Written by 9:32 pm analytics, Supply Chain Excellence • One Comment

Welcome Puff, The Magic Dragon, to Your Team

Remember the Peter, Paul and Mary song, Puff the Magic Dragon? It was a song from my youth. While there is a raging debate on song intent—Was the goal of the song to celebrate a children’s story from Oliver Wilde about the loss of innocence or is it a statement about the use of drugs?—in this blog, I am going to rely on the account of the songwriter, that the lyrics tell a childhood story about a dragon that lived in Honalee.

A dragon is often a symbol of evil. It can symbolize supernatural powers, wisdom, and/or strength from hidden knowledge. Here, I postulate that the supply chain dragon is data. Supply chain leaders are drowning in data and low in insights. Each survey completed over the last decade comes back with data as a major challenge for supply chain leaders. In Figure 1, we show results from a recent S&OP study.

Figure 1. S&OP Challenges

How do we harness the power of data with new forms of analytics? How do we slay the dragon and kill bad processes? As I think about this topic, I find it useful to listen to this song from 1963.  Read the lyrics of this old favorite. The focus is on the loss of innocence. …the dragon lives forever, but not so for little boys… It is time for us to lose our innocence–traditional analytics thinking– and ride the data dragon to new horizons while building powerful organizational capabilities.

The lyrics for you to ponder:

Puff, the magic dragon lived by the sea
And frolicked in the autumn mist in a land called Honahlee
Little Jackie paper loved that rascal puff
And brought him strings and sealing wax and other fancy stuff oh

Puff, the magic dragon lived by the sea
And frolicked in the autumn mist in a land called Honahlee
Puff, the magic dragon lived by the sea
And frolicked in the autumn mist in a land called Honahlee

Together they would travel on a boat with billowed sail
Jackie kept a lookout perched on puff’s gigantic tail
Noble kings and princes would bow whene’er they came
Pirate ships would lower their flag when puff roared out his name oh

Puff, the magic dragon lived by the sea
And frolicked in the autumn mist in a land called Honahlee
Puff, the magic dragon lived by the sea
And frolicked in the autumn mist in a land called Honahlee

A dragon lives forever but not so little boys
Painted wings and giant rings make way for other toys



Today, technology providers are selling tools/software. The opportunity is to unleash the dragon to enable a learning organization. This journey is about more than visualization, simulation and pattern recognition. It is about answering the questions that we do not know to ask and testing and learning from market data (in vitro) to understand effective frontiers in the complex non-linear systems known as supply chain.

Why is this important? Organizations are larger. This is due to mergers and acquisitions and globalization. The focus is on growth, but supply chains are not fit for purpose. The leaders’ goal is customer-centric and agile, but the discussions quickly become political with a wrong focus. Functional silos are a barrier, but the ability to obtain and use data effectively is a major stumbling block.

This week, I received this email from a financial institution questioning why business leaders are not harnessing more insights and redesigning processes based on analytics. The frustration by the investor is that the many promises made surrounding ERP and advanced planning did not come to fruition. As you read the financial analyst’s views of the industry summon your inner dragon:

Clearly, the whole space has moved along very rapidly in recent years.  I think one of the huge problems is that US-centric food companies (Kellogg, General Mills, Smucker’s, Conagra etc.) are working on a“if all you have is a hammer, all you see is nails” problem – they have been making boxed, canned and otherwise shelf-stable packaged food for well over a century but now consumers and retailers are looking for fresh foods and they don’t know how to adapt.  Certainly, Campbell’s foray into its c-Fresh business ended in tears.

We have the rise of new channels, particularly eCommerce. And again, companies aren’t sure how to configure themselves to play profitably in there.

Meanwhile, retailers are getting far more sophisticated in their analytical capabilities, which is reducing the importance of the category captain role for the largest CPG brands in each category.

Everything is getting faster – new products are introduced and eliminated more quickly, better analytics are enabling better real-time feedback on what should go where on a shelf and at what price at a much more granular level – may be down to individual stores. Yet, companies are blind to these insights.

I’m sure there are a lot of other big themes. Not sure what big topics you’ve been looking into recently, but if we can find some topics for discussion that would be relevant to investors in the Consumer area, that would be great.  We don’t understand why the investments of ERP and advanced analytics are not yielding better results…

Stories from the Field

To stir up debate, let me share some stories. In 2015, I worked with a manufacturer of men’s underwear. (My stepson calls them tidy whiteys.) Imagine a boring, basic apparel item. The company had the major market share in their category, but they had a problem. The average buyer was 55-year old men. Older men buy less underwear than younger males and their packaging was not as attractive to the female shopper buying for the family.

