The Push vs. Pull Diversion Diversion

I’m still scratching my head over an IndustryWeek article titled The Great Push vs. Pull Diversion by By Edward S. Pound and Mark L. Spearman of Factory Physics, consultants and writers of one of my favorite Lean books. They state:
Executives and managers in manufacturing have been subject to a great diversion ever since the advent of the Toyota Production System. In an effort to improve performance, many have wasted inordinate amounts of time and money in organizational struggles over push systems versus pull systems.
* If you are discussing the benefits of push vs. pull in your organization, stop.
* If you are implementing pull systems and focusing on achieving one-piece flow, know that you are in danger of leading your organization to decreased throughput performance, poor customer service performance, or both.

The authors ask the reader to “Understand that whether WIP is pulled or pushed is not the point and that too little WIP is as bad as too much WIP” and they make the points that we should not ignore the practical science underlying WIP levels in a push or pull system, and that this so-called “push versus pull diversion” is the result of confusion surrounding the definitions.
They proceed to criticize the definition of pull systems give by James Womack, and the fact that the supermarkets in which we shop do not appear to be a true pull system. The supermarket is a metaphor, not an operational model for a pull system. The author’s don’t offer an authoritative definition of push vs pull (perhaps because this is a discussion we should stop having, according to them).
Let’s clarify: the difference between push and pull is not between the quantity of WIP in the system. That is simply one of the end results of converting from push to pull. In fact, in some rare cases we have actually increased inventory levels in order to implement kanban systems.
In general terms, there is MRP-executed planned push production and kanban-driven downstream pull production. In a kanban-based pull system the quantity of kanban cards (and therefore WIP) is calculated based on usage, replenishment lead-time, container quantity, and something called the safety coefficient which takes into account the practical science of process variation mentioned in the article. The quantity of kanban cards in the system is adjusted based on the changes in monthly demand. The push system produces or moves product at the convenience of the producer process. The pull system only produces or moves product at the request of the downstream (customer) process.
Although a kanban system by definition must be a pull system, not all pull systems are kanban systems. You can keep your WIP levels exactly the same and convert form push to pull. You will stabilize your production output by stopping overproduction, as well as identify process problems. When you push, you are making production information and the shop floor condition obscure by de-linking material and information flow.
The key difference is in how information about consumption of the product is given to the producer process. In a pull system, material and information flow are linked. In a push system, they are not necessarily so.
The authors give two examples of companies facing risk through converting from push to pull (the billions saved going from push to pull must be folklore):
In blind pursuit of one-piece flow, many companies have driven throughput levels down to unprofitable levels. Since true one-piece flow is only possible with zero variability, and the laws of nature do not allow zero variability, true one-piece flow is an impractical ideal.
I would remind that all ideals, by definition, are impractical. If you achieve your “ideal” that just means you set your sights too low. Furthermore, since our clients have put one piece flow into practice, it may be a practical (that is to say attainable) ideal, at least to some.
The authors state:
Telling employees that they should “pull to demand” or “pull work only as needed” has a nice intuitive ring to it. Unfortunately, as we have discussed, intuition that is not fact-based can go badly wrong.

Actually we have found that the idea of pull is profoundly counterintuitive to the vast majority of people who we have introduced it to over the years. Pushing feels safer, more convenient, and the way things end up if you let entropy go (if you don’t manage and improve systematically towards the ideal).
The authors are correct in that less WIP is not always better, but zero inventory and one piece flow still is the ideal. The Toyota Production System teaches us to improve towards the ideal, not to apply a theory based on practical science and be satisfied that push is as good as pull. If you are making changes without insisting on improving towards an ideal (takt, flow, pull, built in quality, etc.) then please don’t call it Lean, or refer to it as TPS.
Please continue to have this discussion until it is clearly understood that pull is superior to push, and then return to your study of the practical science of Edward Pound and Mark Spearman.

12 Comments

  1. dave

    September 16, 2007 - 12:36 pm

    I have read Factory Physics, and i know it is obvious (by the practical science discussed in this book) that Pull system is superior to push. it was clear enough. but they have written about WIP in more details than other books i have ever seen. as you have absolutely know, in F.P., WIP is being capped, then it will be reduced by decreasing of variability. most of technics it is used to decrease variability, are in the lean tools packages plus continuous improvement.
    as i see, both of them – traditional lean books and factory physics – accept pull system is better than push system, however it is implemented.
    have i been misunderstood?

