The consumer goods industry has many industry-specific characteristics that present planning challenges. These characteristics include proliferation of stock keeping units (SKUs), promotions, complex disaggregation requirements, shelf life of items, contract manufacturing, and direct store delivery. Leading companies in this industry have designed innovative solutions to overcome 10 common planning challenges.
(Note: Consumer product companies are broadly segmented into two major categories: food and non-food products. Each has several sub-industry verticals. For example, food can be divided into dairy, beverage, packaged foods, or agricultural products. The non-food category also has several sub-verticals, such as tobacco, personal care, household items, consumer electronics, apparel, and footwear. Most of the challenges I describe are common to all sub-verticals in the consumer goods industry. There can be specific planning challenges for particular sub-industry verticals such as milk planning and allocation in the dairy industry or returnable item planning for the beverage industry. The scope of this article does not include sub-industry-specific planning challenges.)
1. Proliferation of SKUs
Using many SKUs is common for the consumer goods industry. However, at any point in time, only one set of SKUs can be active. The proliferation results in a number of planning challenges:
- A need for planning at the product group level instead of for end products. There are too many planning runs if all planning happens at the end product level.
- Lot size issues for low-volume SKUs. It’s common to outsource manufacturing in this industry, and low-volume SKUs create lot size issues in manufacturing. Manufacturing always needs a minimum lot size to manufacture a product. For example, the product can be produced in lots of 10,000. If production is taken in lots of fewer than 10,000, then the setup cost increases. This is not economical. For low-volume SKUs, for example, if the product sells 1,000 per month and is produced in a lot of 10,000, then 10 months’ worth of inventory needs to be maintained, thereby increasing inventory carrying costs.
One of the largest tobacco manufacturers in the world had challenges in terms of too many SKUs under each product group while implementing SAP Advanced Planning and Optimization (APO) Demand Planning (DP). Using a large number of SKUs creates challenges in terms of forecasting, disaggregating, stock planning, and introducing new products. APO DP helps consumer goods companies in terms of forecasting at different levels to get higher forecast accuracy, measuring forecast accuracy through APO alerts, and better new product planning using APO life cycle planning functionality. Part of the reason for so many SKUs was innovation to keep the tobacco manufacturer’s products relevant. The rest was outside its control, because, among other issues:
- Tobacco is a highly regulated business. There are frequent directions from governments to put up new warnings or graphics on the packages, and therefore, packaging specifications change.
- There are changes in excise duty, leading to a product code change along with price. (The company mentioned in this case study has global operations. It has operations in countries in which the excise duty changes twice a year and also operates in countries in which excise duty changes happen once a year.)
- Every promotional SKU needs to have a separate code, although in many cases the product is fundamentally the same and only the price is different.
The company had the unique experience of seeing that none of its final SKUs were valid for more than four to six months, although in the brand level, it had been selling almost 80 percent of the same brands for the last 10 years.
The company had used APO DP statistical forecasting tools for forecasting end product demand. For large selling SKUs, the forecasting was done at the individual SKU level, whereas for small selling SKUs, forecasting was done at the product group level to achieve higher forecast accuracy. The company had developed a phase-in, phase-out planning book in APO DP where with the help of a set of DP macros, the company modeled stock planning requirements for new products to be introduced and for existing products that would be withdrawn in next few months.
Promotion is the lifeblood for the consumer goods industry. Although promotions for some sub-verticals such as tobacco and brewery are highly regulated, promotions for most other categories, including packaged food, personnel care, and household items, are not. Promotions may cause planning challenges in several ways:
- Estimation of promotion volume: This is mostly a manual process. No planning tool can calculate promotion quantity and it is generally the gut feeling of the demand planner or the salesperson in the field.
- A number of promotions active at the same time: It is common for promotions to be active at the retail, regional, or national levels at the same time for the same SKU. A good example of this can be a company that is promoting a particular detergent brand in national media and running some sales promotion for the same product at particular stores with both campaigns happening at the same time. This can cause an issue in terms of determining the effect of each of these types of promotions on the SKU volume. In the given example, if there is an increase in sales for the detergent brand at the month end, it’s difficult to tell how much of this increase happened from promotion in the national media and how much of it happened because of promotion at the store level. SAP APO offers a functionality called promotion base that helps in segregating the effect of such promotions on the same product.
