| A Control System Perspective | | | | the real differentiator is the nature of the |
| | | | system-type. A push system is a feed-forward |
| Software can be a real pain. But before your tear | | | | control system where as a pull system is a |
| the rest of your hair out or call your software | | | | feed-back control system. |
| vendor in frustration you might want to think | | | | For those whom have spent time and energy |
| about some of the nasty little details that can | | | | implementing a so-called closed loop MRP system, |
| cause the pain to begin with. The concepts below | | | | and constantly frustrated by the results, you can |
| are not necessarily life changing. Yet many | | | | take solace in that there is a big difference in |
| manufacturing specialists have ignored them and | | | | calling something ‘closed-loop and actually |
| have needlessly suffered as a consequence. | | | | having a true feed-back system. |
| | | | Recall that the first iteration of MRP in the late |
| First, let me be clear that software programmers | | | | 60’s was followed very closely by |
| and providers are NOT evil people. In fact, I am | | | | ‘revised’ versions called “Closed-Loop |
| amazed and quite proud of how dedicated they | | | | MRP” in the early 70’s. It became |
| are and how much progress has been made | | | | apparent that without some feedback mechanism |
| during the last three decades. Unfortunately, like | | | | built into the system it was unworkable. By the |
| any automated system, there are inherent | | | | 80’s the inclusion of several feedback |
| limitations to what these systems can and can't | | | | modules (i.e. shop floor control and input/output |
| do. The purpose of this article is to help you | | | | control) was necessary to assure the planning side |
| understand that how you think about your | | | | was kept valid. And from Closed MRP a new |
| operations and organize your work is as much of | | | | improved methodology manufacturing resource |
| a factor for success of a software | | | | planning (MRP II) emerged to dominate U.S. |
| implementation as what code you choose and | | | | discrete manufacturing in the 80’s and |
| how you choose to configure it. | | | | 90’s. |
| Special Note: Up until now, I may have given you | | | | Unfortunately, the term ‘closed-loop’ was |
| the impression that ERP (Enterprise Resource | | | | a bit of an exaggeration. Yes, the system |
| Planning) is synonymous with MRP (Material | | | | technically has a feedback loop. It is self correcting |
| Requirements Planning) and MRP II (Manufacturing | | | | over time, but that’s the catch - how much |
| Resource Planning); although I realize that they are | | | | time? |
| distinct. I will address this in more detail later. | | | | In an effective control system, the feedback has |
| Here is some background... | | | | to be time relevant. For example, in your |
| I once worked in a process-based company | | | | car’s cruise control system, the feedback is |
| where several process engineers worked to keep | | | | virtually ‘real-time’. As the car senses a |
| the complex process operational. Over lunch one | | | | hill, it adjusts quickly. Imagine, if the delay |
| day, the engineers were discussing the pros and | | | | between heading down hill and easing off the |
| cons of feed-forward and feed back control | | | | throttle was two minutes? Would you define this |
| systems. Not being a process engineer, they | | | | as being “in control”? |
| were eager to rid me of my ignorance. | | | | Yet, this is exactly what happens in MRP II, the |
| A feed-forward control system uses pre-defined | | | | feedback is simply not responsive to the real |
| inputs to manage the system. A good example of | | | | world changes that impact production. |
| a feed-forward system would be your typical | | | | Thus despite the implied ‘closed loop’ |
| lawn sprinkler arrangement. You determine how | | | | name tag, MRP II is in practice a feed forward |
| much water you want, the time of day to start | | | | system, with occasionally tweaks after the fact. |
| the watering, which part of the lawn goes first | | | | Back to the sprinkler analogy, if we do hit a rainy |
| etc. All in all a very nice set up, unless of course, | | | | period, we may adjust the sprinkler system |
| it rains. I’m sure many of us have seen | | | | accordingly – but usually, a dollar short and a |
| sprinklers running in the rain and wondered why. | | | | day late. |
| That’s the pitfall of a feed forward system; it | | | | Contrast that to a pull system. Here the system |
| assumes that certain operating conditions are | | | | feedback is based on the material status itself |
