considered. How each line item is routed depends upon whether the customer requires a single consolidated shipment. If consolidation is required, additional candidate fulfillment plan alternatives need to be generated to consider a sourcing point at which to consolidate the order. For this purpose, three sourcing points can be considered for the point of consolidation: 1) the sourcing point from which the most line items are being sourced; 2) the sourcing point that minimizes the time required to consolidate all items not sourced at that sourcing point; and 3) the sourcing point that minimizes the cost required to consolidate all items not sourced at the sourcing point. Using this approach, each non-routed fulfillment plan alternative will be expanded into three routed alternatives, i.e., one for each of the possible consolidation points.
[0160] If the customer does not require a consolidated shipment, the system may nevertheless attempt consolidation should consolidation be determined to be advantageous. For example, consolidation may be advantageous if there are multiple line items with small-sized and/or low-valued items since the shipping costs to the customer of such items may exceed the combined shipping costs of moving the item using LTL to a consolidation point and then using package shipment for the final shipment to the customer. Since this trade-off is very situation dependent, it may be preferred to create these consolidated alternatives and then compare the costs associated with the consolidated alternatives to the costs associated with the non-consolidated alternatives. If the consolidated alternatives are determined to be more cost effective, the consolidated alternatives will be further considered even if the customer does not require consolidation.
[0161] To determine the activity costs for each line item within an order, the activity costs are accumulated as the item movement is traced from the point of sourcing to the point of final delivery. Activity costs include, but are not limited to: pick-up costs; put-away costs; cross-docking costs (if this occurs en-route to consolidation); carrying costs (if the item is inventoried temporarily during the consolidation process); shipping costs (e.g., LTL costs between logistic points and/or Package Shipment costs between logistic points and to the customer if required); and holding costs of the item's value over the delivery fulfillment period. The activity costs may be kept in parameter tables by location and by physical attributes of the items (e.g., size, weight, Hazmat, etc.). parameter values used in computing the activity costs may be kept in the ontology
[0162] Further factors that may be considered when selecting a fulfillment plan may be used depending upon the situation and the business. For example, if there are other known orders for another customer, then there would be an opportunity cost associated with using the candidate item for the subject order and not fulfilling the order for the other customer. The magnitude of this opportunity can be partially based on the relative preferences of the two customers. In the case of advance demand notice orders, the difference between the probability of need for each advance demand notice order needs to be considered (regular orders are considered to have a 100% probability of need).
[0163] Another form of opportunity cost that can be considered is an anticipated peak in demand for the item at the sourcing location. This can occur either because of seasonal changes or because recent demand has been significantly below expectation and the law of averages expects that the short-term demand will be larger in order to reach the overall average demand over time. In this case, the computed opportunity cost will tend to favor alternatives where the item is sourced from a point not anticipating a short-term increase in demand over a sourcing point that is anticipating such a demand.
[0164] Yet another factor that may be considered is the degree to which the alternative fulfillment plans uses excess inventory. Excess inventory is the amount of unreserved inventory on-hand at a souring point that exceeds the stocking level for the location. An item may also be considered to be excess inventory if the item has not been demanded at a particular sourcing point for a given period of time (e.g., x months). If the use of excess inventory will be a factor in the evaluation of alternative fulfillment plans, then the cost associated with using excess inventory can be weighted to favor the selection of such a fulfillment plan. This can be accomplished by setting the cost equal to the negative of the carrying cost for a predetermined period of time, preferably, ½ the time period established above with respect to determining when an item is excess inventory.
[0165] Yet another factor that may be considered is the age of the inventory being used to source an order. In this case, the alternative fulfillment plan that uses the oldest inventory can be weighted to favor the selection of such a fulfillment plan. In order to provide this weighting, the ages of the inventory in all the alternative fulfillment plans will be considered. The age is the current date less the date at which the item was added to the inventory at the sourcing point. The cost may then be the maximum age minus the age of the candidate items multiplied by the daily carry cost amount for the item at that location. The following Table 1 provides an example.
TABLE 1 Fulfillment Plan Age Of Daily Carry Cost At Alternative Inventory Location Relative Cost 1 15 $0.0100 $0.72 2 5 $0.0140 $1.15 3 45 $0.0110 $0.46 4 87 $0.0100 $0.00
[0166] In the example provided, the youngest alternative (5 days) has the largest cost while the oldest alternative (87 days) has zero cost. This type of factor is particularly useful if the distributor was concerned about an item becoming excess.