- Difficult to clean promotion: Cleaning promotion from actual sales is important to get a baseline if tools such as statistical forecasting are used. Cleaning promotion means taking out promotion from actual historical sales. This is required to get the sales history baseline as statistical forecasting tools use this baseline for forecasting future sales. Once the baseline is forecast, future promotions are added to this. Cleaning promotion is always a challenge as it is difficult to exactly estimate its effect on past sales.
- Managing active SKUs: Promotion can cause a certain number of SKUs to be active under a particular product group at any point. This causes planning challenges in terms of disaggregating planning results at the product group level to the right set of SKUs.
- Promotion can involve a number of applications: Consumer goods companies can use a number of tools for managing promotion — for example, a trade promotion tool, a CRM tool for creating promotion, third-party promotion tools, and APO for planning promotion volumes. In certain cases, this necessitates the integration of promotion planning tools with other applications.
One of the largest consumer goods companies selling packaged food, personal care, and tobacco products saw that almost 25 percent of its product portfolio was continuously on promotion. This caused several planning challenges in terms of a good baseline history or forecast and stock planning.
The company designed a unique solution with its implementation partner to address this and came up with the concept of market SKU and system SKU. A system SKU is the actual SKU that is shipped to the customer, and these keep on changing throughout the year. A market SKU is a product group, the level at which the product is known in the market. All planning and forecasting are done at this level; however, this may not be the level at which actual shipment happens. For example, in this case the company’s product is a 100-gram glucose biscuit. This is a market SKU at which all planning and forecasting are done.
However, many variations of this product are available throughout the year (for example, a 100-gram glucose biscuit with a 25-gram extra glucose biscuit free in the pack or a buy one, get one free offer such as a 100-gram glucose biscuit sold along with a 50-gram salted biscuit). These variants are system SKUs at which the product actually is sold throughout the year. These system SKUs keep changing due to different promotions at different times. Doing any forecasting at the system SKU level results in forecast inaccuracy because of the short life cycle of these SKUs, which may be available only for a few months of the year.
By contrast, planning and forecasting at the market SKU level is much easier because these SKUs are steady throughout the year. It is easier to get a good history of past sales and hence relatively easier to do a forecast with good accuracy. The company designed a process to disaggregate the forecast at market SKU level to system SKU.
A complex product and location hierarchy for consumer goods companies necessitates multiple levels of disaggregation. A brand may have multiple product groups under it. Each product group has multiple products, and each product may have many SKUs.
For example, the Colgate brand can have two product groups: toothpaste and toothbrushes. One of the product groups, toothpaste, can have many products such as white, gel, variety, or peppermint, and each of the products (for example, the white toothpaste) can have multiple SKUs in the form of 50 grams, 100 grams, or 200 grams.
If the planning or forecasting is done at a higher level (in this case at the product group toothpaste level) for better forecast accuracy, then these forecast numbers need to be split up to the lowest individual SKU level. This process is called disaggregation. As discussed earlier, a proliferation of SKUs leads to a multilevel product hierarchy for companies in the consumer goods industry. Although having a wide distribution network is considered a strength for competing companies in this industry, it also brings challenges in terms of complex multilevel location hierarchies. Multilevel product and location hierarchies can cause the following challenges in disaggregation:
- Small-selling SKUs and rounding effect: Every consumer goods company eventually has some small-selling SKUs in its portfolio. Multiple levels of product and location hierarchies-based disaggregation have a poor effect on these SKUs. For example, if a company does a statistical forecasting run at the region and product group levels, the values now need to be disaggregated at the individual distribution center and SKU level. For small-selling SKUs this multiple level of disaggregation can result in several rounding effects at the lower level and distorted values resulting in poor forecast accuracy.
- Time-based disaggregation: In addition to product and location hierarchy-based disaggregation, time-based disaggregation is also a requirement from several consumer goods companies as the month- or quarter-end effect is very high in this industry. Up to 40 percent of the actual sales can happen in the last week of the month or quarter, and these sales are generally driven by a push from the sales team to achieve its sales quota or incentive targets.
