Advanced Planning and Optimization (APO) extensions for particular industries are designed to work with SAP ERP Central Component (ECC) Industry Specific (IS) solutions. For example, many retail clients are using APO Forecasting and Replenishment (F&R) with IS Retail, and it’s common for users to work with APO Rapid Planning Matrix (RPM) with IS Auto. As the retail industry does not have any manufacturing operations, F&R does not have any Production Planning/Detailed Scheduling (PP/DS) component and mainly focuses on F&R between retail distribution centers (DC) and stores. Retailers often use F&R either with the IS Retail solution or with the traditional ECC solution. APO offers two extensions for the auto industry: Rapid Planning Matrix (RPM) and Multi Mix Scheduling (MMS). These extensions address planning and scheduling challenges faced by the auto industry, such as complex multi-level bill of material, line balancing, issues with MRP performance, and scheduling constraints. Auto companies often use RPM and MMS either with the IS Auto solution or with the traditional ECC solution. Table 1 provides more detail on the retail and auto extensions. For a table showing the major APO extensions, click here.    

Forecasting and Replenishment for the Retail Industry

    F&R, an APO extension for the retail industry, has a lot of similarities with DP and Supply Network Planning (SNP) applications. However, F&R offers additional capabilities/functionalities over the traditional APO solution:

• Many retailers are using F&R with the IS Retail solution at the back end. For APO, the most common back end is ERP Central Component.
• Unlike DP, demand-influencing factors (DIF) is a term used in F&R that indicates any past and future factors – such as price changes, promotions, holidays, and emergency situations – that affect regular demand.
• F&R does not have a capable-to-match engine like SNP.
• APO Safety Stock Planning functionality is not available in F&R. However, F&R allows modeling of specialized stock requirements for the retail industry, like presentation stock.
• F&R has a specialized replenishment algorithm for the retail industry.
• F&R has specialized analytics for the retail industry.

Following is more detail about the major solution components of F&R.  

Demand Planning Capabilities of F&R

The forecasting tool of F&R supports statistical forecasting (specific models that consider trends or seasonality), causal modeling, promotion planning, and consensus planning. This component is similar to the DP solution. F&R statistical forecasting should be used for products that are replenished on a regular basis from external and/or internal vendors because products must have a measurable and repeated sales history for statistical forecasting to work.  

Calculating and Optimizing Replenishment Quantity

  F&R helps in calculating the replenishment quantity of a product for each store or distribution center. There can be different models for replenishment calculations. In one model, quantity calculations can be based on the forecast/orders of the distribution center. In another model, such replenishment calculations can be based on aggregated store forecast/orders, when store-level forecasts get rolled up to the level of the supplying distribution center. This process is known as multi-echelon replenishment. Requirement quantity calculation in F&R uses almost the same logic as SNP Heuristics. The gross requirement for a period (week or month) is calculated based on future demand projections for the period and target/presentation stock requirements at the end of the period. The net requirement is calculated by deducting on-hand stock/stock projections from the gross requirement. The requirement quantity calculation also takes into account order and delivery cycles and corresponding lead time. As discussed earlier, a good part of this logic is the same as SNP Heuristics, except for a few specialized requirements of the retail industry. Following a different strategy for replenishment can affect the quantity calculation. For example, a strategy of pre-allocation enables retailers to build up stock earlier than when it is required, which can be useful for situations such as a Christmas promotion, when a large quantity of stock is required over a short  period of time. After the net requirement calculation, requirement quantity optimization is a step in which net requirements are combined with order proposals for a particular supplier and optimized according to different rules. These rules can include the following:

• Pack size rounding
• Matching order proposals against vendor minimums
• Truck load building
• Economic Order Quantity (Ordering costs are balanced with the cost of capital for inventory)

After optimized order proposals are created in F&R, they must be released (automatically or manually) to the purchasing solution. There can be different release strategies, such as maximum or minimum rules based on specific values, quantities based on which order proposal release for a particular item can happen, and whether proposals must be released manually or automatically. Beyond automatic replenishment quantity calculations, F&R also has a replenishment workbench to support manual replenishment for the replenishment planner/analyst, who can use this to create, alter, review, group or split order proposals for items based on criteria such as products with an overdue delivery date or products with unexpected sales. This workbench is useful for managing exceptional situations. Cause determination helps with in-depth analysis of all information for a specific location or product.  