  2. Mike

    September 16, 2007 - 1:07 pm

    Isn’t it ironic that working towards the ideal is more practical!
    So many things are theoretically equivalent when you have perfect knowledge.

  3. Ed Pound

    September 17, 2007 - 8:41 am

    If Dr. Spearman and I have been unclear in our explanation of the use of push vs. pull terminology in our Industry Week article (http://www.industryweek.com/ReadArticle.aspx?ArticleID=14900), hopefully we can clear up any misunderstanding. Also, Mr. Miller has made a number of misstatements about the article content and I would like to address those here.
    First, some background. I have over 18 years manufacturing experience and, until I started working with Factory Physics Inc., had spent all my professional life in manufacturing operations and management. This has included:
    – two and a half years in Japan working in production at a Japanese company—Kaizen was a daily occurrence
    – Lean transformation work with Shingijitsu at AlliedSignal (now Honeywell)
    – Supervision of Six Sigma efforts at a line management and executive management levels.
    Dr. Spearman has over 29 years experience in research and consulting with manufacturing companies and as an award-winning professor at Northwestern University (where I was a student of his) and at the Georgia Institute of Technology.
    We understand the use of the classic concepts of push and pull and have used them in practice. Our work with Factory Physics concepts has shown us that there are improvements to be made in the terminology, understanding and practices of operations productivity improvement which led us to write the Industry Week article.
    The following is incorrect, “Edward S. Pound and Mark L. Spearman of Factory Physics, consultants and writers of one of my favorite lean books.” Mr. Miller doesn’t name the lean book but the assumption here is that Mr. Miller is referring to the book Factory Physics.
    – The book “Factory Physics” was written by Mark Spearman and Wally Hopp. I was not one of the authors.
    Mr. Miller states: …”The author’s don’t offer an authoritative definition of push vs pull (perhaps because this is a discussion we should stop having, according to them).”
    – In fact, in the article is the following: “A clear definition states, ‘A push system schedules the release of work based on demand while a pull system authorizes the release of work based on system status.’” This definition is out of one of Mr. Miller’s favorite books, “Factory Physics” 2nd Edition, page 162.
    – The book “Factory Physics” is a fairly authoritative source. It is widely used in industry and at over 100 academic institutions to teach the practical science behind manufacturing management. It was also selected as Book of the Year by the Institute of Industrial Engineers shortly after it was published. A third edition should be coming out at the end of this year or in early 2008.
    Mr. Miller states, “Let’s clarify: the difference between push and pull is not between the quantity of WIP in the system. That is simply one of the end results of converting from push to pull.”
    – This is exactly the kind of statement that causes emotional diversion discussions that the article was addressing. The article states,
    o Do not be diverted by underproductive discussions of push versus pull. The practical science that governs your operations performance is not as simple as push versus pull.
    o WIP control is a strong determinant of performance and, in the real world, less WIP is not always better.
    – The article never states that the difference between push and pull is the quantity of WIP in the system. The article states that WIP control is a fundamental control for determining operations performance so whether you call it push or pull is irrelevant. Since the discussion of push versus pull typically leads to these type of discussions, we recommend companies talk about WIP control, understand the science behind WIP control and avoid this type of debate. Our experience with leading companies around the world has shown that this leads much more quickly to achieving the results that companies want. After all, companies don’t want to be the leanest or have the lowest variability or to be like Toyota, companies want to make profit over the long term. Toyota is just one company that has achieved long term profitability.
    Mr. Miller’s last statement exemplifies the type of approach we are trying to avoid, “Please continue to have this discussion until it is clearly understood that pull is superior to push, and then return to your study of the practical science of Edward Pound and Mark Spearman.”
    – Simply saying, “Believe what I say,” is not a learning and improvement approach. We are saying that the use of the terms push and pull is not the point. WIP control and the mechanics of WIP control are determinants of operations performance. So understanding the practical science that describes WIP control is a competitive advantage in improving operations performance. If an approach is to be objective and productive, conjecture and refutation are encouraged to ensure the usefulness and credibility of the approach. To quote Daniel Boorstein, “The greatest obstacle to discovery is not ignorance—it is the illusion of knowledge.”
    – Finally, we are major supporters of good Lean Manufacturing practices. Indeed, many companies have been very successful with Lean Manufacturing techniques. Both Dr. Spearman and I have worked successfully with many companies’ Lean Manufacturing personnel and we are strong supporters of good productivity practices. It is our experience that the use of the terms push and pull are typically loosely defined and tend to divert operations strategy discussions, manufacturing logistics design efforts and operations productivity improvement efforts towards the type of discussions illustrated here whereas a clearly defined practical scientific approach moves companies more quickly towards the results they desire.
    For an even more thorough discussion of the history of push versus pull, go to the following link: http://www.factoryphysics.com/member.cfm?mode=default&view=whitepapers and download the white paper titled, “To Pull or Not to Pull, What is the Question?” A two minute registration process is required. This paper was commissioned, refereed and appeared in the journal of Manufacturing and Service Operations Management in 2004.
    Sincerely,
    Ed Pound