[0167] To compute the total weighted cost for the alternative fulfillment plans
[0168] To select one of the alternative fulfillment plans for implementation, the system determines if the alternative fulfillment plans meet a gross margin threshold for the preference category selected by the customer. The gross margin is computed for each of the alternative fulfillment plans as the total revenue to be received by the distributor (i.e., item price plus customer-paid shipping plus any special service charges) less the activity costs. For the advance demand notice situation, the distributor revenue is computed as though the customer has a 100% probability of need and will take delivery of the item. If the gross margin threshold is not met, the alternative fulfillment plan will not be considered to be profitable and will be disqualified from further consideration. If all of the alternative fulfillment plans fail to meet the gross margin threshold, the customer will be told that the order cannot be filled. On the other hand, if an alternative fulfillment plan is determined to meet the gross margin test, has the lowest total weighted cost to the supplier, and meets all of the delivery and price quote requirements of the customer, the alternative fulfillment plan is selected as the best plan for implementation. Alternatively, the system can eliminate those line items that cannot be fulfilled within the customer's requirement constraints leaving only a partial fulfillment to the order. The partial fulfillment can then be re-evaluated to determine if it now falls within the customer's requirements.
[0169] When determining if an alternative fulfillment plan meets the price quote requirements of the customer, it is determined whether the order quoted price and the actual fulfillment cost to the customer falls within a customer established tolerance. Tolerances for each customer may be maintained in the database. It will be appreciated that the actual fulfillment cost may vary from the quoted price owing to the need to meet the order from secondary sources and/or by substituting items. If the tolerance is exceeded, further customer authorization may be required to fulfill the order. If no authorization can be obtained, the fulfillment plan should be disqualified from consideration. When communications are possible with the customer, data should be solicited regarding the customer's ability to consider upgrade/downgrade item substitutions and/or the ability to backorder line items or the entire order. The backorder time can be either the time to fulfill the problem line item from the preferred remote sourcing point or the time required to order the line item from the manufacturer and either drop ship the item directly to the customer or deliver the item through the distributor's logistic network.
[0170] If no primary sourcing point fulfillment plan or tier one sourcing point fulfillment plan alternatives are selected as a viable fulfillment plan using the above described methodology, a second iteration of the above described steps may be used to create further sourcing fulfillment plan alternatives
[0171] If a line item cannot be filled using either the first or second iterations described above, a third iteration of the steps described previously may be employed wherein substitute items are considered
[0172] In sum, while the described intelligent order fulfillment plan methodology provides many alternatives to arrive at a viable order fulfillment plan, in practice, the vast majority of orders that will be entered into the system will be for single line items that may be fulfilled using the first or second iteration of the steps above-described. Therefore, in practice, there will not be a large number of alternatives that are created and evaluated. Furthermore, the branch and bound structure of the four described iterations will tend to limit the number of alternatives that will be generated and considered. Therefore, even if third and fourth iterations are utilized, the need to use these iterations will be reached only because there is limited item availability within the distributor logistics network. The first and second iterations will not have created a large number of alternatives in such instance because there will not have been a large number of possible sourcing points with available inventory.
[0173] For managing inventory within the n-tier logistics network, the system may also include agents for performing intelligent inventory management. The inventory management agents are provided to forecast item demand and establish base stocking levels and re-order points for items throughout a distributor's logistics network. The inventory management logic thus allows the distributor to position items within the logistics network in a manner that should maximize efficiency and profitability when selected sourcing plan alternatives are utilized.
[0174] The processes involved in managing the inventory are presented in a generalized manner so that they can be applied to any logistics network topology. Consequently, they are implemented using a high degree of table-driven and parameter driven software engineering techniques. In connection with the software, the described system is desired to operate in conjunction with the ontology
[0175] As will become apparent from the description that follows, the emphasis of the inventory management process is to allow the distributor to deal with probability and uncertainty in demand. Another purpose of the described process is to give suppliers/manufacturers advanced information regarding forecasts of future demand. After each forecast iteration, the manufacturer can be given consolidated demand and replenishment forecasts for their items over the forecasting horizon. The consolidated data can be further broken down by anticipated delivery point. Some manufacturers may desire the data for the immediate next quarter to be broken down in a finer level of detail such as forecasts by week.
[0176] In the performance of the inventory management process, illustrated in
[0177] The system then forecasts the combined demand and determines the critical stocking ratio that will indicate the total quantity the distributor can afford to hold in inventory during the forecast period. Using the determined critical stocking ratio, the system allocates the permitted inventory level among the various distribution points in the logistics network by assigning over the forecast time period the base stock level and the reorder point for each item at each distribution point. If a item is not to be allocated to a particular point, then the base stocking level for that point will be zero.
