Search By Topic The Green Supply Chain Distribution Digest
Supply Chain Digest Logo

Dr. Michael Watson
Northwestern University


Supply Chain by Design

Dr. Michael Watson, one of the industry’s foremost experts on supply chain network design and advanced analytics, is a columnist and subject matter expert (SME) for Supply Chain Digest.

Dr. Watson, of Northwestern University, was the lead author of the just released book Supply Chain Network Design, co-authored with Sara Lewis, Peter Cacioppi, and Jay Jayaraman, all of IBM. (See Supply Chain Network Design – the Book.)

Prior to his current role at Northwestern, Watson was a key manager in IBM's network optimization group. In addition to his roles at IBM and now at Northwestern, Watson is director of The Optimization and Analytics Group. 


August 30, 2016

You Don't Need the Optimization in Multi-Echelon Inventory Optimization


The Truth is That Multi-Echelon Inventory Optimization is Widely Misunderstood and Chances are, you Don't Need Optimization

 


Multi-Echelon Inventory Optimization (MEIO) has been a hot topic for the last 15 years—software vendors push it, industry analysts tout it, and customers demand it.

The surprising truth behind all this, however, is that very few firms really need the ‘optimization’ feature in MEIO. It is fair for you to be skeptical about this claim. Optimization is considered the ‘gold standard’ of planning and scheduling and multi-echelon inventory optimization has been ingrained in our heads for years. But let’s dig in a little further.

Watson Says...

The surprising truth behind all this, however, is that very few firms really need the ‘optimization’ feature in MEIO.

What do you say?

Click here to send us your comments

Optimization, at its core, is the ability to sift through innumerable combinations in order to determine ‘the best’ solution.  So how does this relate to our inventory problems?

The key is to understand where the “optimization” in MEIO actually takes place:  The math behind MEIO helps determine where your inventory buffers should be (and, to a lesser extent, how large they should be).  The levers that the optimization engine pulls in order to make these decisions are the fill rates at each node and the promised lead-times between nodes. It sets these variables to optimize the good of the entire system (or across all echelons).  Often, this results in some nodes holding very little inventory and therefore providing poor fill rates or long promised lead-times.  For example, it may be optimal for an upstream node to promise only a fill rate of 80% to the downstream node.  Or, it may make sense to promise that downstream node a lead-time of 16 days.

As a note, these models assume that the patterns of flows are fixed in the supply chain—that is, you know which plant provides components to which assembly plant.  The flow is not optimized.So why don’t you need this optimization capability then?

  1.  You might not have multiple echelons.  For these decisions to make sense you have to have multiple echelons that carry inventory.  Not just multiple echelons as part of the supply chain.  Many retailers only have one inventory echelon—they buy from suppliers and store products in warehouses.  Similarly, make-to-order manufacturing firms have only one pile of component inventory.   A retailer with a hub-and-spoke system or a CPG manufacturing firm with raw material and finished good warehouses would each have two echelons, although it’s still a bit of a stretch to call them “multi” echelon.

  2. The decision the optimization is making is not one you want to make. Most people don’t realize that the optimization is changing the fill rates or changing the promised lead-times from one node to the next.   You may not want to change these hard-fought commitments or attempt to reprogram the IT systems that link the echelons.

  3. The implementation may prove to be politically difficult. Even if you are, in theory, willing to change the fill rates or the promised lead-times, you may find that the implementation goes against every good instinct that your managers have.  For example, if it is optimal that an upstream node should serve a downstream node with an 80% fill rate, chaos may ensue.  The upstream node will have a low amount of safety stock (because their fill rate target is 80%).  The downstream manager may not quietly live with the 80% fill rate.  Either he or she will complain loudly to someone higher up or they will game the system and place orders for 125% of what they want so they actually receive closer to 100% of what they want.  And, the problem isn’t any better when you consider the upstream node.  There is more to good fill rates than just the safety stock—you have to make sure you have enough trained people and good enough systems to actually ship the products.  So, having a manager with a fill rate target of only 80% may not help you assess whether he or she is doing a good job on getting orders shipped when the inventory is on-hand.  And, managers who have been trained to do their best with fill rates may find that trying to stick to an 80% fill rate is not going to look good in their performance reviews. 

