Remember the model airplanes we used to build as a kid. At the time it was a lot of fun. Probably our first experience at creating something out of seemingly unrelated parts. Slowly it took shape and began to look like the picture on the box. Lucky for us or our parents, there were instructions and a picture.
Just looking at a picture of the plane would not be quite the experience of putting together the parts. The process of building the model conveyed relationships. At the simplest level, we saw what parts were connected to other parts. From a child's perspective that was more than enough, and probably for the parent helping.
Models of new plane designs are constructed that serve an expanded function. Relationships are tested to determine performance of a unique design. A new wing design might be tested in a wind tunnel to check the relationship between the wing design and drag. The entire plane design might be tested to check for stability. In this context, the model is serving the purpose of quickly and inexpensively gathering design information for the future construction of a full scale plane.
Sometimes these same models are used to help diagnose what went wrong in the event of a failure. Maybe the same design is to be used in a different application or environment (whether deliberate or by circumstance). It would be worth while in these situations to get an idea about the affect on performance of the changes on the design. Why risk the full scale equipment, unnecessarily?
This is one of the key points. Models are used to minimize risk. They help anticipate the affect of different conditions on equipment. The word equipment could easily be replaced with either process, system, manufacturing, time, costs or profits.
Testing & anticipation are not always negative terms. At the same time risk is minimized, opportunities are identified for improvement. Models will do both, sometimes simultaneously. External environmental factors are going to have an affect on a design, either positively or negatively.
There may also be several alternative design modifications that could correct or improve performance. Testing these on the model are rarely perfect, so the modeling may not be accurate enough to indicate the best alternative. However, it may narrow down the field of possibilities. The new prioritized list thereby directs further investigation.
Testing ideas and theories is expensive and time consuming. There are limits to what a model can emulate. The challenge is to determine the situations that can be reasonably modeled to take full advantage of the savings of modeling verses testing at the full scale. This is particularly true when there are allot of influencing variables and alternate responses.
Last but by no means least, models help people visualize complex situations. The object here is not to convince you to go out and build airplane models, although it may sound like that up till now. This is serious business. Opportunities that are not visualized are opportunities missed. In a competitive environment, the critical steps are identifying every available opportunity, seeing them faster, and taking the best advantage of them by taking the best action. Models are very important to staying competitive.
The people who are close to the data can intuitively see relationships between the market and manufacturing (supply chain capability). Management as well as others will not see these relationships as quick or as easily because they don't handle the same amount of information. Visual (graphical) models will help others see and understand supply demand relationships. Supply/Demand models are the models that "business" need to stay competitive.
There are three potential Supply/Demand models that can be the most useful - strategic, tactical (Sales Operational Planning) and operational (Scheduling). The major difference is the level of detailed information used and the intended question.
Strategic planning deals with long range resource planning for large capital expansion. Most times this analysis is relatively simple. When a manufacturing site is highly integrated, with recycle and waste streams that are highly dependent and limited, then a model is warranted. The data used is at the highest level of aggregation, analyzing one time period, a year's capacity; or three to five years in yearly buckets.
Tactical planning tends to handle more variables than strategic planning. Incremental adjustments to inventories, storage, capacity, raw materials and transportation need time to arrange - 3 months plus. So, the primary function is resource planning in the intermediate term, that will help the business respond to changing market conditions. Some product family production planning may also be done. The data is aggregated by product families and is done on a monthly basis, going out 12 to 24 months.
Strategic and tactical planning were done on an entire plant's capacity or critical resources. Operational planning is done on work centers within a plant. Specific products are modeled at the highest level of detail. The intent is to determine the best time to manufacture, schedule production.
Stay simple if you can avoid building complicated models. It is really dependent on the complexity of the supply chain or specific manufacturing operations. This is what will determine the complexity of the models and the need for them.
Spreadsheets have become a very popular way to model strategic and tactical plans. The one handicap of spreadsheets is that they wont time-phase the data, emulating the lead-times of material movement, purchasing and production time. This not usually a problem for strategic and tactical planning due to the large time buckets, months and years.
When working at the weekly, daily or hourly level spreadsheets are inappropriate, except for the simplest situations where time-phasing might be done in your head. You can determine how often that is.
Linear programming is an advanced mathematical routine for finding the optimum solution for a specific situation. An expert system could apply logic to the conditions and variables used in arriving at the final solution. Complex problems may take advantage of these types of programming methods for making the job easier. The models are typically very customized and require considerable expertise to maintain.
Operational scheduling handles the most input elements, variables and possible objectives, thereby, making it a good candidate for linear programming. Linear programming has been the mainstay for scheduling for many years and more recently expert programs have been used to supplement it.
There are a number of ready made software packages that were designed specifically for Supply/Demand modeling. All have strengths and weaknesses relative to what are your specific needs and requirements. My best suggestion is to carefully analyze the intended use and find some help from an experienced person in the field. The spreadsheets and other programming packages mentioned above run in the couple hundred to two thousand dollar price range with more expertise required to do the programming relative to the Off-the-Shelf packages. The Off-the-Shelf software will range between $25,000 to $100,000. The savings over spreadsheets come from the programming & maintenance end and enhanced capability. They require less expertise and are quicker to set up.
These packages usually include a graphical interface which will be very important to the end user. They also have variations of linear programming to help solve generic optimization objectives. The cost of the software is somewhat related to the sophistication of these algorithms.
The current direction of Off-the-Shelf packages is to allow for advanced programming with outside linear programming routines and expert system interfaces. The advise is the same as described above. Be careful not to over do how much support is needed to maintain the program. Customization extracts a price.
Computer models of business supply/demand dynamics can yield the same benefits of airplane models. They can allow the business to fly faster and higher. Businesses need to test and anticipate the impact of market conditions on their ability to respond.
Supply/Demand models are needed to design and optimize Kanban and JIT systems. MRP systems don't feasibilize constraints therefore suggest unrealistic schedules. Tying a scheduling model into a MRP system is essential for making it operate properly.
We have arrived at why I wrote this article. All manner of planning systems have been put in but most move bad information around faster and don't improve the quality of the information. Quality decisions come from quality inputs and being able to understand the relationships between cause and affect. It is apparent that many implementers of MRP systems have not taken the time to model the business' supply/demand relationships for that business, so that the MRP system could be properly programmed and its limitations understood. The businesses that want to survive and be the best will learn how to build their model airplanes.