DDMRP: The Good, the Bad, and the Ugly

DDMRP: The Good, the Bad, and the Ugly

There has been a lot of activity on the LinkedIn feeds of this thing called DDMRP. Now that SAP has embraced it, I expect this activity to get much stronger. But what is it really? What does it do? Is it the magic elixir to solve all supply, inventory and service issues once and for all as its proponents boast? Or is it just another fad as its opponents claim? In this article I provide a more balanced perspective.

Note: DDMRP stands for Demand-Driven Material Requirements Planning, which was introduced and is promoted by the Demand Driven Institute (DDI).

The Good

Starting with the good. DDMRP does some things traditional MRP and many other traditional approaches did not focus on or ignored completely. Notably, a lot of approaches have historically been inspired by innovative system capabilities, not so much by innovative processes. Take for example MRP II, ERP, APS, FCS and similar system types. Implementations of such systems generally take the form of input what the supply chain looks like today, and they will output what you should do to manage the day-to-day operations. You tell the system where stock is held and relevant supply parameters, and it will tell you how much to stock there. DDMRP instead takes companies back to the drawing board. Taking inspiration from queuing theory, LEAN, JIT and other schools it enables companies to determine where product should be stocked and in what form. Its focus on flow and reduction of queues around and between each stocking location, allows for effective lead times to be reduced more than would otherwise be possible. A main objective is to reduce the maximum effective lead time between any two sequential stocking locations across all the steps a product traverses from supplier to customer. As soon as a sales order comes in for an item, every stocking location in its path independently and simultaneously determines if it needs to replenish or not. This behavior is why they dubbed the approach "demand-driven", the DD in DDMRP. Since the lead time to replenish is a significant factor of safety stock, minimizing the longest lead time achieves a minimal inventory coupled with fastest possible replenishment, as needed only. The beauty of this concept should be evident to those experienced in inventory management.

The other strength of DDMRP is that it provides an easy methodology. In principle it requires no specialized systems, just follow the rules. All the parts are provided: concepts, calculations, and processes. This makes DDMRP accessible to both companies without the budget for advanced systems and their expensive implementations, as well as those without good processes in place, or bad adherence to processes. Regardless of other benefits of DDMRP, taking a company's supply chain from chaos to a controlled state will in itself be of huge value to those companies, both financially and otherwise.

The Bad

DDMRP does a lot to reduce lead times, which is beneficial in any supply chain. However, there are some lead times that cannot easily be reduced, and these limit the extent to which DDMRP can provide value. One example is overseas supply. If it takes effectively 90 days to get product ordered and delivered from China to the USA what are the options to reduce this? Options may be:

  • Intermediate stocking points along the globe are costly and are not helpful in this regard. The ships themselves are moving stocking points, serving the same purpose, but better and cheaper.
  • Local sourcing is sometimes an option, but not in most cases. In those cases it is inevitable to have safety stock on the US side, and plenty of stock in-transit, all well before sales orders for the same are received.
  • Another option is to offload the sourcing to a third party, but that is only making it someone else's problem, not solving it.

This means DDMRP will be useful to companies that source everything locally, or could change their sourcing to do so. For all others, there is a big gap that must be addressed.

Whilst a strength of DDMRP is that it can determine WHERE product should be stocked, one of its biggest weaknesses is determining HOW MUCH to stock in such locations. Safety buffers are determined backward looking, using archaic logic with wet-finger parameter values. This is equivalent to using a naive safety stock, worse than using a traditional safety stock and certainly much worse than determining stocking levels through a multi-echelon inventory optimization. This is masked rather than addressed by the use of multiple so-called zones of inventory levels.

Whenever product demand patterns show trends or seasonality, inventory buffers should inherently flex along with the demand, otherwise service levels will fluctuate unmanaged. DDMRP acknowledges this fact through a seasonality factor that may be applied to the safety buffer levels. Since DDMRP takes no forecast as input, it should be evident this equates to using a naive forecast, worse than using a traditional statistical or judgmental forecast and certainly much worse than using a probabilistic forecast.

Similarly its use of so-called average daily usage (ADU) is nothing more than a naive forecast.

And this brings us to the ugly...

The Ugly

The ugly side of DDMRP actually has nothing to do with its concepts. Those are solid, with its strengths and weaknesses, similar to any other approach. The ugly of DDMRP is in how it is promoted and marketed, with lots of exaggeration, contradictions, and outrageous claims. This has very much been the case in the past, but still is to a degree today. In my opinion this is the primary reason DDMRP is getting so much push-back from the industry experts and practitioners. Some outrageousness in marketing is fine in that it sparks discussion, and brings attention of a new concept to a larger audience. A moderate dose of outrageousness will allow a majority of experts to overcome their initial knee-jerk reactions and accept, maybe even embrace the new concept in time. In talking to many experts I find that the sheer magnitude of outrageousness of the claims made by the DDMRP crowd, especially the early claims, have made lifelong enemies of many such experts.