  4. There aren’t enough choices to require optimization.  At the most basic level, the multi-echelon optimization is telling you whether a node in the supply chain should be make-to-order (hold no inventory/have low fill rate) or make-to-stock (hold inventory/maintain higher fill rate).  In a CPG supply chain consisting of a buffer of raw materials and a buffer of finished goods, you simply have two scenarios—you hold raw materials or you don’t.  You don’t need an optimization engine to tell you which is better.  You just calculate the costs and benefits of the two cases.  Now, you may also have an opportunity to hold WIP and differentiate the finished good closer to when the demand comes in.  This gives you three choices to evaluate.  You may think this sounds trickier but the thing is, it is almost always obvious what the inventory benefits are going to be—it is better to hold inventory as raw materials, and if that isn’t possible, then as WIP is the next best choice.  Remember that optimization comes in handy when you have too many combinations to figure out on your own.  When you have 3 choices, you want to know the costs of each of the 3 anyway.

  5. You don’t make these decisions that frequently.  Finally, even if you still want the optimization feature, you don’t need it as part of the tool that you use to set inventory levels on a monthly basis.  It’s not likely your supply chain can handle the chaos introduced by changing your fill rates or promised lead times more than once a year.  So in reality you can just do a quick off-line study every 12-18 months and reset the buffers based on those project results.

So why does it exist? It must have some value…

OK, In a Few Special Cases You May Need It.  Multi-echelon inventory optimization has shown value.  In many cases, this is in very long, usually manufacturing-centric, supply chains with high cost items, a complicated BOM, lots of choices of where to hold inventory, and lots of common components for different finished goods.   

Final Thoughts
 

If I shouldn’t be focusing on the optimization in multi-echelon inventory optimization, where should I focus?


Admittedly this blog is focused on just debunking and clarifying the “optimization” feature of MEIO.  It’s important to point out that the “multi-echelon inventory” portion of these solutions is very important, however.  There is tremendous value in using the right science to determine how much inventory you should carry (not whether to carry or what the fill rate should be) at each node in your supply chain.  We should also point out that there are many interesting optimization problems within these supply chains that relate to inventory (but aren’t what people talk about when they say “multi-echelon inventory optimization”).  We’ll continue to cover more on those topics in future posts.

(A special thanks to Ganesh Ramakrishna, Larry Snyder, Sara Hoormann, and Austin Haygood for helping write and edit this article.)


Any reaction to this Expert Insight column?
Send below.


Your Comments/Feedback

Peter Cacioppi

Chief Scientist, Opalytics Inc
Posted on: Nov, 12 2016
"There are many interesting optimization problems ...that relate to inventory."

Typically non-linear optimization problems, no? 

So even if you concede that true MEIO has a niche utility (a debatable point in its own right) why hire a company that lacks the expertise to build such optimization? It seems far fetched to imagine that an organization that lacks the resources and know-how to build MEIO would be fully prepared to build more esoteric optimization engines covering similar terrain.

Edith

President, Opalytics
Posted on: Nov, 15 2016
While implementing the results of the Optimization part of MEIO can sometimes be hard, it is definitely worth while to go through the effort. We expand on this in "The Optimization of MEIO"
 
 
 
 
 

Features

Resources

Follow Us

Supply Chain Digest news is available via RSS
RSS facebook twitter youtube
bloglines my yahoo
news gator

Newsletter

Subscribe to our insightful weekly newsletter. Get immediate access to premium contents. Its's easy and free
Enter your email below to subscribe:
submit
Join the thousands of supply chain, logistics, technology and marketing professionals who rely on Supply Chain Digest for the best in insight, news, tools, opinion, education and solution.
 
Home | Subscribe | Advertise | Contact Us | Sitemap | Privacy Policy
© Supply Chain Digest 2006-2013 - All rights reserved
.