[0178] In connection with the allocation of inventory levels among the distribution points in the logistics network, the system determines the replenishment method that will be used for the items. If the use-one, replenish-one method is used, then the reorder point will be set to the base stock level minus 1. If the “(r,s)” policy is used, the re-order point, “r,” will be set to whatever the chosen mathematical or heuristic algorithm determines. At this point, the system may run the initial replenishment processes that will create orders at any point where the unreserved on-hand inventory plus orders-in-transit quantities are less than the designated reorder point. This initial run will also determine the disposition of any item that has been reclassified as excess because the new base stock level for the item has been reduced or set to zero and the unreserved on-hand inventory of the item plus orders-in-transit of the item exceeds the new base stock level. Subsequent replenishments will be based on the current unreserved inventory position and the setting of the re-order points.
[0179] To determine the size of each forecast period for use in performing the inventory management process, system parameters are used. Generally forecasting will be done in monthly periods unless the inventory management team has a particular reason to choose a different period size. Other system parameters are used to determine how many periods to include in the forecast horizon. Generally, this will be either twelve or thirteen months, although there are many situations where the inventory management team will want to forecast only for a quarter or half-year. Still other system parameters may be used determine the frequency with which history data is extracted, aggregated and loaded in the database(s) that will drive forecasting. These activities will generally be tied into the distributors data warehouse Extract-Translate-Load (ETL) schedule since the same data is used for decision support analysis as is used for forecasting.
[0180] Further system parameters can be used to determine the frequency with which forecasting is done and which items are included in each forecasting cycle. Usually in the MRO environment, the distributor's ontology
[0181] To compile the historical demand data, items are categorized according to characteristics that effect the demand for items in a logistics network. The items can be categorized across the following segmentation dimensions: 1) moving category; 2) demand rate; 3) number per order; 4) world factors; and 5) lead-time. Preferably, the categorizations of an item are kept in system parameter tables so that the inventory management team can easily change them. The historical demand data for the items may also be maintained for each souring point within the logistics network. By way of example only, demand data may be maintained for each sourcing point and broken down as follows: 1) demand for replenishing lower tier sourcing points; and 2) demand for fulfilling customer orders (e.g., demand for fulfilling speed and convenience orders and demand for fulfilling advance demand notice orders).
[0182] Turning first to the moving categorization, an item will be classified as either “fast” or “slow” moving based on the level of demand for the item over a given time period
[0183] Within the demand rate category an item will be classified as having a “fixed” or “variable” demand rate based upon the rate that items experiences demand over a given period of time
[0184] Within the number per order category an item will be classified as being either “single” or “lumpy” based upon the number of units of the item typically ordered
[0185] Within the world factors category, an item is annotated as to whether its demand is impacted by external world factors such as the weather, the economy, competition, change in customer status (where demand comes from only a small number of customers), etc
[0186] To consider the historical effect of world factors, it is necessary for the distributor to have recorded the particular world factor events for each historical period. Each type of event has its own method of recording and association with demand. For example, with hurricanes, the occurrences are measured in “events.” For severe summer heat, the world factor can be measured in degree-days for the region serviced by the distribution point. The parameters which define each world factor will also designate how it is to be associated with the demand. The historical data for all periods in which the world factor did not have an occurrence is compared against the periods during which such occurrences happened. The differences between these two demands represents the impact of the world factor during the historical horizon.
[0187] The exact method of associating world factor affect on demand is specified by parameters. Possibly ways include “percent increase per hurricane occurrence” or a table which indicates the incremental amount of demand of item in a period based on the number of days the temperature was a certain amount above the normal. Such a table might be as illustrated in Table 2.
TABLE 2 Number of Days of 10 to 15 degrees 5% increase in demand per day above normal Number of Days of 16 to 20 degrees 10% increase in demand per day above normal Number of Days of over 20 degrees 15% increase in demand per day above normal
[0188] Recall that the item information database has the particular world factor (if any) that applies to the individual item. This will have been established by the item management team.
