We live in a data- and algorithm-driven economic world. Data and algorithms also play an essential role in the planning and control of value chains in companies and of global supply chains. Work on improving algorithms has been ongoing for many years; the functionality of supply chain management software, from ERP systems to specialised APS systems, has been constantly refined.
However, the quality of the data used by the algorithms has not kept pace in many companies. This applies both to the input data that is processed by a software function and to the control data that regulates the behaviour of the software functions themselves.
Fig. 1: Every software system ultimately consists of a sequence of algorithms that process input data and are influenced by control parameters in terms of internal behaviour and sequence.
Particularly critical in this context are the data that control the sequence of calculation steps in an ERP[1] -, inventory management or APS system or that regulate the behaviour of a functionality itself. Although these data are not only used in MRP algorithms, but also in scheduling functions and all kinds of planning functions, in technical jargon they are generally referred to as scheduling parameters or MRP parameters.
Nothing works without correct MRP parameters
Without “correctly” set MRP parameters, an ERP system does not generate any useful forecast values and no sensible MRP suggestions, so that the users are forced to make their own considerations about future material requirements and sensible replenishment stocks. Instead of using the expensive functionality of the ERP system, the companies are thus thrown back on the gut feeling of their MRP controllers and the ERP system is only used as a golden typewriter.
Stable planning and scheduling processes that achieve identical results under the same boundary conditions cannot be achieved in this way. The fact that planning decisions are dependent on people and coincidence and change with every holiday or sickness replacement is a daily affliction in countless companies. A sustainable economic value chain in which the desired delivery readiness is continuously maintained with the lowest possible inventories is also a long way off with manual processes.
Fewer errors in the planning and control processes, faster processes and differentiated processes can only be achieved with more process automation, for which good MRP parameters are crucial.
Up-to-date, complete, and error-free scheduling parameters also have strategic importance for the future productivity of our companies. A large part of today’s labour costs in our companies are incurred in the administrative areas and no longer on the shop floor. Urgently needed productivity gains can therefore only be achieved by streamlining and automating the administrative areas, which include demand planning, operational procurement and production control.
Ultimately, demographic trends will force us to automate more in the coming years, as the staff needed for these supply chain management tasks will become increasingly difficult to obtain and more expensive to pay.
Conventional maintenance of disposition parameters does not lead to success
In discussions with companies, the responsibility for poor MRP parameters is usually seen in the hands of the users: “Unfortunately, our people do not maintain their scheduling parameters in a disciplined manner”, they say. However, a simple calculation shows that manual master data maintenance would take far too much time and is therefore unrealistic.
Fig. 2: Conventional maintenance of MRP parameters does not lead to success
To do such a calculation, it is necessary to understand that all articles are subject to a life cycle and that market demand is shaped by many influencing factors. It is therefore not enough to set the MRP parameters of an article once, when creating the article. They must be readjusted regularly. If a user is responsible for 1,000 material numbers, has to maintain 10 MRP parameters for each material number, spends an average of 60 seconds per MRP parameter for calling up, thinking about and setting the parameters, and has to check the MRP parameters four times a year, this requires 667 hours per year. This corresponds to an additional 40% job per planner, purchaser or production controller just for maintaining the MRP parameters! In practice, planners are usually responsible for far more articles, have to maintain more than 10 parameters and do this, if possible, monthly and not just once a quarter.
Some argue that the staff would definitely have the time to maintain the MRP parameters, since better MRP parameters would allow for higher automation. However, if every user maintained his or her own master data, neither reproducible scheduling decisions nor stable scheduling processes would be possible. Every user behaves differently, has different experiences and a different perception of risk, and therefore adjusts the MRP parameters differently than his or her colleagues. Can there be 10 different correct solutions for one planning decision? Probably not!
It makes more sense to define uniform criteria for all users on how MRP parameters must be set. Therefore, in the 1990s the idea of the MRP parameter maintenance manual came up, in which, among other things, specifications for the maintenance of MRP parameters were defined. These manuals did not have widespread success; today we only very rarely find corresponding work instructions. The first reason for the failure of the MRP parameter maintenance manuals was their manual application, which made the time needed to maintain the scheduling settings more time-consuming than with individual data maintenance according to the users’ gut feeling. But even where this hurdle was overcome by appropriate software support, no satisfactory planning results were and are achieved.
