Supply Chain Shaman

Close this search box.
Lora's Latest Post

Better Math makes Better Networks

While supply chain leaders design and fine-tune their networks, supply chain laggards inherit and accept their supply chain networks as a given.  Leaders understand the power of strategic planning.  However, this realization has not happened quickly; instead it has happened VERY slowly over the twenty-five years that I have been studying supply chain management. 
Few people understand this better than the team at Chainalytics.  I recently ran into Jeff Metersky, one of the founders of Chainalytics, at the S&OP IE event in Las Vegas.  During the event, we shared a cup of coffee and discussed the history of network design software, projects and technology capabilities.  I liked Jeff’s insight SO much that I thought that I would share it with my readers.  This blog is written in a Q&A format.  The referenced insights are based on Chainalytics’ work over the past nine years.

Q: Jeff, tell us about your background.
 I graduated as an Industrial Engineer from the University of Illinois in 1986. My first job was with the AC Spark Plug Division of General Motors.I was one of the first industrial engineering hires into a materials management group. At that time, there was no supply chain group. We were the closest thing to it and we reported into manufacturing. 
At that time, I did not know much about transportation, logistics or warehousing. This job helped me learn, and as I became more interested, I got a MBA in materials and logistics management from Michigan State. After graduating with my MBA, I worked for IBM in Princeton, NJ and was part of a corporate transportation and distribution group supporting the United States market.  This is when I was first exposed to strategic network design work.  For my first project, I used Bender Management Consultants and their network optimization offering; and 4.5 years later, I went to work for Paul Bender. I ran Network Strategy projects for them for about 4 years. 
My next job was with Inter-Trans Logistics (ITLS). I was in charge of professional services for the decision sciences suite (Supply Chain Strategist, Transportation Modeler, Carrier Bid Optimizer) which was sold to i2 in 1998). After the acquisition, I had a similar role at i2 Technologies and for a short while focused on Product Lifecycle Management (PLM).  Nine years ago, I joined with Mike Kilgore as a founder of Chainalytics.  So, you might say that I have spent a lifetime in the industry doing strategic network design work.
Q. Over your career, how has the role of network strategy changed?
In the beginning, there were VERY few supply chain organizations.  And, when someone did speak of supply chain back then, it was not holistic. The discussion was more limited and focused on a function.  The notion of a supply chain organization and a Chief Supply Chain Officer is relatively new, and strategic network projects have changed in five ways:

  1.  Holistic.  Companies now embrace a more holistic view of the supply chain.  While it used to be a focused on functional excellence for a single organization (source, make or deliver), it is now about making the right trade-offs between source, make and deliver.  This shift is major but has only happened very recently at more than a handful of companies.
  2. Global.  Early on, the scope of network design projects was very regional.   Today, most of our projects are dealing with a more global view and greater intricacies.
  3. Systemic.  The first projects were ad hoc.  Often the genesis of a project was to analyze a capital expansion or a new product introduction.  Now it has become more of a part of the ongoing business planning activity.
  4. In-house.  Twenty years ago, the projects were rarely done in house and they were stand-alone one- time studies.  Now the best companies have in-house strategic teams with deep core competencies always looking for ways to tweak the network. About 25% of our clients now have some form of in-house capabilities and they typically report to supply chain planning or supply chain strategy organizations.  It is not part of a corporate strategy group.  We see this capability the most often in consumer products food and beverage manufacturers and the big box retailers.  For a company, to have an in-house function, it is usually greater than 3 billion in revenue with six people gainfully employed.
  5. Deeper Analysis.  Historically it was small projects on an ad-hoc basis, now it is ongoing with quarterly refreshes. The analysis is 2-4 times a year to refine the big picture.  And, it is at a much more granular level.  Today, technology is less of a barrier to understand the impact of variability, constraints and flows. We play the roles from strategic advisor—anything new and different to staffing augmentation—to process outsourcing. (Some of this change is outlined in figure 1).