The Company’s problem was how to change the demographic and sell more products. Their question? Could analytics in the supply chain help? The company sold on Amazon, operated its own website, managed outlet stores, and could purchase end-aisle displays in major retail chains. We formulated a test plan to test and learn how the combination of price, artwork, color, package count and style attracted a different demographic. We established e-commerce tests, studied basket behavior and took the lessons learned to the bricks and mortar displays. We mined unstructured data to understand the brand preference of ethnicity and age. The product development group advocated a long-tail supply chain strategy with many combinations of color, packaging sizes, and artwork. We rationalized product complexity. We moved the needle when a major sports star became the brand advocate. The discovery of the connection of the sports star to the brand positioning came through unstructured text mining of the social and interest graphs of the potential buyer.

A second story is of an electronic distributor. The company’s goal was to improve customer loyalty. The measurement was net-promoter scores and results of an annual survey. The problem? These were annual assessments and measurements that were not actionable. The question was, “How could the organization harness data and drive loyalty?” We worked on the mining of email data using sentiment analysis to understand distributor opportunities by segment. The organization had 12,000 emails/month sitting in in boxes that were not used.

Mired by product complexity–product portfolios not driving growth–supply chain costs are increasing. How many organizations struggle to formulate the right questions and use new forms of data. My answer. The problem is pervasive. We need to harness the dragon and yield new results. The answer is not with one technology or approach.

We need to harness known data to answer known questions and drive insights through visualization. We also need to ask ourselves are there new ways to use data that we do not understand and answer the questions that we do not know to ask. And, is there an opportunity to build a semantic reasoning layer to enable continuous learning? Discovery, test & learn, sensing and opportunity identification are all possible if we are willing to change how we work.

Table 1. Untapping the Potential of Using Analytics to Build the Learning Organization

Visualization Visualization of an Answer Visualization of a Learning
Questions Asked Known Unknown
Data Known Unknown

Let me give you an example. In 2010, Kellogg produced cereal with liners in the box with an odor. The Company published the statement in response to customer sentiment, Kellogg’s company statement says that “We have identified a substance in the package liners that can produce an uncharacteristic waxy-like off taste and smell.” In this situation, as sales declined, the company might question:

  • Is the product positioned properly on the shelf?
  • Was the promotion effective?
  • What was the impact of a competitive product?

The team would never have asked the question, “Is the downturn in sales due to quality defect caused by a supplier providing liners for the boxes that stunk?” Similarly, in the world of transactional data, they would not have been mining social sentiment to see the onslaught of twitter complaints on odoriferous liners.

Once I ate a Pillsbury breakfast bar and crunched on an insect. The manufacturer had an infestation problem. Coming from the world of manufacturing, I know things happen, but my decision to never buy the product again was emotional, not intellectual. I lost trust in the brand. I never bought the product again, and still, shiver at the thought.  When I visited Pillsbury and asked the team about their infestation problem, they hung their heads. A case of not getting data quickly to understand how this infestation became a brand problem…

How to Get Started

The evolving dragon in my story is the “supply chain engine that could.” The promise of mathematical engines to improve business decision-making is the basis of the supply chain planning, trade promotion management, revenue optimization, and supply chain execution markets for transportation/logistics. The concept is to take data, push it through an engine and gain insights from a better output. Historically, the design of the data model drove software market positioning. With each evolution of the market for “engines” the market moved through a hype cycle. In the 1980s, it was optimization overlaid on AS-400 operation systems. Then there was the move to client-server and the deepening of “engine math.”

Over the last two decades, I followed market shifts. These included cloud-based computing, in-memory processing, parallel computation, and open-source analytics. Traditional decision support software applications were “functional” serving

The assumptions were:

  • Clean data
  • Seamless processing in and out of the engine
  • The fit of the data model
  • The ability of the model to drive Insights (Requiring testing of the output to understand if the technology is yielding better decisions.)
  • Timing: Available output at the speed of business. (Early solutions were not scaleable.)
  • Visualization: Usability and ease of use
  • Interoperability: The ability to move data with context to business teams based on requirements.
  • Use of the insights in a process to improve processes

This dragon story has many twists and turns. The rise of pattern recognition, machine learning, semantic reasoning, text mining, and open source analytics capabilities opened-up new possibilities to drive improvement. Smart guys peddled new approaches attempting to land new big deals. In short, cool techniques looking for a home—entrepreneurs looking for a problem to solve.