| predictable. | | | | – if an upstream resource has insufficient |
| A feed back system, on the other hand, takes | | | | inventory, it automatically sends the signal to the |
| information of the existing process (often | | | | downstream operation to stop. In this case the |
| real-time) and adjusts inputs accordingly. For | | | | information and material are one in the same. Pull |
| example, in the sprinkler system example above, | | | | is a true closed loop system since the feedback |
| if we placed a moisture sensor to detect the | | | | mechanism is time relevant and the signal is very |
| current lawn moisture content, and programmed | | | | clear. |
| the sensor to allow water only below some set | | | | The real difference between push and pull is not |
| point, we would than have a feed-back control | | | | whether we are building to ‘demand’ or |
| system. Feed-back systems are much more | | | | to ‘forecast’. Rather the difference is |
| effective when the environment has several | | | | because PUSH is inherently a feed-forward |
| variables. The downside is that feedback systems | | | | system. So, let’s be clear that the primary |
| tend to be more complex. | | | | differentiator between Push and Pull is not what |
| Recently, a client of mine was enthusiastically | | | | triggers the process, but what mechanism is used |
| espousing the virtues of moving away from a | | | | to control it. In MRP II, the ‘feedback’ |
| traditional pull planning system (MRP) | | | | mechanism is generally unreliable or not timely and |
| and towards a Kanban-based Pull system. He said | | | | thus is better described as a feed forward |
| that using actual customer demand, as opposed | | | | system. Pull systems (whether Kanban, or Drum |
| to the master production schedule, would be a | | | | Buffer Rope) are designed to adjust to immediate |
| blessing. But you may know the familiar story… | | | | conditions. |
| “this will be GRRR-E-A-T once some of the | | | | Here are a few things to keep in mind: |
| kinks are worked out for getting it in place!” | | | | - Purge and clean up any WIP that is physically on |
| But the devil of these things is ALWAYS in the | | | | the floor and assure that the necessary MRP |
| details. | | | | transactions reflect this reality. |
| Working with clients over the years I hear this | | | | - If you have a resource that is at or near its |
| quite a bit. Push is passé while Pull is ‘in’ | | | | capacity, use this resource to implement traditional |
| because Pull is synched to true demand. Checking | | | | Input-Output control. The concept of Input-Output |
| the APICS dictionary we get: PUSH | | | | control is valid, but is too complex if used in more |
| -“production of items at times required by a | | | | than a couple of key resources. Pick one or two |
| given schedule planned in advance” and PULL - | | | | strategic resources and use these to pace the |
| “the production of items only as demanded | | | | inputs into the system. |
| for use or to replace those taken for use”. | | | | - Do not release new orders into the system |
| The great debate between so-called | | | | without validating that existing orders were |
| “Push” systems such as MRP and | | | | actually completed and out of the system. |
| “Pull” systems such as Kanban is old news. | | | | - Avoid the temptation of releasing any work that |
| Most practitioners by now would likely agree that | | | | does not have all the needed material or resource |
| demand pull is superior to push in that lead-time is | | | | time, since this order will invariably end up stalled |
| reduced, work in progress (WIP) is reduced etc. | | | | in the process and add to confusion on the floor. |
| However, the actual reason for the overall | | | | Knowing that MRP is not a true closed loop |
| superiority, at least in terms of scheduling final | | | | system can help guide improvement efforts. Here |
| assembly, is not what the common wisdom holds, | | | | is what you can do to improve your existing MRP: |
| it really has much less to do with | | | | The real secret of Kanban as a scheduling tool is |
| ‘demand’ as it does with | | | | in its inherent ability to choke off release of new |
| ‘feedback”. | | | | orders. This automatic ability of restricting WIP is |
| So, what does this have to do with Push and Pull | | | | the real Trump card of Pull. And while MRP is not |
| planning systems? | | | | as robust in doing this, with some awareness of |
| Despite the conventional wisdom that Pull and | | | | the importance of input-output control at least |
| Push is differentiated by the nature of demand, | | | | you will be heading in the right direction. |