I. Overview of Network Analysis
When we decide to build a factory or distribution center, we must consider how to design the building structure and what kind of material handling equipment and systems to use, but we must first answer some basic strategic questions. Should we build a new warehouse or expand the existing warehouse? How much do we need to build? Do we need to integrate or close several warehouses? Where should the warehouse be built? What kind of products does the warehouse need to handle? What kind of customers should the warehouse serve? Questions like this are usually part of network analysis.
Network analysis
Simply put, network analysis is to support the decision-making process of a given supply chain with appropriate physical equipment (planning, product line, distribution center). This process is driven by a series of cost factors and operational constraints. Cost variables vary with the scope and nature of the study (manufacturing and distribution, single warehouse and multi-warehouse). But in general, the general categories of costs include the following categories:
1, raw material procurement cost
2. Fixed costs
3. Variable cost
4. Inventory holding cost
5, installation and transportation costs
6, outbound costs
Operational constraints are those requirements that do not consider cost. There are many operational restrictions, but they usually include the following factors:
1, facility status (locked open/locked closed)
2. Facility capacity (product type and load capacity)
3. Storage and production capacity of facilities
4. Customer service requirements
5. Procurement requirements (single supplier and multiple suppliers)
6. Minimum and maximum number of facilities
Network analysis is also constrained by demand factors (demand quantity, customer location, product structure) and optional networks (alternative location of facilities, transportation situation).
Model tool
In addition to the simplest network, the existence of a large number of special networks, the need to evaluate many cost variables at the same time, and the need to meet operational constraints make it increasingly difficult to solve problems with traditional model methods (calculators, spreadsheets, etc.). ). Making the best choice (minimum cost or maximum profit to some extent) requires the use of network model tools. There are many business model tools available. Most tools consist of three basic parts: a user interface that can input data of demand, cost and constraint variables; A converter for converting these data into corresponding mathematical functions; And analysis engine to make the final solution. The solution engine uses powerful modeling tools and special mixed integer linear programming theory to calculate the real optimal solution, so it is called "optimizer". Most tools also have statistical data and graphic output functions.
Modeling and analysis
Network modeling and network analysis are generally considered to be the same. In fact, network modeling is a part of the network analysis process, and it is also a very important part. We make this distinction mainly because there is usually a misunderstanding: a truly optimal network can be determined by establishing a complete model. However, the model is only a mathematical tool to calculate and optimize the objective function under a series of constraints and given data. It should also include users making a large number of operational assumptions, inputting relevant data for each hypothetical situation, and correctly understanding the model results. It should also consider those factors that cannot be quantified in the model (such as risk management, human impact, sales and market impact, etc.). ).
Second, the benefits of network analysis
The first benefit of network analysis should be cost saving. Of course, there are many other benefits. At least this can bring good communication and interaction between departments.
cost saving
Network analysis may bring 5%- 15% logistics cost savings. Of course, this will change with the actual situation, assuming that the current network is sub-optimal. It also depends on the changing ability of the internal network. For example, according to the provisions, a specific distribution center must continue to operate, or users mainly protect the original factory, so it is difficult to get cost savings. Finally, cost saving lies in avoiding some costs rather than just reducing them. Usually, network analysis is to find new facilities to adapt to new growth, rather than integrating existing facilities to reduce costs. In this case, it is difficult to quantify the cost savings because there is no clear benchmark to measure the solution.
Other benefits
In addition to cost-saving opportunities, a mature network model can bring many other benefits. The optimized network can improve the customer service level by shortening the delivery cycle and improving the order satisfaction rate. The network model is also a good budgeting tool, which can predict the future capital demand and operating costs. It is also an ideal testing tool to quickly test other operational scenarios and the impact of acquisitions, new products and other business changes. Most importantly, the network model is also a catalyst to encourage internal communication. When constructing and evaluating the network model, many discussions are needed, including strategic planning, finance, sales and marketing, customer service, information system, procurement, inventory control, manufacturing, distribution, transportation and other departments that affect or are affected by the changes of the logistics network. Because these people express their views from the perspective of the whole organization, they can form some new perspectives and information. Finally, after collecting and analyzing operational data, some new improvement opportunities may appear.