F&R Analytics

  F&R provides a set of standard content in SAP Business Intelligence Warehouse (BW),  and this enables a set of readily built queries, including history of manually changed order proposals, forecast quality, range of coverage, overstocks, understocks, stock outs and undelivered products, stock outs and lost sales, number of stock exceptions, replenishment exceptions, and dead stock report.  

F&R Data – Master Data, Transactional Data, Parameters, Profiles

 

Like APO, F&R also needs a set of master and transaction data to operate. Typical master data that is relevant here include products (in this case, retail articles), locations (in this case, retail stores, distribution centers or supplier locations), scheduling information, and product hierarchies. As there is no production activity in retail, production related master data like BOM, Routing, and Work Center are not needed here. Typical transaction data for F&R include sales/goods issue data, on-hand stock positions, and orders.

Beyond master and transaction data, F&R needs a set of planning parameters to operate. Parameters either can be sent from an external system by interface or manually maintained within F&R. Specific parameters can be maintained in a time-phased manner that allows the system to consider new parameter values when a specific point in time is reached. For easier maintenance, parameters sometimes are grouped in a profile within one specific process area. For example, a “forecasting profile” bundles parameters regarding the forecast calculation. Profiles need to be maintained only once and then can be assigned to all suitable master data elements. For easier maintenance of profiles, these can be mass maintained within F&R.  

F&R Supports Retail Industry CPFR Standards

  Collaborative Planning Forecasting and Replenishment (CPFR) standards are supported by F&R, which supports this process by creation of both a demand and an order forecast, and by allowing the retailer to share this data with defined vendors. This is enabled by a specially created function module which can be called using Remote Function Call.  

Case Study

  One of the largest home furnishing retailers in North America implemented F&R for forecasting article requirements and replenishing its stores. The solution was initially deployed for 168 stores, six distribution centers, 30,000 SKUs, and 1,500 vendors. The solution went live with 7,000 to 14,000 purchase orders daily (50 to 100 order proposals per store). For the stores, the F&R application was integrated with a store point-of-sale solution. A replenishment workbench of F&R used by nearly 40 central planners was the main replenishment engine for the retailer, doing all quantity calculations according to 20 different business rules. This was a multi-lingual implementation (English and French) on SCM version 5.0, on DB2 database and tight integration with SAP ERP. A replenishment run for 100-plus stores takes around two hours. The project had unique analytics requirements in terms of in-stock position, inventory turns, late POs, and auto versus manual order proposals. The solution helped the retailer to provide optimal inventory levels at stores, reduce lost sales at the store, reduce the number of undelivered products at the distribution centers, and improve its forecast.  

Rapid Planning Matrix

  RPM is a different technique for bill of material (BOM) explosion in the liveCache to allow the APO to process large amounts of data much faster. This was developed for the auto industry but subsequently used in many other industries that have similar requirements. This works with Integrated Product and Process Engineering (IPPE) and is a useful functionality for certain business requirements, as described below.  

Situations in which RPM Can Be Useful

 

• When dealing with high-volume data
• When the product structure is complex
• When very frequent and very fast material requirements planning is needed
• When there are innumerable individual customer orders with configuration – that is, there are innumerable variations of the end product.

This form of planning is recommended for make-to-order repetitive manufacturing, in which large numbers of models are used, as in automotive, high-tech and consumer durable companies.  

Advantages

  RPM offers several advantages over conventional Material Requirements Planning (MRP). RPM plans the requirement quantities and dates for components of a product very quickly. The planning is single level. Single-level explosion of the IPPE data consists of the product variant structure, line design, and the process structure. The subordinate BOM levels are planned in APO using PP/DS. Another reason for the RPM’s speed advantage is that the planning matrix for the RPM product is generated in the liveCache for the APO. The liveCache contains the data relevant to planning, so time-consuming database accesses are no longer necessary. The RPM algorithm stores data in optimized form in Random Access Memory (RAM) in the liveCache main memory database, instead of on disk. The RPM uses matrix form for data storage and allows simultaneous BOM explosion of all orders, instead of order by order. In addition, the valid BOM item is determined simultaneously for all orders. RPM helps to improve customer satisfaction by offering fast reaction to order changes. In the trucking industry, it is common for a customer to order a truck early and make last-minute changes to its configuration. RPM uses the order start date and the master data that is maintained in IPPE to determine the exact requirement dates for the components. RPM automatically changes the requirement dates of the components to match any changes to the start date that are caused by sequence changes in model mix planning. There are several downstream functions that use data from RPM:

• Determination of the component requirements for each order for further planning in the APO and MRP in the ECC system
• Generation of order BOMs for backflush at reporting points
• Transfer of the characteristic value matrices as the basis for sequencing
• Forecast delivery schedule/Just-in-time (JIT) delivery schedule/sequence JIT call.

