Comparing PP/DS to Simio for Production Scheduling

Executive Summary

• We cover the differences in production scheduling between PP/DS and Simio, what each system is designed to model, and what types of manufacturing environments work for each.

Introduction

PP/DS is a well-known advanced planning application for production planning and scheduling offered by SAP which is part of the APO suite. SAP is the largest enterprise (i.e., corporate) software company globally, with revenues close to $13 billion. The PP/DS product was introduced in roughly 1999, and I believe based on i2 Technologies’ Factory Planner product, which was one of the benefits which SAP received from their partnership, albeit brief with i2 back in the late 1990s. a

PP/DS provides both a cost optimizer (which is very rarely implemented) and both production planning and detailed scheduling heuristics. There are heuristics both for the initial planning run and for capacity leveling.

The Background on Simio

Simio is a relatively new company but comprises individuals with a great deal of experience in simulation and have worked for and developed significant intellectual property for some very well-known simulation applications. Simio is one of the few simulation systems that provides an easy-to-use integrated development environment for 3D simulation and is one of the few companies to provide 3D simulation for the production scheduling market. While Simio is a “simulation” environment, it is also a planning environment. Planning is essentially a simulation or prediction of the future. However, as I will discuss, there are several ways to interact with and gain information from Simio. Using the 3D simulation view is just one way.

Manufacturing Environments

As with most production planning and detailed scheduling applications, PP/DS works best in discrete manufacturing environments. Discrete manufacturing environments are the easiest to model. SAP sales are extremely effective at selling PP/DS into different manufacturing environments than this, but it is not successfully implemented outside of discrete manufacturing unless it is completely customized. Simio, on the other hand, has a much wider breadth of industry implementation. However, Simio tends to focus on the most severe production scheduling problems.

As I will describe in the probabilistic modeling section of this article, Simio is mainly designed for manufacturing environments with a high degree of variability, which is the type of situations that have up until very recently not been addressed with packaged applications. Currently, these environments are, of course, planned. Still, they tend to be designed with a great deal of manual effort in spreadsheets or with custom-developed optimizers using off-the-shelf general optimizers like CPLEX. Easy-to-schedule production environments are not Simio’s target market (although it could be utilized in such situations). Simio’s prototypical customer has very high scheduling complexity, and for which there are substantial penalties for being late.

Available Views

PP/DS has two basic views. One is the Detailed Scheduling Board, which is your standard Gantt chart representing resources, which can allow jobs to be moved around between resources. The second is the Product View which shows the movements in and out of products into a facility. PP/DS has more functionality than many companies implement. One of the reasons that PP/DS often is implemented with a small subset of its functionality is the shortage of views available in the application.

One of the things that strike us about Simio is how many views it has. Simio’s “showcase” view is 3D modeling. However, there are many views in addition to this. The model building view also can be run at any time, so the modeler can see the flow of material as there are building the model.

Simio can be used in this 3D mode or interacted with more in a traditional advanced planning system. For instance, Simio also has a Gantt Chart but shows the probabilities or percentage likelihood with any outcome. Each job shows its end date and then its need date so that the scheduler can see how much lag of flex they have in the schedule for that product. Simio has a resource view, which shows the sequence and length of time resources are being utilized.

Reports

One of the common needs of PP/DS projects is reports. The primary report available within PP/DS is the Alert Monitor. This is a listing of the areas of concern which the planner needs to address. The Alert Monitor was better back several releases ago. I wrote a paper about using the APO Alert Monitor as an adjunct reporting system back in 2003 for SAP Tips. However, the interface was changed along the way, and I no longer use it in that capacity.

On all APO implementations, regardless of the module, it is necessary to provide additional views. This applies in particular for aggregated or cross relationship views. This means typically working with SAP BI resources to develop the report design in some mockup applications. I have used Excel for this several times, which I describe in this article. However, this is a time-consuming process that I would prefer not to go through continually. Secondly, the reports take a great deal of time to build the BI teams at every project I have been on. It is quite typical for the BI reports to lag the PP/DS go-live by at least six months. A final issue is that BI is an external system to APO and PP/DS. This means that there is a lag in the report’s currency so that the report frequently only reflects information up to the previous day.

Reports are a solid area of Simio. There are pie charts for resource utilization and aggregated views of the probabilities of the planned production order.

Capacity Constraining

PP/DS is often sold as a tool for performing constraint-based planning. However, it is more accurate to say that PP/DS can perform constraint-based planning, but this is only effective if the cost optimizer is used. Quite a few companies have attempted to implement the cost optimizer in PP/DS, but very few stay with it. It should be said that this is not a problem that is unique to PP/DS. Optimization in many applications has often run into problems in implementation.

In a forward-looking industry like enterprise software for supply chain planning, there has been very little analysis of why this has been the case and very little done to improve implementation methodologies or even slightly adjust strategies on new deployments. For this reason, most of the optimization projects repeat the same mistakes that I first observed on projects back in 1998. That period was my first exposure to optimization in packaged software, so there it is quite probable that the same mistakes have been made since optimization was first introduced in supply chain planning.