- Disaggregation needs to be on the right active SKUs: As discussed earlier, it’s common to have multiple SKUs available under a product group with each SKU having a defined validity period. For example, several promotional SKUs available under a product group may be valid only for a month or for a few weeks. Promotional SKUs are valid for a very short time — for example, for a month. A product group is valid for a longer period. Disaggregation needs to ensure that values at the higher level are disaggregated only on active SKUs during a particular time bucket. The proportional factor calculation needs to take this into account.
- The proportional factor calculation rules can vary per product group: Different product groups for a consumer goods company can have different proportional factor calculation logic — for example, for one product group the company might want to calculate the proportional factor for disaggregation based on last six months, whereas another might be based on the last 12 months.
A planner’s manual intervention for future disaggregation can overrule system-based disaggregation. This is a requirement at several companies. Although the system can calculate the proportional factors based on a defined business rule (such as based on the historical proportion of the last six or 12 months), the planner can always change the forecast numbers for an SKU based on market reality. The system then calculates the new disaggregation factors based on the changed proportion and populates all future buckets. Planners should be able to change the proportional factor values that are populated in the future buckets at the beginning of the next month. This is to factor in market reality as planners would know it better. The system is guided by a static rule defined in the system.
One of the largest consumer goods companies had several challenges with disaggregation while rolling out DP globally. The company had close to 40 percent of its SKUs in C class with a very small monthly volume. For getting a good statistical forecast, the company decided to do forecasting at the regional level. However, this led to a number of disaggregations, first from its regional to branch offices and from there to individual distribution centers, resulting in rounding effects and even zero values for certain SKUs.
(Note: C Class is a product classification. The largest selling SKUs are in A class, the next level is B class, and the smallest selling SKUs are in C class. Typically, C Class SKUs are many in number, but they together contribute less than 20 percent of a company’s total sales.)
The company was traditionally a tobacco company and had recently entered into the packaged food business. The tobacco business was steady, but food was growing rapidly. The company wanted to build business rules for proportional factor calculations specific to each category. It wanted the tobacco proportional factor to be calculated based on the last 12 months, the packaged food category to be based on the last six months, and the ready-to-eat category to be based on the last three months. It believed that data older than the last three to six months did not make sense for the food business.
In every planning run, the system populates planning results for the next few months, while disaggregating the values at a lower level. In a planning run, the system can disaggregate based on the proportional factor calculation for that category. However, if planners made a manual change in the past for any category on disaggregated values, the disaggregation would happen based on the planner’s revised proportion for current and future time buckets. Disaggregation of the split weeks (when a week is split between two months) needs to be based on selling days available in each month.
In this project, a number of additional key figures were used to store and populate different proportional factor values. A set of macros was created to meet the requirements of different business rules.
4. Managing the Product Life Cycle
Frequent new product introduction and retiring of existing products are at the core of the consumer goods industry. Rising competition, changes in regulations, and frequent changes in packaging specifications necessitate the regular introduction or discontinuation of products. Managing this product life cycle can pose several challenges from a planning perspective:
- Forecasting of new products: Most planning tools forecast based on history, and new products do not have a reasonable history. Planning tools provide like modeling, a method of forecasting new products based on the history of similar products. However, the challenge is always to get a similar product whose history can be mimicked.
- Stock planning of new products and products to be discontinued: New products need to be introduced in a phased manner, and some of the existing products need to be phased out gradually. Inventory planning becomes complex in such cases.
A large consumer goods company specializing in personal care products planned to use an APO solution for managing its supply chain. One of the challenges the company faced was how to forecast for new products as there was not enough history. After analysis, the company identified that its new products fell into three different categories: The first category is those that are a different variant of one of its products already existing in its product portfolio. The second category is those products that the company is introducing into the market in response to its competitor’s action (i.e., a similar product is available in the market from its competitor, but the company does not have an offering in the same category). The third category is the innovative product category (i.e., the products that are completely new to the market).
The company adopted three different strategies for planning and forecasting these three categories. For the first category, it used the like modeling approach in which an existing similar product variant’s historical sales were taken as the base for forecasting demand for the new variant.