[0189] Finally, within the lead-time category an item will be classified as either “fixed” or “variable” as a function of the item's lead-time from the supplier
[0190] For those dimensions above-described that have a variable alternative, the item information database should also contain the designation of which stochastic distribution function is the best fit along with the applicable parameter values for that distribution function. Furthermore, it will be appreciated that, when taking the value combinations for these dimensions (moving is fast or slow, demand is fixed or variable, number per order is fixed or lumpy, world factors is yes or none, and lead-time is fixed or variable), all the combinations are considered and there will be at least one mathematical or heuristic forecasting method applicable to each combination. Some combinations will have several possible methods. Also, any particular mathematics or heuristic forecasting method may be applicable to more than one combination. Since the particular mathematics and heuristic forecasting algorithms and methods are well known, they will not be discussed in greater detail herein. Nevertheless, for further information on mathematical and heuristic algorithms and methods for inventory management, the reader is referred to
[0191] When forecasting the combined demand, each item has at least one and ideally several different forecasts computed or determined
[0192] When forecasting the demand to fulfill customer speed and convenience orders, the mathematical algorithm(s) or heuristic(s) used depends on the settings for the moving category, the demand rate, and the number of units per order. If the item is coded as historically experiencing a constant number of items per order, then this constant is used as the number of items per order during the given forecast time period. If the item is coded for a lumpy number of items per order, then the fitted stochastic distribution function is used to determine the probability of “k” number of items per order during the given forecast time period. The value of “k” is determined so that the probability meets a threshold designated by a system parameter. This parameter is closely associated with the desired fill rate. For example, the threshold might be set by the inventory management team to be 95%. This establishes the value of “k” such that the probability that the order will contain “k” or fewer items per order is at least 95%. The value of the number of items per order for the given forecast time period is multiplied by the forecasted number of orders to arrive at the forecasted number of speed and convenience demand for the period.
[0193] When forecasting the demand for fulfilling advance demand notice orders, the system utilizes the Equipment Knowledge Base to analyze the customers who have equipment that might need the item in a maintenance task
[0194] The probabilities of need are added for all the customers in the distribution point's scope. This sum is then rounded to the nearest integer using a threshold parameter as the basis for rounding. This is the threshold that the inventory management team believes is critical in order to comply with the level of service agreements with the customer. For example, this threshold might be specified as 80%. This means that if the probability of need is 80% or more at a distribution point, then we will consider this as a demand for 1 unit for forecasting purposes. The inventory management team considers the 80% likelihood as mandating the physical presence of the item for possible fulfillment. So, in this example, the rounding would occur as follows: 0 to 0.79 would round to zero; 0.8 to 1.79 would round to 1; 1.8 to 2.79 would round to 2; etc.
[0195] Since the distributor will not know if the item is actually needed even when the advance demand notice order is submitted for the maintenance work by the customer, the advance demand notice order rounding is done separately because the probability of need is different in concept from the probability of demand. The probability of need is based on the age and condition of equipment. It may be 10% in one period, 25% in the next period, 65% in the next period, 90% in the next period, etc. This does not mean that there will be demand for the item in each of the periods. Instead, it means that the likelihood of demand is increasing in each period. The actual demand and need will not be determined until the maintenance team inspects the equipment as part of the maintenance work. It is important that this threshold be implemented as a system parameter since it is likely that the inventory management team will have to modify its value from time to time as experience with the customers is gained. While the rounded amount is used as the forecast to the target distribution point, the unrounded sum is sent up to the next tier for accumulation summing over all the lower tier points. Separate rounding will occur at each tier in the logistics network. The rounded advance demand notice forecast is then added to the speed and convenience demand forecast computed earlier for the forecast period. Because the speed and convenience demand may still be a real number (whole integer plus a fraction), this sum may also be a real number.
[0196] The world factor (if any) is then applied to the combined customer fulfillment demand forecast. When world demand is involved, a Markov chain technique may be used
[0197] The next step in determining the combined demand forecast is to consider the tier replenishment forecasted demand. This requires that the demand for all the lower tier distribution points that replenish from the subject distribution point be already computed. The replenishment demand for those lower tiers is added together to create the total forecasted demand for the subject distribution point for the forecast period.
[0198] Finally, the historical data will indicate how much of this subject distribution point's inventory is replenished from higher level tier distribution points as oppose to be replenished directly by the supplier. The percent that is replenished through higher level tier distribution points is multiplied by the total forecasted demand to arrive at the forecasted replenishment demand that will be passed up to the higher level distribution point for its forecast computations.
[0199] It should be noted that for slow moving item, forecasted demand may only be for the given forecast period fractional (that is either less than 1 or a whole number plus some additional fractional amount such as 2.56). This is especially likely to occur at the lowest tier levels. If this is the situation, then the fractional demand is accumulated over chronological periods until it reaches a quantity greater than or equal to one. This whole unit of forecasted demand is then assigned to the forecast period where the accumulated fraction first exceeded 50%. Any rounded off residual fractional demand is assigned to the last chronological forecast period for the next chronological accumulation. An example is illustrated in Table 3.
TABLE 3 Period 0.15 0.35 0.20 0.45 0.15 0.15 0.20 0.10 0.15 0.25 0.20 Forecast Accumulated 0.15 0.50 0.70 1.15 1.30 1.45 1.65 1.75 1.90 2.15 2.35 Forecasted 0 1 0 0 0 0 1 0 0 0 0 Units
[0200] Another example is illustrated in Table 4.