After a good deal of analysis, I have concluded that a major reason for optimization failure has been the over-application of a single optimization objective function, namely, cost optimization to every supply chain domain. This conclusion is documented in this article, which describes the many limitations in getting cost optimizers to work in production environments.

Since the vast majority of companies implementing PP/DS do not use the optimizer (they either use heuristics or no method, just allowing planners to move jobs around manually to resources), PP/DS should be considered an unconstrained tool. However, PP/DS does have a wide variety of heuristics. In fact, there are over fifty, which are described in this article. Many heuristics, many directed just towards specific manufacturing environments (for instance, many of PP/DS’s heuristics are for repetitive manufacturing), is probably the strongest point for PP/DS. Most PP/DS projects result in the PP/DS Consultant matching the combination of heuristics and their sequence to the production process’s particular requirements. However, heuristics are not a “great” approach to production planning. They provide decent results; they are fast to run, but they are frequently used because they are simple to implement. They do, however, offer a great deal more flexibility than MRP (if we compare them to the production planning heuristics, rather than the scheduling heuristics in PP/DS). The vast majority of PP/DS implementations do not use the cost optimizer and are not constrained to reiterate. I run into clients who think they are performing constraint-based planning when running the PP/DS heuristics, which is not correct.

Simio uses optimization-simulation and is commonly implemented with constraints. The data setup is relatively simple in Simio, and this enabled resources to be effectively updated, which is a common problem with all planning systems that rely upon resources.

Model Setup

The model setup in PP/DS is the most complicated and time-consuming of any production planning and scheduling tool that I have ever had exposure to. PP/DS projects require a minimum of one year to implement. Most PP/DS projects take longer than this, with the first go-live not being a correct measurement of the project duration because few PP/DS projects attain very much planning buy-in until later rollouts. Some data that populates PP/DS is brought over “automatically” from SAP ERP using an SAP integration product named the CIF. However, as the next section describes, this is not at all as straightforward as it sounds. PP/DS also has data that does not exist in SAP ERP and is maintained in PP/DS. The primary method of managing this data is with the MASSD transaction in SAP APO.

Simio has two major areas that require setup. One is the data, such as the materials and BOMs.

The other is the actual model building.

Integration

PP/DS is only is implemented on accounts that have implemented or are implementing SAP ERP. Therefore, PP/DS’s integration to SAP ERP is relevant, and its integration ease to other ERP systems can be considered irrelevant. PP/DS is connected to SAP ERP (as is the rest of the APO suite) through SAP middleware called the CIF. I have been part of teams that built more reliable adapters between SAP and a best of breed application than the CIF, and which outperform the CIF in both performance and reliability, and I have done so with standard integration languages and UNIX, combined with ABAP (the code required to extract data from SAP).

Issues with long-term maintenance of the CIF can be understood from reading this article. Because of my exposure to custom-built tools and too refined middleware products like Informatica, the CIF has always left me nonplussed. I have written on several occasions of my concern for the extremely high maintenance required to run the CIF, and it is a level of maintenance that never seems to tail off as I think it should.

Probabilistic Modeling

PP/DS, as most production planning and scheduling applications, does not address probabilities. All assumptions are “deterministic,” which means they have no range of values. One can, for instance, set up alternative resources; the plan will only pick the number of resources necessary per run. There is no variability during one run. A range can only be developed through making changes to master data. Sometimes this is kept in different versions of the application, which is a process called simulation.

In practice, this is rarely done. I have been writing about the lack of simulation used in PP/DS and other APO applications for some years. Performing simulation in PP/DS is an entirely manual process. The vast majority of companies are not staffed to perform simulation in any module in APO.

The complexity of APO combined with the lean staffing at businesses in planning departments and the resources even the largest companies seem willing to allocate to these agencies fairly well limits simulation in APO to being occasional affairs. Knowledge of how to document and archive simulation results is significantly lacking in the industry. I have explained in this article. PP/DS reduces the simulation performed at companies because of its overhead and how simulation is set up. I describe how much easier simulation is incomparable products in my book on supply planning.

Newer production planning and scheduling applications allow the simulation to be performed by simply having the simulation as a separate tab in the planner’s interface. The planner was able to switch between what-if scenarios. This is a functionality resident within the production planning and scheduling application PlanetTogether. Simio’s approach is entirely different from all other applications in this space that I am aware of.

Simio iterates through multiple scenarios are eventually coming to a probabilistic production schedule. That is, a planned order has a percentage likelihood of being late. This brings down the effort to deal with a probabilistic program because probabilities Rather than certainties are built right into the assumptions. I also mean that a separate simulation process, where certain assumptions are changed less necessary. I say less and not “unnecessary” because classical simulation can still be useful when testing a proposed switch to the master data or configuration.

Conclusion

There are quite a few things to consider when comparing a standard packaged application like PP/DS and comparing it against something which is a modeling environment like Simio. Unfortunately, too many companies move directly to using a packaged application without considering the alternatives.