Third, the modeling requirements
In order to establish an effective model, it is necessary to collect and verify a large amount of data. Network analysis has three basic driving factors: demand, cost and constraint. Efforts must be made to find data related to each hypothetical scenario. In addition, the integrity and representative data in the model must be considered. The model is carried out at the product group level (dry goods versus freezing, pallet picking versus unpacking picking) rather than at the SKU level, and scattered customers are divided according to type (big versus small, retail) and geographical location.
ask
Demand data describes the basic information of customers and reflects the characteristics of orders. These data are generally obtained from historical customer purchase data, preferably 12 months, in order to grasp those seasonal purchase characteristics. Sort and divide the data according to products, customer types, geographical locations and modes of transportation (parcel delivery, LTL, vehicles, etc.). ).
expense
The quantity and type of cost data depend on the scope of analysis. Generally speaking, costs include fixed costs (unrelated to demand) and variable costs (a function of demand). Fixed costs include the capital of facilities and equipment, as well as indirect costs, such as managing labor. Variable costs are generally equivalent to operating costs, such as direct labor and transportation. Other costs, such as inventory holding costs, can be said to include fixed costs and variable costs, and modeled accordingly. One of the tasks of this model is to analyze the trade-off between fixed cost and variable cost. Take the newly-built distribution center as an example. Assuming that this distribution center does not need to be built in operation, it should be built only when variable cost savings can make up for fixed costs. Fixed costs include facilities, equipment, additional administrative staff and related inventory costs. This may reduce the outbound transportation costs of local customers. The increase or decrease of inward transportation cost depends on the whole network, as does the direct labor cost. If variable cost savings can make up for fixed costs, then this distribution center can be built, otherwise it should not be built. In some cases, cost data is not so easy to obtain, especially if you want to obtain cost data through product mix or customer classification. General manufacturing and distribution costs can be obtained from business schedules, income statements and other reports, and overall transportation data. The difficulty lies in how to get the rate of transportation mode and route. In some cases, especially parcel delivery and LTL delivery, this information can be obtained from the published price list. However, it takes a lot of time and energy to obtain this information in the way of vehicle transportation and railway. Finally, determine whether to consider or not to consider some cost factors in the analysis. Those costs that are not considered are not important, and some may be important, but they should be excluded because we don't want them to have an impact on the network. The latest typical example is the inward transportation cost provided by the third party logistics. Although these costs are important, they are usually ignored because we want to design the network around customers rather than suppliers. In this case, sensitivity analysis is usually used to determine the impact of these decisions. It is also used to evaluate the sensitivity of the scheme in order to increase or decrease different cost drivers.
Constraint condition
Constraints are factors that users add to the model regardless of cost. There are many forms of constraints, of which four are the most common. The first is the capacity limitation of production line, workshop or distribution center, and the second is the qualification limitation. Qualification restrictions may prevent warehouses storing frozen products from storing dry goods and production lines producing glass bottles from producing canned goods. The third is the limitation of customer service. The limitation of service level is that the construction of facilities can not only consider the cost. Finally, the on/off limit. It limits the maximum or minimum number of facilities, and/or some facilities remain open or closed.
challenge
The two biggest challenges of successful network analysis are the incompleteness of data and the inability to consistently run through the research objectives. The latter is a problem of project management. Due to the lack of corresponding experience of a large number of people involved in the research, network analysis can easily fall into inappropriate data collection and analysis, and may turn the process to other directions.
On the other hand, the data problem is not artificially controllable. There are three solutions to deal with data incompleteness. First, you must ensure that these data are available. In the long-term strategic analysis, incomplete and unnecessary data can also get correct results. The second is to leave room for missing information. These spaces take many forms, and usually use the most optimistic estimates rather than specific information. Finally, it is the key data of analysis, so we should study and analyze it hard to get useful information.
For some international models, these challenges will be even greater. Project managers also face greater challenges, including language barriers and jet lag. For some reasons, data collection has also become difficult. The biggest problem is the lack of standard transportation price list. For example, unlike the United States, there are basically no LTL transportation price lists in other countries. In addition, the transportation infrastructure in different countries and places is very different, so it is difficult to estimate the transportation time and distance. The gap between regional labor and facility costs is also more significant than that in the United States. Of course, tax rates and business rules are different in different countries and different regions of the same country. Tax considerations can greatly change the research direction. In most cases, the best way to deal with these data problems is to rely on the opinions of local experts and take the time to thoroughly study those factors that have significant costs or restrictive effects.
Fourth, the key to success.
The key to the success of network analysis lies in data and manpower. Judging from the data, the analysis must be accurate and reasonable. The data may be incomplete or hypothetical to some extent, but it must not violate the overall research direction and must win the trust of those managers who may take action according to the plan. More important than data is people. First of all, a successful analysis needs an experienced analytical person, whether internal or external, to process data, build models and lead the whole process. Secondly, the project team should be composed of a group of people from all over the country who can handle all kinds of business problems and logistics problems of impact analysis. Usually, the project manager coordinates the whole team. Thirdly, it is the support of senior management. If the research is not well affirmed, team members will not participate and the project will soon lose momentum. Finally, we must set certain goals and strive for them strictly. Network model analysis can easily be misunderstood as something else, so do some unnecessary analysis. This can be avoided by these methods, such as explaining the strategy of network analysis in the early stage, defining the boundary of network model and defining the purpose of analysis.