At a recent SAP user’s conference, a speaker discussed how his company (a Fortune 50 company) had challenges managing inventory and production scheduling. They heeded management’s and SAP’s exhortation to respond to all MRP exception messages. The company hired Accenture as a third-party resource to manage approximately 25,000 MRP messages per week. At the end of every week, the company would have the India-based Accenture resources process all 25,000 or so messages. On Monday, the planners would come back in to see the latest planning run and the new messages.
First of all, what a great business model for Accenture. They hire low cost labor to do highly manual, tedious work and charge a premium. Be careful if Accenture is also your system integrator, system consultant and 3rd party processor of ERP messages as there is a strong conflict of interest. The more messages the system produces, the more money Accenture makes.
Second, what a mind-numbing task for planners. If management direction is to respond to all MRP messages, because this is the only way thought to properly control the system, a planner’s job is filled with drudgery. See the message, investigate the message, respond to the message. Repeat.
MRP/ERP are deterministic systems. Trying to follow all the messages generated by MRP is to assume that the future can be predicted with enough data and technology. This is flat wrong and a source of unhappiness for many a manager and planner. MRP nervousness is a fact of life. See Spearman and Hopp, Factory Physics (2008, Waveland Press) p136, for more description. With MRP, feasible plans can become infeasible even when demand drops. More data and more data faster does not provide better control.
The secret to happiness with your ERP system is to structure planning practices and your ERP systems to naturally manage stochastic behavior. Stochastic systems are those that exhibit variability1. Variability is a fact of life. Taking a stochastic approach to using ERP systems allows both reduction of MRP messages and strong prioritization of remaining messages. It’s not that MRP messages are bad, they serve a useful function. It’s the way companies manage their systems that creates message management chaos.
Here’s a quick guide to achieving MRP happiness and better performance. It’s call Dynamic Risk-based Scheduling (“DRS”) and is described in detail in Chapter 7 of Factory Physics for Managers:
- Set policies accounting for the levels of variability in demand and supply you are willing to tolerate in your business. These are set in three main areas:
- Capacity – planning to run at 100% capacity utilization is a recipe for failure. There must be some capacity buffer to enable predictable variability management.
- Inventory policies – where raw material or finished goods stock inventory is used, optimal policies (when to order, how much to order) by part should be implemented.
- WIP policy – for a given level of capacity utilization and variability (in demand and supply), the amount of WIP in a process is a primary design parameter. Whether you call it push or pull is not so important, what is important is maintaining the right amount of WIP to get maximum throughput with minimum cycle time2.
- Establish a structure of CONstant Work-in-Process ("CONWIP") flows as illustrated below with stock points at each end and a virtual queue in front of each flow. CONWIP controls total WIP level in the flow, not at each station as does classic kanban. Optimize WIP levels to get maximum throughput with minimum cycle time. Factory Physics Inc.'s CSUITE Operations Analytics provide the powerful Flow Optimizer and Inventory Optimizer performance curves to ensure best possible performance and do rapid improvement identification and scenario analysis.
- Use MRP work orders in the virtual queue ("VQ") as a capacity trigger
- The virtual queue is like a barometer. When work orders in the VQ go up or down past calculated trigger points, it’s a signal to adjust capacity.
- Work orders in the VQ can stay as planned WOs and don’t need to be firmed until they are due to be released from the queue into production. This way, MRP message management activity is greatly reduced.
- Once a work order is moved into the CONWIP flow, nearly all messages except for cancellations would be ignored. Once work is in the production flow, very little expediting or schedule changes are allowed.
- Since optimal WIP levels are maintained, the cycle time to complete for the work order will be as short as it can be which also helps with minimizing MRP message management.
Following is a chart of on-time delivery for a major medical products company that implemented this practice. The initial thought was that the approach would require more planning time but the exact opposite happened. Planning and control time actually dropped because of the reduced level of management activity required and the improved predictability of performance. On-time delivery performance increased nearly to 100% even as demand increased. WIP inventory dropped by $6 million (60%).
The result of Dynamic Risk-based Scheduling is a system that:
- is easy and predictable to manage
- results in much less MRP message management than standard production control
- provides maximum throughput with minimum cycle time and minimum cost tied up in working capital and capacity
- implements and controls optimal inventory levels while delivering highest desired service levels.
These characteristics will greatly reduce drudgery and stress in managers’ and planners’ jobs leading to a much-improved sense of control, well-being and happiness. Additionally, DRS should drastically reduce, or even eliminate, the need for a third party message manager—a move that will certainly make executives happy with the reduced cost. –ESP
For more information on Dynamic Risk-based Scheduling, CSUITE Operations Analytics and Factory Physics services, send an email to Ed Pound at email@example.com
1 The term variability is used loosely here, as is common, to include variability and randomness. A sequence of numbers 1,1,2, 1,1,2,1,1,2… has variability but it’s not random. There is variability because all the numbers are not the same. However, it’s a non-random repetition of 1,1,2.
2 Cycle time here means the time it takes a job from the beginning of a routing till the job completes all steps of the routing—typically across different work stations or machines.