I will not hash out all the faulty claims, since that would fill another few articles. But I would highlight one that has practical implications. This claim, still maintained today, is that with DDMRP you do not need a demand forecast. This claim has been nuanced more recently in that you still need forecasts for long-term planning and capacity planning, but not for replenishment. From the above it should be clear this is only true for a small subset of companies that can make ALL their lead times so short to become effectively a make-to-order business. For the majority of companies this is a falsehood. Then even for those companies for which this claim may be true, not using a forecast leads to significant inventory inefficiencies for all items except the ones with the most steady fast-moving demand patterns.

Another downside of this sharp demarcation is that it dooms every DDMRP implementation to become an additional functional silo in the supply chain, where most companies are working very hard to remove the silos they have already.

It is clear though why DDMRP followers would persist this claim. First, its outrageousnous gives it traction and visibility, sparking discussions. This may even have been the direct impetus to SAP embracing the concept. One cannot fault them for doing so. Second, it is my opinion that they simply haven't figured out how to make it work with a forecast. Unfortunately, it perpetuates an us-versus-them situation, preventing more experts and practitioners to embrace it. It does not need to be this way however; in a future article I will lay out the missed opportunity of how DDMRP and demand forecasts can not only work side-by-side, but in fact enhance each other.

The term DDMRP itself is both a misnomer and misleading. Without going into all detail, again I wish to highlight only one issue of this. DDMRP is not a flavor of MRP. It is again clear why the term was chosen: existing MRP implementations are the primary target to rip and replace. And again DDI cannot be faulted for picking this name for this reason. However, all the boasts of value it brings are in comparison to the value of MRP. That is setting the bar so very, very low. It is like boasting to a college math major that you are better at math than a third grader. Similar and even greater value than DDMRP has been achieved for decades compared to MRP. As early as the 1990's the APS systems introduced (the real ones: FYGIR, Manugistics, Numetrix, not the big ERP tag-alongs of the time) were all providing financial value of similar size as DDMRP is claiming today. On this count too, I believe DDMRP does bring additional value, but they are short-changing themselves by not comparing to the best of the rest. Whilst DDMRP brings similar order of magnitude of value, it does so with simple rules rather than the brute force of complex expensive systems. It also does a better job at controlling the bullwhip effect than those systems did. Its benefits are not as impressive as forecast-driven multi-echelon inventory optimization (MEIO) offerings of today though, caused directly by the shortcomings described above. If you already have such systems the incremental value of DDMRP may be non-existent or even negative, if you were to replace them.

However, there is good news here too. The value both approaches bring (DDMRP, MEIO) are for a large part complementary. This would indicate that combining both approaches would bring even greater combined value.

Conclusion

DDMRP is built on strong core concepts and it has a lot of potential. However, the way it is marketed and promoted today pretends to be more than it is. They paint a black-and-white picture where shades of gray are possible. This is unfortunate. It makes opponents of people that could be proponents. It allows them to hammer away at weaknesses that do not need to be weaknesses, and it draws attention away from the strengths that it certainly possesses.

Rather than us-versus-them I believe a hybrid model is possible. DDI and other DDMRP evangelists will need to open up to this possibility before it can happen. Until then, I would strongly caution companies considering DDMRP to validate it could work for them and question every boast made about it. If you have very long lead times, question if it is at all possible to reduce the longest ones and by how much. Is it sufficient to cover order lead times? Are you already in a state where supply chain processes are under control? Then DDMRP value may be very limited. Use what makes sense, but do not go all in until you have ascertained it is not a dead-end street.

As mentioned above, I will publish more articles in the near future that explain in increasing detail how DDMRP can play nice with demand forecasting and inventory optimization systems. Maybe it can persuade DDI of the same...

UPDATE:

Following the discussion caused by this article I have published a rectification:

DDMRP: A Fistful of Considerations

Find all my articles by category here. Also listing outstanding articles by other authors.

John Allan Loucks

Heart driven leader. Developer of people, teams and organizations. AI innovator. Visionary who can reverse engineer the idea into execution.

4 年

Well this is completely untrue.

回复
Sam Graham

Taking a break from Linkedin; but I'm an ERP observer, blogger and author. Please contact me if you think that I might have knowledge that is useful to you.

7 年

A very valuable contribution to the debate: that you Stefan. I'm still reading the second book, but I am seeing DDMRP as being an excellent tool, some of the time. I want it as an option in my tool bag but I'll pick and choose when I use it.

Stefan de Kok

? Supply Chain Innovator ?

7 年

for everyone following this discussion, I have just published a formal rectification here: https://www.dhirubhai.net/pulse/ddmrp-fistful-considerations-stefan-de-kok/

Ian Henderson

Interim Manufacturing Line and Project Manager, Consultant. Manager of people and sophisticated systems; equally adept at changing things so that sophisticated systems aren't needed.