For the second category, the company used market data from AC Nielsen to get actual and projected sales data of its competitive offering. This data was loaded into APO, and a like modeling tool was used to estimate initial demand for its new products in this category. The initial estimate was further modified after a consensus planning process in which effects of future changes in market share or effects on demand of the company’s existing product portfolio were factored into the estimate.
For the third category, the company planned to forecast these products for the first year manually. The forecast files were created in Excel, and the same files were uploaded in APO DP for the first 12 months. Once the system had actual sales data for one year for these products, the company planned to use statistical forecasting tools available in the solution to forecast demand.
5. Managing Item Shelf Life
Certain consumer food product categories such as personal care products, household products, or packaged foods have a long shelf life. Shelf life is a requirement mostly for food companies and specifically for certain food sub-industry verticals such as dairy. For these products, shelf life needs to be considered for planning stock for the entire supply chain network. Managing shelf life causes several planning challenges:
- Item shelf life is not considered by certain planning tools such as the Optimizer.
- Shelf-life constraints need to be considered in all other planning runs.
- Planners need to be alerted in advance about which products are expiring.
A leading packaged food company had challenges in terms of managing shelf life of its items. Being in the food business, the company is monitored by many government agencies such as the FDA for compliance and needs to ensure that none of its items are available in stores after the product’s expiration dates have passed.
The company used APO shelf life functionality to model the following shelf life scenarios:
- Stock needs to be projected considering the stop sell date (i.e., if any product has expired, it should not appear in a stock figure). A stop sell date is defined as a date beyond which the product is not considered sellable to customers (for example, hospitals or pharmacies). However, under extreme situations, the product can be sold based on an agreement with the customer. This constraint is relevant at the market demand chain level.
- Imported products need to obey a minimum remaining shelf life constraint due to import restrictions setup by countries (for example, a product cannot be sold to country markets if the remaining shelf life doesn’t fulfill import restrictions). This constraint is relevant at the central warehouse level.
- Monitoring shelf life of raw materials also during planning (i.e., shelf life planning) needs to extend across bill of material (BOM) levels.
The company had used APO shelf life planning extensively during planning its supply chain network. The minimum remaining shelf life of products was considered during replenishment planning at market distribution centers and regional warehouses. APO provides project shrinkage that is expected to be generated in every time bucket across the network because of shelf life issues. Shelf life is also propagated across BOMs to ensure that any of the raw materials nearing their expiration dates are not being used in the manufacturing of the finished products.
6. Truck Load Building (TLB) and Route Optimization
Reducing transport costs is a top priority for consumer goods companies. Ensuring that transport vehicles are fully used and that they are following the most optimal route is a way to achieve lower transport costs. Consumer goods companies face several challenges in TLB, vehicle space optimization, and route optimization:
- TLB is a manual process in APO. TLB is also a functionality in APO Supply Network Planning (SNP). Loads can be combined based on truck capacity obeying volume and weight constraints. However, selection of loads to be combined in a truck is still a manual selection process, and for years, consumer goods companies expected a more robust solution here. SAP recently came up with a strong offering in terms of Load Builder. Load Builder is part of Transportation Management (TM) 9.1. The solution offers load-building functionality that supports a variety of load-building rules. The strong visualization capability of the tool allows you to see how the loads are actually being placed in a vehicle.
- Route optimization: This is to ensure that the vehicles are following the lowest distance path on a regular basis while delivering material from the factory to the warehouse to the distribution center and eventually to key customers and retailers. Route optimization poses several challenges to consumer goods companies in terms of maintaining a huge route master with their distance and costs for transportation, maintaining all alternate routes in the system, keeping this database updated, doing a daily optimization run, and analyzing the results. Many consumer goods companies use APO Transport Planning and Vehicle Scheduling solution for route optimization, which helps them to model a variety of route and vehicle constraints.
A large consumer goods company implementing APO Transportation Planning/Vehicle Scheduling (APO TPVS) for transportation planning and route optimization had a complex requirement in vehicle space optimization. It was delivering a mix of packaged food products (noodles, spices, salt, wheat, biscuits), tobacco products, and personal care products (soap, shampoo) in the same vehicle to its distributor. There are several rules on how these products need to be placed inside the vehicle. For example, a salt bag cannot be placed on top of a tobacco pack, and some products should not travel together.