TABLE 4 Period 1.15 0.35 3.20 0.45 0.15 1.15 0.20 1.10 0.15 0.25 0.20 Forecast Accumulated 0.15 1.50 4.7 5.15 5.30 6.45 6.65 7.75 7.90 8.15 8.35 Forecasted 1 1 3 0 0 1 1 1 0 0 0 Units
[0201] Note that the forecasted number of units of demand for each period is the accumulated amount rounded to the nearest whole integer less the accumulated forecasted units from prior periods. As was done with the advance demand notice fractional amounts, the rounded, whole integers are used for the target distribution point, but the unrounded amounts are forwarded upwards to the next higher tier distribution point when considering replenishment requirements. This enables the system to accumulate fractional amounts and do the rounding at the next tier which may result is a more precise network-wide computation of true demand. At this point in the process, there is a consolidated forecast of demand for the item at each point in the logistics network.
[0202] To determine the critical stocking ratios
[0203] For determining allocations and base stocking levels a preliminary base stocking level is first computed for each distribution point by using the mathematical algorithms and heuristics prescribed by the classification of the item along the Moving-Demand Rate Number Per Order-World Factor-Lead-time dimensions. These mathematical algorithms and heuristics will also take into account the lead-time data discussed earlier. Once the preliminary base stocking levels are computed, they are accumulated over the entire logistics network and compared to the maximum stocking level allowed by the critical stocking ratio. If the total is less than the maximum allowed by the critical stocking ratio, then the base stocking levels are finalized and no further adjustments are needed.
[0204] If, however, the total exceeds the maximum allowed by the critical stocking ratio, then allocation must take place
[0205] If the critical stocking ratio does not permit all the distribution points with a preliminary base stocking level to have at least 1 unit stocked, then allocation will be done by giving 1 unit of stock to the distribution points in the order of the distribution points with the largest preliminary stocking levels. The limited stocking amount is allocated in units of 1 to the distribution points in the sequence of decreasing amounts of preliminary stocking levels. For example, if the critical stocking ratio allows for 5 units total, then an exemplary allocation is illustrated in Table 5.
TABLE 5 Inventory Point A B C D Preliminary 2 2 1 1 Base Stocking Level Allocated Base 2 1 1 1 Stocking Level
[0206] However, if the critical stocking ration only allows 3 units, then the allocation will look like as illustrated in Table 6.
TABLE 6 Inventory Point A B C D Preliminary 2 2 1 1 Base Stocking Level Allocated Base 1 1 1 0 Stocking Level
[0207] A system parameter is used to permit the inventory management team to have the allocation process give preference to allocating stock to the middle and higher tier distribution points rather than to the lower tier points. This concentrates the limited stocking ability at points that can be quickly dispatched to the lower tiers of the network when the lower tiers experience real demand. If there is a large difference between the total preliminary base stocking levels and the maximum allowed by the critical stocking ratio, then this situation is brought to the attention of the inventory management team for manual intervention and resolution. There is a system parameter which specifies this difference threshold.
[0208] It is to be understood that the inventory management team has the ability to manually override these automatic base level computations if they choose to impose specific stocking models on certain distribution points. For example, certain lowest level tier points may have a list of “never out” items that is maintained for marketing purposes regardless of the financial considerations embraced in the inventory management processes. Or there may be a minimal “national stocking model” that is applied to all branches to have these branches stock a minimal set of common item.
[0209] To determine the replenishment method
[0210] Once the base stocking levels and re-order points have been set, the first replenishment process that is performed will place the replenishment orders to bring the logistics network into some initial state of readiness. Because of the domino effect in a logistics network topology of many tiers, it may require several replenishment iterations before the inventory state is brought into conformance with the desired state as specified in the inventory planning processes, algorithms and heuristics. During this process a situation may develop wherein forecasted demand from the current iteration may be substantially less than that forecasted in prior iterations. This may result in on-hand inventory becoming excess. When there is excess inventory, the system will execute a “reverse fulfillment planning process” to determine the disposition of the excess inventory. This process will take into account any future expected increases in demand, costs of handling excess inventory, and candidate consolidation points for the excess. It is quite possible that the process may decide that some excess inventory should simply be left in place. This may be the case if there is still forecasted demand, but the new demand is less than the previous demand. This may also be the case if the costs of moving the excess exceeds the value of the excess.