  4. Jon Miller

    September 17, 2007 - 9:16 am

    Hello Mr. Pound!
    Thank you for continuing this discussion on push vs. pull.
    Factory Physics was indeed the fine book I was referring to, apologies for not attributing authorship correctly.
    You and Dr. Spearman are clearly authorities, and I did not mean to imply otherwise. Yet we have some differences of opinion, and I fully respect yours.
    At no time above did I say “believe what I say” any more than you and Dr. Spearman did in your article.
    Unless you disagree that pull is superior to push, I would think that as supporters of Lean manufacturing, you would want people to continue having this discussion, until they see that pull is superior. That is my central point.
    WIP control is wonderful, but if you are on the never-ending Lean journey, TPS is your target, and pull is the ideal to pursue through kaizen.

  5. Mark Graban

    September 17, 2007 - 4:00 pm

    Dave is correct — Factory Physics (the book) does discuss push vs pull in some detail, but the key is limiting WIP. Most push systems (run by ERP or otherwise) assume infinite capacity and will “stuff the goose” by continuing to cram more into the system, which increases WIP beyond the optimal level and increases CT.
    The “CONWIP” system (as described in the book) is a hybrid push/pull system, were you only start something new into the process when something has finished, but you might “push” within the process (say an assembly line). Having “flow” in a “push” process that has its WIP limited by the CONWIP rules (a different type of pull, since the start into the process is authorized by a card/signal) can be very efficient. The Factory Physics rules help you calculate the optimal WIP for the given system. But, they don’t assume the system’s variation is a given (as push manufacturers would), they would recommend reducing variation by using Lean and Kaizen methods.
    Let’s not bow down at the altar of “pull.” What’s Womack’s phrase, “flow where you can, pull where you must”?? Pull isn’t a cure all and pull isn’t the right approach for all systems (try it in semiconductor fabrication… not practical).
    I think the article, and Pound/Spearman, are correct that adopting “single piece flow” pull systems without first addressing the underlying variation can be suicidal to a business. I’ve seen it first hand.
    (Full disclosure, I was a student of Spearman’s as an undergrad at Northwestern, but my defending them isn’t out of personal loyalty, but out of a belief that their ideas are correct).

  6. Jon

    September 17, 2007 - 10:04 pm

    There’s nothing wrong with bowing down in respect for pull. I have no respect for push, as it takes us in the direction away from ideal.
    The quote you’re looking for is by John Shook:
    “Flow where you can, pull where you can’t.”
    Lean production does not advise push.

  7. Mark Graban

    September 18, 2007 - 4:26 am

    A toyota assembly line is not strictly “pull.” A continuously moving line is a “push,” one could argue, no?

  8. Jon

    September 18, 2007 - 5:39 am

    The difference between push and pull is not how much material there is between upstream and downstream, or how it is physically moved. Just because you stand at the front of a cart full of materials you don’t need and “pull” it downstream does not make it a pull system. Likewise standing behind a cart of material that you need next and “pushing” it downstream to your workstation would not make it any less of a pull system.
    A continuous moving line would be push if it is literally continuous, and never stops even if there is no demand or if there is a problem downstream. The paced conveyor lines at the engine assembly and the vehicle assembly plants at Toyota pull one piece at a time.
    Even in their machining lines they have something variously called full work system (a.k.a. two-point control, or AB control) which stops the machine from making and moving the next piece if there is full work on the line (no gap in the staged work on the conveyor). This is also pull.
    A push line would keep on producing and moving product downstream regardless of whether the downstream process is taking what the upstream produces. A pull line stops if the downstream does not take what the upstream produces.
    I have only seen pull assembly lines at Toyota. That does not mean that push assembly lines do not exist at Toyota, but it would surprise me greatly.