 

RPM Structure

  The advantage of RPM comes mainly from its structure. RPM consists of data from two matrices:

• Characteristic value matrix. The characteristic value matrix represents a method of storing the object dependencies for the individual orders in the liveCache. The rows of the RPM contain all the components of a product that the system determines from the component variants of the product variant structure (PVS) for the product, so the rows of the characteristic value matrix contain the maximum number of possible characteristics. The columns contain the lined-up APO planned orders for the product – that is, the configurations from the orders are in the columns.
• Component variant matrix and component validity. The component variant matrix is used to determine component requirements. The system determines all the components possible for a product (super BOM) and the quantities and writes them in the matrix rows. All the components for the product and the order data are in the component variant matrix. The system writes the APO planned orders for the product in the columns. The system uses the object dependencies and time-based validity to determine the components that are necessary for the order and puts an X at the corresponding component. The use of a particular component in an APO order is indicated by an X in the appropriate row of the rapid planning matrix.

The process of making this RPM involves the following three steps:

• Convert the assembly view of the PVS into a flat list of the master data: material number, quantity, selection condition, effectivity parameters
• Include the assembly orders: due date, quantity, configuration
• Explode the BOM for the orders. As indicated earlier, a cross-means component selected for order.

 

Case Study 1

  A big-three auto manufacturer implemented an RPM solution for its manufacturing facility in Australia. Traditionally, the company was using MRP for its planning. It had a complex BOM structure and its maximum BOM used to have up to 40,000 components, including all historical changes. The company handled up to 30,000 customer orders regularly. This required a huge amount of storage and up to 60 Gigabyte dependent requirements in ECC MRP in database in relational form. Explosion of this maximum BOM order by order means more than one billion checks of effectivity parameters and selection conditions. This caused huge performance issues in the company’s MRP run – taking a very long time and causing frequent terminations of MRP runs. This performance problem justified an investment in a solution such as RPM, which keeps data in an optimized form in the liveCache main memory database, stores data in matrix form, and does simultaneous BOM explosion for all orders. The RPM solution significantly reduced the BOM explosion time, to less than an hour.  

Case Study 2

  An automobile company was manufacturing two of its highest-selling car models. The size of maximum BOM for two variants of the model was 70,000 and 12,000 component variants. There were up to 23,794 individual orders for car 1 and 19,495 individual orders for car 2. The planning needed to manage complex selection conditions. The RPM method completed the complex BOM explosion in only 19 minutes, and the complete planning run in 48 minutes.  

Model Mix Planning and Scheduling

 

MMPS is a planning and scheduling technique for configurable products with a high volume of orders. Like RPM, this solution was initially developed for the auto industry, but today is used in many other production scenarios in which different products are produced together on the same production/assembly line.

MMPS helps create a production schedule in the medium- to long-term planning horizon, taking into account delivery dates, available capacities, and any existing restrictions. MMPS can plan individual lines or a complete line network. A line network is a series of lines one after the other, or parallel lines that can be used as alternatives. MMPS uses master data of the IPPE and the PP/DS Optimizer to determine the best solution.

 

Handling Constraints

 

MMPS can handle a variety of constraints or restrictions. Restrictions can be grouped in two categories: Hard and soft. Hard restrictions should not be violated at all, and soft restrictions should be respected to the extent possible. For an automotive company, the following can be examples of such restrictions:

• Particular types of cars cannot be combined in the same sequence. For example, no two white cars in a sequence
• Particular types of cars must be combined in the same sequence. Example: At least two red cars in a sequence
• Quantity (position) restrictions, such as the total number of a particular engine. For example, cars with V8 engines <= 100 pieces per day
• Spacing restrictions. For example, every third engine should be the U.S. version of left-hand drive.
• K in M restrictions. For example, every three vehicles out of five should be fitted with air conditioners.
• Block restrictions. For example, at least two engines for right-hand drive vehicles in a row.

MMPS helps calculate times for the preceding and following lines, and the exact start and end times for the order sequence. This order sequence takes into account restrictions and customer-required dates, and fulfills business requirements like an equal load of the line segments, a minimization of restriction violations, and a minimization of the absolute schedule deviation.