Today, as outlined in figure 1, most companies have completed a network analysis.  It is no longer about excess capacity, instead it is still all about reducing costs.
Figure 1:  Change in Network Design Analysis

Q.  How have technologies changed?
The technologies per say have not changed. Most are still mixed integer linear programming. The change has come with what we can do with them. Based on the changes in computing power, we can run bigger models with more detail.  In the past, due to computing limitations, we could only analyze a simplified model:  a look at only a part of the supply chain with the other parts “locked-in”. Today with these changes, I can model develop more granular models and analyze more trade-offs simultaneously.  The power is in the views and the ability to quickly calculate “what-if analysis” to evaluate trade-offs. Hardware and computing power have gotten better, and we are able to push the boundaries for discovery to see new opportunities.
I am also excited about some new and emerging technologies.  Axxom is a company out of Germany.  It offers an alternative to the mixed-integer programming approach called CSA that allows us to solve bigger and more granular problems. I also like Solvoyo. This is a start-up that is placing solvers in the cloud to have deep computing horsepower to throw into the problems.  Today, we need two tools: one for inventory optimization and one for network strategy; however Solvoyo (SAS model all in one) is changing this to enable a single solve.  They leverage the computing horsepower and proprietary algorithms to do it in one solve.  Additionally, LLamasoft’s technology has finally matured.  Previously the company was just known for their simulation abilities; but today, they offer solid network optimization, and they are the most progressive in continuing to advance the strategic design software of a named known player.  The established players –JDA, I2 Technologies, IBM, INFOR and Oracle—have not changed much.  The new players are continuing to move forward and reset the bar.  They have closed the gap on the leaders.  Each are doing creative things to move forward and execute broader and deeper analysis.  They are also finding a way for the non-power user to access to the power of the technology while decoupling the differences between network design and flowpath optimization. I am excited that data models work together and it is easy to use.  Roll it up automatically and focused the energy.  Today, we are looking at having network analysis for the masses…
Q. With the changes in computing power, should network design been done at a granular level? 
There is a fine line.  We like to run the models deep as we can and but then view the data at an aggregate level to develop conclusions and identify trends  The tools now allow us to do this easier.  The more and more you aggregate the less confidence that people have in the answer.  This is particularly true in manufacturing.
However, I have observed, and I think that it is legacy and old paradigms, that people are still modeling at aggregated levels and not pushing to become more granular.
The more that you are trying to refine , the more that I need the detail. It is behavior that matters.
Q. Have you ever worked with a company where the solution was counter-intuitive?
Yes, these are fun projects. It happened during the first project that we did as Chainalytics.  Basically, the company was planning to make a significant capital investment in their flagship and oldest plant: a manufacturing facility that had been around for 50 plus years.  It had multiple level manufacturing, vertical, and at the same time a change was happening in leadership.
When we started the project, the client’s question was “Do I need all these plants?” A couple of paradigms existed:  certain products were being outsourced based upon potential cost savings and the inventory build strategy dictating when in the year inventory should be built.  We found their policies and strategies to be false, including closing the plant they wanted to invest in (see figure 2).
At another company with a high number of shipments not coming from their primary DC’s, and dependent on out-of-region shipping, we discovered that while the company thought that their issue was a “follow the sun strategy” to support high levels of next day shipping, the real issue was that inventory deployment strategies and stocking levels were the drivers.  We asked two questions:
1) What do I do for optimal deployment? What should be stocked where with an appropriate inventory carrying position to minimize overall delivered cost?
2) What is the right level of service to maximize the profit contribution?  They are still using the approach today.
Q. In your work, what is your biggest learning?
I have learned that it is not always about the BEST MATH ANSWER.  Optimization allows insight into possible solutions.  I now know that organizations need to chose the answer that best improves the business (feasible, aligned with strategic initiatives and growth, reflects change management issues and future assumptions), but not necessarily the best mathematically optimal.
Q. What is your recommendation for the reader?
Don’t think of network design it as a project.  While this may be a good reason to initiate it, the greatest value happens when it is an ongoing process. I am still learning and it is fun.
Network design analysis has slowly gained adoption.  I agree with Jeff that while it cannot happen without better math; the best answer is not always the best math.  The good news is that the math is better, the computers are more powerful, and teams are now more willing to use it.  This combination is giving rise to the Business Process Outsourcing (BPO) market for firms like Chainalytics. 
What insights can you share about your work with network analysis?  Or technologies that you think are promising?

Search the Archives
Share this Post
Featured Image
Recent Posts

Redefining Your Relationship with Data

The journey to becoming data-driven requires owning the cycle of ingestion, storage, and usage. The adoption of new technologies makes this easier and drives process flexibility in schema and the preservation of semantics.

Read More »

Rise. Or Rise-up?

The sharing of concerns by the business leader on the evolution of SAP HANA to SAP RISE as heard at the Consumer Goods Forum. Here I share the concern, some options, and some background.

Read More »