  9. Ed Pound

    September 18, 2007 - 8:38 pm

    Jon,
    A few more points:
    > TPS is not the target. The target is to be profitable over the long term. Toyota is just one company that has been profitable over the long term.
    > All of the discussion after my previous posting reinforces the point of the Industry Week article. As an exercise, try describing how an assembly line works without using the terms push or pull. For instance, Toyota and most all automotive assembly lines keep WIP and cycle times low because the stations on the assembly line limit the amount of WIP that can be in the line at any time. Tact time is tied to demand and setting the assembly line speed paces all operations to tact time and thereby provides throughput to meet planned demand (note that planned demand is almost never equal to actual demand). If the assembly line hits the tact time day in and day out, cars will be produced to planned demand with minimum cycle time and minimum amounts of WIP. Of course, we know from Little’s Law and other basic Factory Physics principles that WIP and cycle time are essentially equivalent so limiting WIP limits cycle time. Note that this description of the basic logistical performance of the Toyota assembly line was completed without using the p words.
    This is pretty straight forward for automotive assembly lines but the simplicity of the logisitics does not appear as well in most other industries where there is no assembly line to:
    > provide a pacer for workflow
    > act as a cap on WIP
    Then, the problem becomes determining what level of variability is acceptable and, given that level of variability, determining the minimum amount of WIP required to get maximum throughput (the Optimal WIP zone).
    Thanks for the opportunity to address these issues.
    Regards,
    Ed Pound

  10. Mark Graban

    September 21, 2007 - 9:24 pm

    A toyota assembly line has segments that move together with buffer zones in between. The one zone will keep producing until the buffer zone ahead of it is full. So is the line pushing until WIP hits a certain limit? Or is the next section “pulling” when the previous buffer is full?
    Does “pull” or “push” matter in that example?
    When Toyota uses a supermarket to “pull” parts via kanban, that’s clearly an example of pull instead of push (building components to an MRP schedule) right?

  11. Jon

    September 21, 2007 - 9:37 pm

    Hi Mark,
    Those buffer zones are more like hand-off zones. If I recall correctly, there are no extra vehicles in these zones to act as “buffer”. The hand off zones are there to allow the line to keep moving even if an andon cord is pulled, so long as the problem is solved before exiting that hand off zone.
    The Toyota assembly line is a fixed position stop system rather than a continuous push conveyor or a pull-started conveyor. Even if there was one piece of buffer, if the line remains stopped unless that piece moves, it is a pull system. I’ll ask our Toyota guys for clarification.
    Yes, the kanban system using a supermarket is a pull.

  12. Ron Pereira

    September 28, 2007 - 6:37 am

    Wow, what a conversation!
    Any western manufacturing company should set a goal of TPS. I am shocked to hear someone argue otherwise! If someone is doing it better than Toyota pray tell.
    Also, I will attempt to address some points from my simple “knuckle dragging” manufacturing mind.
    At a minimum, we need to pull “between” processes and flow “within” processes.
    This is NOT the same thing as pushing within a process (i.e. assembly line) since takt paces production and standard WIP (easily calculated just search Jon’s blog for an extensive article) controls inventory levels.
    However, if we can flow (NOT PUSH) between AND within processes this is optimal as Mr. Shook seems to be saying with his excellent quote.
    This is to say our pacemaker is at the front of the factory and it gets its production signal to start (from customer) and then things flow all the way through one piece at a time with standard WIP in place throughout the factory. This is the furthest thing from push which preaches producing on an island with no regard to what downstream processes need.
    One could argue this system (pacemaker at front) is still pull since we don’t start production until our customer asks for the product. I couldn’t and wouldn’t argue this point.
    Finally, Mr. Pound… it is spelled “takt” time not “tact” time.