The company also wanted the load in the vehicle to be built up in a way that considered the drop sequence of the load in the route. The vehicle was delivering to multiple distributors or key accounts on its route every day. To facilitate the handling process, the load that needed to be delivered to the first distributor on the route needed to be kept at the rear end of the vehicle, then the load for the next distributor, and so on.
Although the company could model many of its route optimization requirements with APO TPVS, there were certain gaps in vehicle space optimization and pallet building requirements. These were partially addressed through developments and for a few of these requirements, the company continued the process manually in Excel. The company had recently migrated from APO TVS to TM 9.1. TM 9.1 load builder functionality helped the company to plan the loads in a way that used maximum space in the vehicle and positioned the loads in the truck in a pattern that corresponds to the route drop sequence.
7. Subcontracting and Contract Manufacturing
Many consumer goods companies outsource their core manufacturing operations and focus on things of higher importance to them such as managing the brand and marketing. Beverage companies have outsourced their bottling operations for years. Contract manufacturing can take different versions in the form of complete manufacturing outsourcing, subcontracting a few operations, or subcontracting while the parent company provides all the raw materials. Contract manufacturing and subcontracting can cause the following planning challenges:
- No planning but control: Consumer goods companies do not want to do detailed Production Planning-Detailed Scheduling (PP-DS) for subcontractors, but want to have control of the process. This means that supplier capacities are not a part of the planning system. However, companies need daily feedback on how their vendors are performing against the plan, as this affects their shipment plan.
- Collaboration is more important than planning: Consumer goods companies understand the need for collaboration with their contract manufacturers, which can account for the popularity of Supply Network Collaboration (SNC).
- Subcontractors may not have sophisticated planning systems: Although some subcontractors have ERP or EDI systems in place, in most cases, they have only Internet connectivity and basic IT systems. This calls for building simple planning and scheduling system integration with them.
A large packaged food manufacturer uses contract manufacturing extensively for producing most of its products. The company wanted visibility and proactive monitoring of all its outsource manufacturing. To make it a reality, the company used a mix of technology offerings from SAP, such as APO and SNC. Although APO Heuristics was used for BOM explosion and to create purchase requisitions or purchase orders for contract manufacturers, SNC work order collaboration helped in monitoring and controlling work orders of suppliers. The work order is based on a purchase order. The contract manufacturer regularly sends data about the progress of the production process in the work order (component quantities used, work in progress [WIP], scrap, or completed final product quantities). The company can then react in sufficient time to potential exception situations (for example, if components are missing or a low yield is expected).
8. Forecast Accuracy
Consumer goods companies primarily operate in a make-to-stock (MTS) environment. This makes forecasting for finished goods of prime importance. Many companies in this industry are gradually moving toward a replenishment model (i.e., they replenish the inventory of their distributors or retailers), but this does not make forecasting less important. Consumer goods companies still need to forecast for their production planning. Several companies have implemented DP and statistical forecasting to improve their forecast accuracy. However, getting good forecast accuracy presents companies with the following challenges:
- Getting a good baseline history is always a challenge for certain SKUs: In some cases this boils down to the issue of history cleaning (i.e., history could not be cleaned as desired). A baseline means normal historical sales of a product. Generally, companies run a lot of promotions to increase sales for consumer goods. The effect of all such promotions needs to be taken out from historical sales to get a good baseline, so history cleaning becomes important. A good baseline is the prerequisite to run statistical forecasting process in APO. For other cases it may be the history itself that is not showing any clear trend, and therefore, it is difficult to baseline. Without a good baseline, statistical forecast tools cannot generate a good forecast.
- Promotion cannot be forecast: Promotion is a major part of sales for consumer goods companies, and it cannot be forecast by any statistical tool. The accuracy of the final outcome depends on how well promotion is estimated.
- Selecting the right level and tool for forecasting: Forecasting accuracy depends a lot on at what level the forecasting is done and which forecasting model is used. Forecasting at a higher level increases the accuracy, but creates challenges because of multiple levels of disaggregation. Forecasting at a lower level decreases accuracy and creates the issue of too many forecast runs. Different forecast models can create very different forecast numbers. Therefore, selecting the right level and model is important for getting good forecast accuracy.
