Written by 2:02 pm analytics, Big data supply chains, Demand

AI This. If Only.


After a dramatic drop at the beginning of the pandemic, Supply Chain Management (SCM) and Enterprise Resource Planning (ERP) software public valuations are near twice the ten-year average. A consolidation wave continued with 299 Mergers and Acquisitions in 2020. The majority of the transactions are in Enterprise Resource Planning (91) versus supply chain planning (31). The high activity levels in the transportation and  Retail/Ecommerce sectors mirror consumers’ movement into new channels. As I look at the financials, I question what the value could be if we got serious about innovation through new forms of analytics.

The gap between what companies need and what is available has never been higher. The current market drumbeat—sustained by high salaries for software sales and consultants—is focused on traditional process automation. Despite the proliferation of new technologies, supply chain planning, designed in the 1980s, remains unchanged. For a gal like me, innovation is painfully slow.

Yes, Artificial Intelligence (AI) is new. The acronym AI hangs on technology vendor websites like shiny icicles on the holiday tree –or gaudy neon lights on a house during the holidays– but when you roll-back the covers and ask fundamental questions, the basics are unchanged.  Companies are hanging their hats on marketing buzzwords, not innovation. Most of the hype is focused on very basic approaches to machine learning. Despite the failure of supply chain planning in demand and supply planning during the pandemic, we cannot and should not try to AI the current state.

Imagining What Could Be

The journey starts with challenging mental models. The current state of automation–Customer Relationship Management, Marketing Management, Revenue Management, Supplier Relationship Management, and Transportation Management–automates decisions within a function but does not enable the seamless flow of data and information organizations to improve organizational outcomes. Today, organizations are larger, more complex, and the speed of business is faster. Companies want to make better decisions faster.

An exercise that I love to do in my training is to ask business leaders to draw data flow in their current decision processes. Within companies, there is a river of demand with multiple parties using data within their individual roles. Each group—marketing, sales, finance, and supply chain — takes a time-phased snapshot of the river using different models, granularities, and biases. I liken it to taking multiple Polaroid pictures. It is quick but cannot stand the test of time.  Within an organization, as shown in Figure 1, there are patterns:

Figure 1. A Participant’s View of the River of Demand In a Large Consumer Organization

While Figure 1 is the planner’s view, Figure 2 shows the executive’s mental model. The gaps are larger and more complex.

Figure 2. The View Of The Supply Chain Executive

Up and down the organization, the focus is on model building and the tight integration of dissimilar models. The issue? Companies do not know how to design an effective flow of data between organizations to use market data.

The industry is stuck in traditional paradigms resulting in many gaps:

  • Lost Opportunity. The larger corporation does not use data collected in consumer and digital marketing programs. 
  • Creating Meaningful Conversations. Today’s approaches do not effectively enable the right conversations between functions (marketing, sales, finance, and supply chain) to drive synergy between groups with differing points of view.
  • Lack Of Grounding. There is no demand visibility signal grounding all groups in baseline demand or market potential.
  • Planning Is Labor Intensive and Slow. Data does not move at the speed of business.
  • Insular. Decisions are inside-out, not outside-in. (No company that I study can effectively use market data.)
  • IBP Is A Deadend Street. S&OP becomes a Ferris Wheel of activity, creating a Deadpool of information. The reason? A focus on financial incentives versus the translation of demand.
  • Procurement Operates On An Island. Material Requirement Planning (MRP) is not sufficient. Business process outsourcing and procurement automation evolved to make the procurement organization more efficient and less effective.
  • Broad Strokes. All items are treated equally when they are not. Cannibalization is not measured.
  • Usefulness. Companies are unable to determine an effective supply model using segmentation, constraints, and capabilities.
  • Usability. The linkage between planning horizons is difficult. 
  • Opportunities Abound. In the organization, lots of data is not used. For example, Vendor Managed Inventory (VMI) systems, a useful stream of customer data, do not connect to planning.

New forms of analytics offer opportunities, if we can step out of our box. This is the goal of this blog post.

AI This. Build New Capabilities

To start the journey, to step out of the box, we must start by forming the right question. Success happens when the right question is aligned to the right analytical technique with a clear definition of what makes a better decision. Each of these steps is hard, but we cannot start the journey without learning a new language to discuss analytical capabilities.