[0211] To determine the degree to which the mathematical algorithms and heuristics used to create the forecast will be employed in the future or replaced with other alternatives, the system may capture the accuracy of the forecasts. In this regard, accuracy may be measured in a couple of ways. One is the amount of overstock that occurs from over forecasting. It will be most apparent in metrics such as total inventory value, inventory turns, amount of excess, etc. Another way is caused by under forecasting which results in inventory outages and is reflected in order fulfillment rates. The historical accuracy of the forecasts can be used to affect the assignment of weighting factors the inventory management team uses to determine the weighted average forecasted demand.
[0212] Another factor considered in determining forecasts may be correlated demand. In this regard, some items have a demand that is correlated to the demand for other items. To determine the impact of correlated demand, the historical orders need to be analyzed to determine what items are bought on the same order as other items
[0213] More specifically, to evaluate correlation for a given item a list of correlated item pairs will be created based on what other items were bought on the same order as the target item. The number of occurrences of this pair of purchases will be tallied over a predetermined time period, e.g., a year. A system parameter will specify a threshold of occurrences that must be met in order to consider a correlation between an item pair. This threshold can either be an absolute amount (such as 10 occurrences) or a percentage of the total number of orders involving the target item (such as 60% of all orders of the target item involved also buying the correlated item). Any item pairs failing the threshold test are discarded.
[0214] If a correlation is determined to exist, the system then determines the percentage of total demand for the target item that is correlated demand. Product managers will have to provide input as to which of the two items in the pair is the primary (the item purchased independently) and secondary (the one purchased because the other was purchased). This information is then recorded in the product database. When customer speed and convenience demand is computed for the item during the forecasting cycle, it is computed in two parts—the part of demand that comes from primary demand and the part of demand that comes from secondary demand. In order to compute the secondary demand, the demand for the other item in the pair must first be computed. Then the correlation factor from the item database is applied to arrive at the correlated or secondary demand for the target item. The primary and secondary demands are added together to form the total forecasted speed and convenience demand. When the system determines and analyzes the demand profile for the target item, only primary demand is considered. The demand rate for the secondary demand will follow the demand profile of the other correlated item (which is the primary item of the pair). If the target item is the primary item of the pair, then its entire demand is considered primary and included in the demand profile analysis.
[0215] Another factor to be considered when determining demand forecast is promotions. In this regard, promotions need to be handled manually in the inventory management processes. The marketing team will design and schedule promotions and the inventory management team will work with the marketing team to determine the expected impact of the promotion campaign on demand. This impact must be manually entered into the system as a specific forecast that is considered in the weighted average forecasted demand. It is also necessary for the system to track the actual results of the promotion so that promotion-generated demand can be excluded from the historical database used in forecasting.
[0216] Often an M/M/M, M/M/G, G/M/M, etc. queuing models can be used to verify or confirm forecasted amounts
[0217] To assist in determining the probabilities that a customer will need a particular product, i.e., to perform a maintenance task on a piece of equipment or components of it facilities, the supplier spoke supports the Equipment Knowledge Base. The Equipment Knowledge Base is illustrated in context as element KB-3 in
[0218] Specific instances of equipment will be associated with generic types of equipment. The maintenance tasks for the specific equipment will reference their counterparts in the generic area and will specify replacement, deletion, insertion of the steps and parts, materials, etc. To create a maintenance task for a specific piece of equipment, the system will start with the associated generic data (if it exists) and replace, delete and insert details for the specific equipment as indicated.
[0219] Actual maintenance history data which includes the use of products will be kept on a specific equipment basis. Where a customer enrolls in the supply chain management program, their equipment and facilities inventory will be registered with the distributor. The Equipment Knowledge Base will include roll-up records to consolidate dates from multiple instances of the same specific equipment across customers and customer sites. Equipment and facilities can be comprised of hierarchically organized components. Maintenance history is generally compiled at the lowest level of hierarchical detail as appropriate.
[0220] Certain of the key features of the Equipment Knowledge Base are based on the Machinery Information Management Open System Alliance (“MIMOSA”) specification. The MIMOSA web site contains the details of this specification:
[0221] http://www.mimosa.org/. In this regard, MIMOSA has specified both data base design schema and data exchange definitions that facilitate the open electronic exchange of data for equipment characterization, operation, and maintenance. Where the customer's systems are not MIMOSA compliant, the intelligent software agents will translate back and forth between the MIMOSA standard and the native interface of the system of the customer.
[0222] With reference to
[0223] If a customer chooses not to include a specific piece of equipment in the program, the customer can still order products for the equipment/facility but will have to order the product directly (i.e., walk in order) or provide a probability of need in an advance demand notice order. A customer provided probability of need is obviously required in this instance since the system will not have the capability of determining the probability of need of a product not included in the Equipment Knowledge Base. Marketing and business policies, rules and procedures may also need to be developed on the fly when the system handles transactions where the customer's equipment is not included in the Equipment Knowledge Base.