- Forecasting for products not having sufficient history and new products: Statistical forecasting and forecast accuracy are always a challenge as the tool needs a minimum of two to three years of history. In today’s dynamic business, getting a good baseline based on two to three years of history for all SKUs can be a challenge.
A large consumer goods company with the objective of achieving higher forecast accuracy started a DP and statistical forecasting project. The company had the usual challenges of forecasting in terms of estimating demand for new products and planning promotion. However, a few of the following initiatives helped the company to achieve high forecast accuracy.
The company developed a 10-step history cleaning methodology and trained planners in that. This helped it to get a reasonably good baseline history. The company had put a separate measure on statistical forecasting accuracy along with overall forecast accuracy. This helped the company to monitor how good or bad its statistical forecast is (before planners do any changes to it) against the actual sales and to put an effort into improving it.
The company designed a simpler simulation planning book that helps planners to do simulation if they see that the forecast result is not as good as expected. Instead of changing the forecast itself manually, adjusting the model was a better approach and helped in improving accuracy over time.
The company made a learning log to analyze the effects of each promotion type, which helped it to clean the past history and to provide planners with a guideline on expected uplift if the same promotion is repeated in the future.
9. Supply Chain Optimization
Optimizing the end-to-end supply chain in terms of cost and customer service is the vision of many consumer goods companies. In practice, the optimization of the entire network may not be a feasible approach, and many consumer good companies opt for optimizing just part of their supply chain networks. Although APO SNP Optimizer is most commonly used across consumer package goods companies, a few of them are also using the PP-DS Optimizer for production scheduling and the TPVS Optimizer for vehicle scheduling and route planning. Deployment optimizers are rarely used by companies in this industry. Typical scenarios in which consumer goods companies choose supply chain optimization solutions are:
- Distribution optimization between factories and warehouses or distribution centers: The focus here is more on transportation cost optimization. On a monthly or weekly basis, a company determines the quantity of materials and decides on the factory that will supply a distribution center or warehouse. The Optimizer in that case runs for a part of the network between warehouse or distribution centers and factories. SNP Optimizers are commonly used for this process.
- Optimization of factory production plan or a factory program: The Optimizer runs on a weekly or monthly basis and decides the production plan at a factory based on distribution demand and cost of production at the particular factory. SNP or PP-DS Optimizers are used for this process.
- Scheduling production lines: This identifies the best schedule for production lines based on setup cost optimization. PP-DS Optimizers are run at the individual factory level.
- Optimization of route and vehicle schedules: This selects the best route and proposes the optimum transport plan and vehicle schedule to make a delivery. TPVS Optimizers are used for this and a typical optimization run is daily.
Supply chain optimization is a need for many consumer goods companies, but companies face a set of challenges while using optimizers:
- Developing a cost model and keeping it updated: Developing part of a cost model is easy (such as transportation or production cost), but some other cost parameters are difficult to ascertain (such as storage or handling costs). Sometimes a company models a cost parameter in an optimizer depending on how it wants the solution to behave. For example, you could increase the priority of a product or customer order to ensure that it is manufactured or delivered first. Another challenge is to keep it updated on an ongoing basis.
- Optimizer runtime: Sometimes the monthly optimization run takes several hours. Several technical solutions are possible to fine-tune this runtime such as using product or time decomposition functionality in the APO Optimizer (that reduces the number of buckets or percent of products that are simultaneously planned) or breaking the Optimizer run into several batch jobs, each of which does optimization for a part of the network.
- How many constraints to model: Modeling too many constraints can increase the problem size to such an extent that it affects performance for the optimizer. Although the number of constraints modeled needs to be representative, if the Optimizer is run for a large network it’s important to model the most critical constraints or break the optimization problem into pieces.
- Analysis of Optimizer result: Several planners from consumer goods companies are facing challenges in terms of interpreting the planning result from the Optimizer (i.e., why it had made a particular decision). This process requires a good amount of training for end users or the Optimizer remains a black box for them.