Everyone uses terms like Machine Learning (ML), Pattern Recognition, Sentiment Analysis, Artificial Intelligence (AI), and Cognitive Computing interchangeably. Stop the frolic. Demand clarity by the technology providers and hold them accountable. (Most overstate capabilities using buzzword bingo.)  Raise your hand and ask for definitions. Here are the ones that I use:

  • Machine Learning: The use of algorithms and statistical models to enable a computer system to perform specific tasks without using explicit instructions,
  • Pattern Recognition: Machine learning based on patterns. The sharing of insights from the patterns. A useful technique for data cleansing.
  • Optimization: The use of math techniques and statistics to improve outcomes.
  • Cloud Computing: Use of in-memory processing using the power of the cloud (storing and accessing data through the internet versus writing to memory on localized disc drives). Third-party hosting is distinctly different.
  • Sentiment Analysis: Mining of insights from unstructured data. Text mining is also a frequently used term. A useful technique for warranty, quality, and consumer data.
  • Cognitive Computing: Use of pattern recognition, machine learning to drive a semantic reasoning layer to answer the questions that we do not know to ask. The ability to discover and learn based on an ontological framework.
  • Open Source Analytics: A set of languages, techniques and tactics to enable parallel processing, schema on read, and data insights. Open-source collaboration drove the evolution.
  • Artificial Intelligence (AI): The development of computer systems able to perform tasks that normally require human intelligence, such as visual perception, speech recognition, decision-making, and language translation. Sentiment analysis, machine learning and cognitive computing are techniques to enable AI.
  • Autonomous Supply Chains:  The ability to use data to sense and respond effectively automatically at the speed of business. Today’s supply chains do not sense and they do not respond at the speed of business.

You might disagree with the definitions, but the first start in your journey of harnessing the power of the dragon is to get clear—and hold technology providers accountable for — a set of consistent definitions.

The second step is to align the techniques to improve known processes. Here is an example as applied to supply chain planning:

Table 2. Sample Use Cases for Advanced Analytics

Goal Technique
Clean Data Pattern recognition and machine learning
Seamless processing Open Source analytics
Drive Insights New Forms of Visualization
Time of Data. Reducing Data Latency. Parallel processing, cloud-based. Open-source analytics
Interoperability Rules-based ontologies to manage multiple ifs to multiple thens and the use of ISO 8000 standards


Stay focused to overcome the industry barriers:

Commercial Models of the Technology Providers. The dirty little secret of the software industry is the commission structure for a sales team. (Most business leaders do not realize that the selection of software lures them into a political snare. Due to the large commission structure, software sales teams politicize the selling process by selling to business leaders egos and polarizing the business teams. The selling of large software deals is very lucrative. As a result, many of the market thrusts are sales-driven—over-hyped promises with many market ups and downs.

The problem? The solution is complex—engines, infrastructure, workflow, rule sets and visualization. The selling cycle attempts to simplify against a business problem. We don’t have well-defined answers on how to redefine processes to improve outcomes. No one has the answers. As a result, over-hyped marketing with over-zealous sales personnel is a barrier to progress.

Emerging Roles of Data Scientists. Form the right team. Realize that data scientists are important to the the team, but success requires a balanced approach. Don’t form a team just of data scientists. The reason? Data scientists speak a different language and lack the process/domain understanding. The tendency of a data scientist is a localized optima focus. Focus on a holistic approach. Rules, engines, policies and metrics need to align. Attempting to build holistic solutions from localized optima is fools play.

Traditional Processes and Technology Standardization. Traditional providers from the ERP and APS markets have been slow to adopt new techniques. Companies sticking to technology standardization will be lag behind.

Steps to Take.

If you are wondering how to ride the dragon, consider taking the steps:

  1. Provide innovation funding. Don’t hamstring the team with having to justify the ROI of each project.
  2. Form a cross-functional team. The ideal group has maverick thinkers—traditional supply chain leaders, data scientists, human resource/change management agents, and IT/analytical members.
  3. Map the issues from the prior year. Identify the gaps and opportunities.
  4. List all forms of data and locations. Spend time to identify unstructured data sources and data in the network surrounding the enterprise.
  5. Explore the art of the possible by having lunch and learn sessions with technologists.
  6. Facilitate a strategy session to brainstorm solutions to explore the Art of the Possible.
  7. Celebrate both success and failure.
  8. Market insights from the group.
  9. Build an education program for supply chain operations.


Ride the dragon. Unleash the potential. I look forward to hearing your thoughts.