[0224] For use in managing the maintenance activities at the customer, the Equipment Knowledge Base interacts with the CMMS legacy system
[0225] When a customer is first installed and registered in the supply chain management system, entries are made in the registry of the distributor as illustrated in the process flow chart of
[0226] When a customer first registers, the software agent of the distributor spoke will request the agents attached to the customer's systems to search their data services for information on equipment. For each piece of equipment found by the agents on the customer side, the agents will create a “new customer equipment event” performative and thus begin the processing depicted in
[0227] When the distributor spoke receives a “new customer equipment event” performative, the customer system agent may or may not have included an identifier for the site. Nevertheless, the customer side agent will have included the identification from the CMMS
[0228] Depending on how the customer has identified the equipment in the CMMS
[0229] Where the depiction of the process illustrated in
[0230] Failure/reliability data in the Equipment Knowledge Base will only be at a summary lever, detailed only to the degree needed to predict probability of need. This is likely to be some kind of metric time between failures. Detailed observation recording need not be kept by the distributor. This level of detail can be kept by the customer, either in the customer's CMMS
[0231] Some customers may choose not to keep any reliability data in their systems. In this case, the distributor and the customer will have to collaborate to attempt to learn what metric to use and what the metric value was the last time a particular part was used for the equipment. This will be the minimum data needed for the supply chain system to attempt a prediction of the probability of need for the part. In the absence of such data, the Equipment Knowledge Base will capture the first metric point when the first maintenance task is performed and work forward from that point. Also, in this case, the confidence factors will need to be set to zero until some basic reference point the metric is established. Once the maintenance data is established for the new equipment, the validation step with the customer will include passing the data on reliability to the customer's agent if the customer's system will also retain this information. The software agents at the customer side will take care of the translation between the reliability data forms and the metrics used by the Equipment Knowledge Base and the forms and metrics used by the customer's systems.
[0232] An equipment retirement is triggered when a customer removes a piece of equipment from its CMMS
[0233] When equipment is retired, the equipment asset records need not be deleted from the Equipment Knowledge Base. Instead, the record's status flag can be set. In accordance with the MIMOSA standards specification, flags can be used to indicate the following record status: 1=active row; 2=inactive row; and 3=soft delete row. In this case, the “customer retire equipment” event will cause the row status code to be set to “2” to indicate that the record is inactive. This enables all the historical data, especially the reliability data, to be retained and associated with the site roll-up record for all instances of that particular equipment. The intelligent agents which are predicting probability of need for an advance demand notice for other instances of the same equipment will likely need to refer to the historical data about the retired piece of equipment. The non-deletion of the record is also preferred since the piece of equipment may subsequently be installed at a different site of the customer or be sold to a different customer. In this case, where a “new customer equipment event” performative is processed, the maintenance programs will find the evidence record by matching the model number and serial number.
[0234] If a site roll-up asset record has all of its records set to the inactive status, then this indicates a situation where there are no longer any instances of equipment use. When this happens, the row status should be set to “3” to indicate a soft delete. At this point, the entire set of affiliated records can be subjected to some kind of records retention policy that indicates the period of time before the records will be physically deleted from the Equipment Knowledge Base. Part of the records retention policy should specify that the equipment manufacturers confirm that the equipment is truly outdated and no longer being used by any of its customers. When a physical deletion actually occurs, the records will be placed in an archives file.
[0235] To create in the Equipment Knowledge Base data about the reliability of the products used to maintain the customer's equipment, intelligent agents are used to capture maintenance data at the customer's site. This reliability data is used to predict the need for a specific product where some maintenance job is performed on the equipment. By predicting the probability of need, safety stocks through the supply chain can be better managed and minimized.
[0236] To predict the probablility of need, data on parts may be kept in the form of quantity needed per maintenance task within the Equipment Knowledge Base. For example, supply parts are typically not re-used in multiple maintenance tasks but are disposed of when the task is completed. While this quantity is often fixed, other times it may be represented by a curve, e.g., a bell curve. The data record for a supply part can thus contain the mean and standard deviation which can be computed from the maintenance history record for that supply part. When an advance demand notice cites a supply part, the Equipment Knowledge Base will return the probability for the cumulative distribution in the quantity on the advance demand notice.
[0237] Data on consumable parts may be kept in the form of quantity consumed in a period of time. The part usage history record in this instance will show the quantity used and the time metric for the part will be the time since the consumable was last replenished in the equipment. The data may be kept as a normal distribution with a mean and standard deviation. When an advance demand notice is received, it will cite the quantity of consumable expedited to be used and the time since the last replenishment.