Modeling business rules: Several consumer goods companies have issues here. The vision is to optimize the global supply chain end to end, but very few global business rules actually drive the supply chain. Rules are often local (i.e., specific to a particular country/business) or are planner specific. It is important that during the business blueprinting phase, you identify these rules early enough and can distinguish which of these rules are local and which ones are global. As an example, one company found its production scheduling rules for optimizing the machine setup of a particular factory are local (specific to that factory), while rules of SNP deployment optimization (i.e., rules for replenishment of stocks for the customer) are global. Ideally, modeling in an optimizer should start once these rules are identified, segregated (local or global), and documented.
A large consumer goods company based in India implemented the SNP Optimizer for its seven manufacturing locations and more than 50 warehouses. The Optimizer runs monthly and decides on the quota (i.e., which warehouse needs to pick up how much stock each month from which factory). The monthly Optimizer run also decides a monthly factory production plan (known as the factory program) based on factory capacity and constraints in the supply chain. The Optimizer runs for the next three months and every time the values of the last run are overwritten.
The company had built up the cost model using transportation cost between warehouses and factories (transportation cost) and production cost of an SKU at a factory (Production Process Model [PPM] cost). As transportation costs change almost every year with fuel price hikes, the company developed an automated approach of uploading the transportation lane cost matrix with revised values when there is a change.
It modeled some other business requirements using cost models. For example one of its manufacturing units was very old and the company did not want to run this unit beyond a certain number of days in a month. Cost were modeled in such a way that for first three weeks of the month actual cost of manufacturing was used and beyond that resource costs in that factory were increased to a very high level so that the Optimizer did not give any plan to that factory beyond the first three weeks.
10. Safety Stock Planning
Ensuring the right stock level at all nodes of a supply chain is a challenge for many consumer goods companies, as any stock-out can be a potential sales loss. APO safety stock planning considers demand and supply variability and takes the required customer service level into account while deciding safety stock at a particular node. Safety stock planning helps companies adopt a scientific method of stock calculation, but while implementing it, companies face the following typical challenges:
- Demand variability depends on forecast accuracy: Poor forecast accuracy may result in higher safety stock calculation. The system can directly calculate forecast accuracy percentage from values in forecast and actual sales key figures. Therefore, before implementing safety stock planning, companies need to have a reasonable level of forecast accuracy.
- Difficult to get data on supply variability: Very few consumer goods companies maintain data on the actual supplier lead time versus the standard supply lead time as maintained in the system. Many companies want to go by gut here by, for example, starting safety stock planning assuming a demand variability of 50 percent.
- Customer service level: This can be different for each product and can vary even by location. For example, a company may want to provide higher customer service in large cities (for competition) and a lower customer service level elsewhere. Often the customer service level for each product at the location service level is not known and developing a strategy and target service level for each customer/product itself can be a project.
- Runtime of safety stock planning run: Some consumer goods companies have experienced very long runtimes for safety stock planning background jobs. Again, many options are available to improve performance of this job, such as breaking it into several batch jobs that calculate safety stock for a set of products or running safety stock jobs separately for each node of the network (such as a separate job for calculating safety stock at distributor and another job for calculating stock at warehouse).
A large consumer goods company was implementing APO SNP safety stock planning for planning stock at more than 1,000 of its distributors. The company had a series of challenges while implementing this.
For certain product categories (especially for its C Class products) where they had a poor forecast accuracy, the system was suggesting a very high safety stock number — much higher than the number of days of safety stock the company was currently maintaining.
The company did not have much data to measure its supply variability.
The company initially wanted to keep 99.99 percent service level for all its customers. This eventually led to very high safety stock requirements in its network. Finally, the company had to do an extensive study to define customer service levels at different locations. For their new product categories and in cities they generally kept a higher service level (99.99 percent) followed by established products and rural markets. Service levels varied between 95 percent and 99.99 percent.
The company came up with a unique design to define a safety stock band for its product categories. Each product category has a higher and lower level safety stock band based on the number of days. For example, for the food dairy stock category, the highest band is three days, and the lowest is one day. If system safety stock calculation by a background job falls within that band (for example, 2.5 days), the system recommendation is taken for the planning run. If the system calculation is beyond this band (for example, four days) system macros automatically correct the calculation to the nearest band (for example, three days) and this is taken in the planning run. Band determination is based on a planner’s experience and is reviewed on a regular basis. However, over time the company gained confidence in the new system, removed these bands for most categories, and went ahead with system-generated numbers.