The biggest challenge in maintaining MRP parameters is that the dynamic interaction of demand, forecasts, MRP decisions and value streams within a company and throughout the supply chain is so complex that it cannot be understood with common sense alone.
It is therefore not enough to standardise setting rules and automate their application. You have to start by setting all the relevant MRP parameters for each individual material number, which interlock with many algorithms, in such a way that they lead to economic planning decisions on average, not in each and every individual planning situation(!). Different companies may well have different objectives. Should the required delivery readiness of an article be achieved with the lowest possible average stocks or with the lowest costs or is the reduction of warehousing and ordering costs more important than delivery readiness? Even within a company it can make sense to pursue different profitability goals for different articles.
Economic settings for disposition parameters can only be found by means of simulation
In order to find the right settings for MRP parameters under the dynamic circumstances described above, in our experience there is no economic way around simulations with empirical data. The only alternative to this approach would be to test parameter settings in practice by trial-and-error; a lengthy, costly and risky undertaking.
Fig. 3: MRP parameters – depending on numerous influencing variables – have to be set and readjusted correctly
For the optimisation of MRP parameters by means of simulation, we draw on the extensive historical data stocks of the ERP system. By linking customer order histories, stock issue histories and inventory histories with article master data, bills of material, routings and other ERP data, a complete dynamic model of the structure and value stream behaviour of a complete supply chain can be built. This digital twin of the physical supply chain contains the special features of the market behaviour as well as the special features of the value stream behaviour. And so, in this digital twin, the effects of different settings of MRP parameters on inventories, delivery readiness, warehousing and ordering costs, number of incoming and outgoing goods and other cost drivers can be empirically determined and improved so that the targeted economic goals are achieved as well as possible.
Fig. 4: In a digital twin, the most economically sensible settings of the scheduling parameters can be determined systematically and empirically.
The strategy and operational behaviour of supply chain management is codified in rules
With the current performance of our computer systems, it is impossible to simulate all the possibilities for setting the MRP parameters for every single of the countless articles in a company. And if it were possible, the result would be completely confusing. Experience shows that this effort is not necessary when proceeding cleverly. Instead of attaching the MRP parameter settings to a material number, it is determined under which circumstances, for which classification characteristics of articles and for which characteristics of articles, which parameter settings provide the best economic results. From this, rules can be derived that say “if an article belongs to this article class, faces these circumstances and has the following properties, then this MRP parameter receives this setting value”. In this way, rules can be structured into decision tables and decision tables that build on each other can be structured into sets of rules.
These sets of rules contain the complete control intelligence of operational planning as well as the considerations for the strategic positioning of a supply chain. This way the supply chain business model is mapped in sets of rules.
If these sets of rules are stored in a suitable software system, such as DISKOVER, they enable the continuous and automatic maintenance of the MRP parameters. Since DISKOVER can not only manage the maintenance of the MRP parameters, but also performs the simulations, the most sensible setting alternatives of essential MRP parameters for each individual article can be automatically and regularly determined by means of empirical simulation and directly applied instead of coding them into rules.
The result is extensive CNC control of the ERP machine
The automatic maintenance of the MRP parameters of an ERP system by DISKOVER runs without users having to intervene and spend time. Since the maintenance process does not require any effort, it is carried out daily. In this way, the parameter settings are readjusted promptly and the number of changes per maintenance run is kept low.
Fig. 5: A powerful system for maintaining MRP parameters is as important for an ERP system as the CNC control is for a machine tool.
Of course, the near future will not consist of fully automated planning processes. Planners and dispatchers will still be required. However, we need to use this precious human resource where human intervention is essential and human experience is indispensable. Today, much of our staff’s intelligence is wasted on tasks and items that well-maintained planning and scheduling algorithms could do on their own. In return, other tasks that urgently need to be tackled by staff fall by the wayside.
So let’s turn the management of our supply chains upside down and redistribute responsibilities between algorithms and humans.
[1] In the following, for the sake of simplicity, the term ERP system is generally used as a summary for different types and characteristics of planning and control systems.