7 年

One glaring point of contention made here is that implementations other than DDMRP begin with current inventory profiles, which are then used to set up the new system. If this is the case then the implementation is being badly managed. Replicating the current model in the future vision is nonsense. We all know the term “keep on doing what you’ve always done and you’ll keep getting what you’ve always got” Implementing any Planning & Control tool has to be based around how you want to run the business. That means understanding the system approaches but also the relevant aspects of Lean and BPR. Dare we say that certain packages often end up being seen as an exercise in IT? That's when companies start with the current and waste money by missing the key element of business improvement. Assessing DDMRP based on the future with any approach replicating today falls into

Good contribution to a debate with some fair points but also some huge factual errors. Let me just make a couple of points if I may please. What the DDI have created is a framework, it is possible to extend the model and adapt it. Orchestr8 have done this in a number of ways, including using a structured mathematical approach to setting inventory buffers and using demand profile segmentation to select the right planning approach at the right time - as you rightly say one approach does not work ALL the time! We embrace flow based planning, forecast based planing and DDMRP demand pull to name a few. We also arent very big fans of MRP and uniquely our clients have the option of switching their legacy MRP tools off completely. Useability is a key factor and I agree with the point about creating another bolt on to legacy systems - planning should be done in a single environment. Now for some negatives. DDMRP DOES NOT reduce lead times. Positioning stock buffers allows each element of the supply chain to operate autonomously and reduce the effective planning lead time as you dont need to look out across the entire supply chain as MRP tries to do. This is inherently a better way to plan, delivers real agility and does away with the bull whip effect. Implementing buffers should be a strategic decision based on time to market and whether the item is to be replenished or not. It works for ALL items - lead time is not a factor in this decision. We see 30%+ inventory reductions with service improvements in cross - continent environment with lead times in months not days. In fact if you think about it, the longer the lead time the worse the forecast tends to be and the better the benefits can be. This point highlights a real understanding problem the author has with buffers - to say they can be linked with Inventory Optimisation is another example and is just false - they are totally different! And a buffer based system will deliver better service and less stock by a significant margin whilst also delivering an easier to run and understand process to the user. We embrace variability and prepare for it - we dont predict it and mitigate "wrongness". Finally, the forecast issue. The author appears to have missed the point entirely. Forecast is ALWAYS used in DDMRP. As in a capacity planning process rough volumes are used to size a capacity which is an ability to make a range of volumes. Inventory buffers at the same - we use an average volume to set the capacity of inventory to allow the supply chain to manage unforecastable variability in demand across an average volume. So this is just misinformation. To conclude we have absolutely seen benefits of the order marketed by the DDI. Our clients see 30 - 70% inventory reductions, better service, more capacity available, reduced operating costs and huge reductions in non-value activity based waste. EVERY time!! and this happens in EVERY environment, including where APS systems have been successfully implemented. Welcome to the debate and I look forward to a little more accuracy from all sides.

要查看或添加评论,请登录

Stefan de Kok的更多文章

  • The Futility of Mapping Forecast Error to Business Value

    The Futility of Mapping Forecast Error to Business Value

    A number of years ago my mom asked me if I held it against her that she beat me so much as a child. My honest answer:…

    21 条评论
  • Safety Stock vs Inventory Optimization

    Safety Stock vs Inventory Optimization

    Many companies go about setting safety stock levels the wrong way. Often this is driven by a misunderstanding by upper…

    24 条评论
  • You Think You Understand Safety Stock?

    You Think You Understand Safety Stock?

    There are a lot of misunderstandings about safety stock. Most people will quibble about what formula makes a safety…

    9 条评论
  • Why Not To Use The Normal Distribution

    Why Not To Use The Normal Distribution

    If you use demand variability, forecast error, or lead time variability to determine safety stocks or other buffers…

    39 条评论
  • Your Forecast Is Already Probabilistic

    Your Forecast Is Already Probabilistic

    If you are like most forecasting practitioners, you may believe probabilistic forecasting is extremely difficult to do,…

    10 条评论
  • How to Measure Forecastability

    How to Measure Forecastability

    In recent weeks the same question in the title above has come up repeatedly. Since I like the DRY principle (don't…

    30 条评论
  • Is Your Accuracy Lagging?

    Is Your Accuracy Lagging?

    A problem I often encounter is companies measuring their demand forecast accuracy incorrectly. Most demand planning…

    37 条评论
  • How to Integrate Information

    How to Integrate Information

    Removing Silos in Planning - Part 6 This article explains how to remove information silos in an easy incremental way…

    5 条评论
  • Decision Silos

    Decision Silos

    Removing Silos in Planning - Part 5 In previous parts, we described process silos, data silos, and information silos…

    2 条评论
  • Information Silos

    Information Silos

    Removing Silos in Planning - Part 4 So, you fixed your process silos and data silos. All your processes are aligned in…

    3 条评论

社区洞察

其他会员也浏览了