Thinking out of the box starts by seeing the box.

A barrier to “out-of-the-box” thinking is getting clear on definitions and gaining clarity on how to align analytical capabilities to solve real-world problems. Before we begin, I want to be sure that we are clear that this piece is written for business leaders. I am not, nor do I claim to be, a data scientist. These definitions are cursory discussions of very complex topics within AI.

Artificial Intelligence (AI), founded as an academic discipline in 1955, is evolving at a rapid pace. In 2015, there was a step-change in the use of machine learning by innovators in supply chain planning.

Machine learning and semantic reasoning are subsets within AI. While Machine Learning uses data to train and find accurate results, semantic reasoning infers logical consequences from a set of facts. The inference logic is usually driven by an ontological language. An ontology is a set of truths to drive reasoning.

Pattern recognition, Natural Language Processing (NLP), and deep learning are subsets of Machine Learning. Pattern recognition is the detection of patterns and regularities in data while natural language processing translates unstructured data into a structured form to enable learning. In contrast, deep learning uses many levels of algorithms to drive insights through neural networks. Deep learning encompasses unstructured data whereas pattern recognition is limited to structured data. Today, when a company in the decision support technology market speaks of AI, it is usually pattern recognition. (We are just dipping our toes in deep water.)

A barrier is the structure of today’s systems with a heavy reliance on relational database structures. Relational database structures hardcode data into formal, and inflexible tables. Schemas store tables, and inside each table, there are pre-defined columns and rows. In contrast, as shown in Figure 3, a graph-based database is a mathematical representation of objects, entities, or nodes and their relationships.

Figure 3. Contrast Between Relational Tables and Graph Databases

Attribution: Cambidge Semantics

Often there is confusion between an ontology and a knowledge graph. An ontology is metadata/schema, whereas the knowledge graph is the data itself. Think about generating a domain ontology and populating it with dynamic facts using a knowledge graph, to create side-by-side collaborative work: machine learning feeding semantic reasoning.

So what does this mean? Most of the time that technology providers use the term AI, the application is simple pattern recognition. We are early in the use of graph databases and even earlier in the application of cognitive or semantic reasoning. From my interviews of asking providers– What Do you mean by AI?–I built the image in Figure 4.

Figure 4. Application of Analytic Techniques

As a simple gal, I am less interested in the verbiage than the application. As a business leader trying to figure out how to drive value, I provide insights in Table 1:

Analytics TechniquePotential Value Proposition (s)
Pattern RecognitionMapping dirty data like master data. Visualization of patterns in data.
Natural Language ProcessingMining unstructured data. Visibility of customer sentiment from email and comments. Listening posts for sentiment, warranty or quality data.
Deep LearningDemand insights generation from ever-changing markets.
Semantic ReasoningRule automation: connection of customer-centric segmentation to Available-to-Promise and Allocation Strategies.

Use Of Analytics to Transform Decisions

While the term AI is bandied-about, innovation is slow. Take the first steps by exploring the use of graph-based databases and building ontologies to automate the river of demand. Start small pilots using semantic reasoning while aggressively exploring machine learning on graph-based database infrastructure. Curtail investment in legacy relational database infrastructure.

In the journey, imagine how to connect graph-based databases across sales, marketing, finance, and supply chain groups to enable demand visibility across multiple models to drive new insights for all participants. And, for the innovator, explore how to use semantic reasoning to tie baseline market sensing to program definition of trade and price policies and connect customer segmentation to rule enablement for order management through ATP and allocation strategies. Having fun yet?

See You At the Supply Chain Insights Global Summit?

We are taking the risk that everyone can get COVID shots/ tests to enable an in-person event in September. We will also have a virtual feed for those unable to travel. The goal of the conference is to Imagine the supply chain of the future. The conference is in Franklin, TN on September 7th-10th, 2021.

In preparation, I am handpicking the speakers and finishing up the Supply Chains to Admire analysis for 2021. The agenda will publish in April.

If you have a story you would like to share; please drop me an email at lora.cecere@supplychaininsights.com.
Please mark your calendars to join us to think differently and Imagine the Supply Chain of the Future. We hope to see you there.