[0238] Since both supply parts and consumable parts use the normal distribution, when the supply quantity or the consumable quantity per time equal their respective means, the probability of need will be 50%. The system will add 50% to this probability of need so that the supply chain probability of need will be 100% of the anticipated demand for the mean quantity or mean quantity per time for the product. In some cases, such as an order for tools, the standard deviation will be zero. In this case, the probability that the advance demand notice quantity will be needed will always be 100%. In most cases, when customers list tools and/or consumable on an advance demand notice order, the customer will be designating a 100% probability of need and what results from the Equipment Knowledge Base will generally not matter.
[0239] For repair parts, the history record may contain records of when specified repair parts are replaced in the equipment. The history record will show the time when the part was replaced and the age of the replaced part. If the user does not specify the age of the replaced part, the system can calculate the age from the previous history records unless this is the first record.
[0240] There are two ways that maintenance data can be captured and sent to the distributor for inclusion in the Equipment Knowledge Base (see
[0241] The maintenance staff can use a variety of methods and tools to create documentation on maintenance actions taken. These can range from totally non-technical paper-based documentation reporting that is manually entered into the system to state-of-the-art multimedia-based personal digital assistants or wearable computers that are RF-connected to system and allow maintenance staff personnel to voice dictate their maintenance curb as it progresses. Maintenance staff personnel can also use PCs which may or may not be connected to the customer's LAN while maintenance is being performed. In this situation, the maintenance staff personnel uses the laptop to record maintenance actions which are subsequently transferred from the PC to the legacy maintenance report system.
[0242] In order to ensure that the right data gets collected by maintenance staff personnel, scripts may be provided that solicit entry of necessary pieces of data. If the data is ultimately going through a CMMS system
[0243] When the intelligent agents extract the above-described maintenance data from the CMMS
[0244] The reconciliation process is an intelligent process that is used to reconcile things like purchase order dates being different from maintenance dates as indicated by maintenance records. Such differences can be reconciled looking for duplicate purchase orders. It is also possible that a product cited in a maintenance history documentation has not yet been included in an asset path bill of materials. In this case, a new asset path bill of materials record can be added for the specific equipment at this customer site and for the site equipment accumulation record.
[0245] Once the maintenance history record has been added to the Equipment Knowledge Base, the reliability/failure status needs to be recomputed for the individual piece of equipment and for the site record. Reliability states need not be computed for “tools” but only for “parts.” Usage status for consumables and supplies will become the basis for subsequent prediction of the probability and need when an advance demand notice order is processed, and ultimately provide a basis for replacing safety stock levels throughout the supply chain.
[0246] With reference to
[0247] The exemplary process begins when the customer's condition monitoring system detects a deterioration of the widget on equipment XYZ and determines that maintenance should be scheduled [CM
[0248] The AIP manager
[0249] The AIP for pre-maintenance planning parts sourcing directs a “planning” agent to collaborate with the CMMS legacy interface agent
[0250] During the planning process, several rule sets are extracted from the ontology
[0251] When the CMMS
[0252] The domain manager
[0253] The “trigger-event” performative is sent to the distributor's agent server
[0254] The distributor's AIP manager
[0255] The “reply” performative containing the distributor's confirmation number is received at the customer agent server
[0256] The broker
[0257] The distributor's “planning” agent which is working with the intelligent order fulfillment planning system
[0258] The customer receives the “evaluate” performative and opens up a new “equipment age/condition query” AIP instance. The agent complex determines that it does not know the exact age of the widget in question. However, a rule set in the ontology
[0259] When the distributor agent server
[0260] Since the distributor's ontology
[0261] When the ordered widgets have been staged according to the level of service specifications and the appropriate status message sent to the customer agent server and entered into the ontology
[0262] The maintenance worker begins the inspection portion of the maintenance task on the designated day [CM
[0263] The CMMS system
[0264] The distributor agent server
[0265] When the customer picks-up widget
[0266] Meanwhile, the distribution center picks, packs and ships widget
[0267] When widget 1 is delivered to the customer [C-S
[0268] While specific embodiments of the invention have been described in detail, it will be appreciated by those skilled in the art that various modifications and alternatives to those details could be developed in light of the overall teachings of the disclosure. For example, the term “intelligent agent” has been used in both the singular and plural senses throughout this disclosure. This usage is not intended to be limiting and, as such, when a task is recited as being performed by an intelligent agent or agents, it will be appreciated that the recited task may be performed by one or a collaboration of intelligent agents residing on one or more computer systems. Accordingly, the particular arrangements disclosed are meant to be illustrative only and not limiting as to the scope of the invention which is to be given the full breadth of the appended claims and any equivalents thereof.