How knowledge graphs help mitigate increasing supply chain complexity

To be future-proof in the new average volatility, supply chains need to address greater complexity quicker than ever. Technology is a critical enabler in solving this challenge. However, companies must select technology solutions that enable them to make faster decisions and break down silos, yet still be flexible enough to adapt to company-specific requirements and address changing business situations. Knowledge graph modeling can do just that and bring a decisive advantage to supply chains.

Volatility, complexity, and speed of decision-making are critical features of the new normal.

Volatility is the new normal and requires speed in decision-making across strategic, planning, and execution horizons. At the same time, supply chains are becoming more complex than ever. The COVID crisis is adding to this, from new multi-sourcing approaches to increase resilience to the accelerated move online. Further complexity is driven by an increasing number of partnerships and mergers.

For example, we’re observing an ongoing wave of partnerships and mergers in the auto industry to reach scale and enable the massive investment required to compete in emerging markets for Electric and Connected Vehicles. However, to achieve the expected scale benefits (e.g., combining purchasing power, manufacturing, and R&D capacity), the auto industry will need to integrate supply chain management across formerly independent supply chains now serving more brands and platforms.

Speed and quality of supply chain planning in increasingly complex value chains are critical to building future-proof supply chains with a competitive edge. Speed, quality, and complexity are competing for needs, and companies tend to prioritize and compromise. However, future-proofing a supply chain requires superiority in all three areas beyond improving current supply chain planning processes and systems.

Technology enables fast decision-making in a complex and volatile environment

To achieve the required decision-making speed, companies are starting to rely on capabilities like machine learning to quickly turn big data into actionable insights and prescriptive actions that enable automated planning and execution processes resulting in supply chain planning man and machine workflows to run at higher clock speed and support faster, more collaborative decision making.

A prerequisite is building an effective digital representation of the increasingly complex end-to-end value chains to provide relevant data and knowledge with the required granularity and latency to these higher-clock speed supply chain planning workflows.

Increasingly complex value chains lead to the increasing complexity of digital representations. This is exacerbated by ever-increasing data volumes and complexity (more external data, disparate data sources, IoT, etc.).

Not all technologies are equal in managing complexity

A critical success factor in building a future-proof supply chain is the ability to represent value chain knowledge and information in a way that can be used for fast and collaborative seamless decision-making across different planning horizons. Suppose technology solutions translate increasing supply chain complexity into a similar or even disproportional increase of data complexity and volumes to be processed. In that case, this puts speed and flexibility at risk.

Traditional data storage and digital representation paradigms (RDMS​, OLAP cubes) often store data at the lowest level, causing artificial data explosion with increasing supply chain complexity. This results in scalability and speed challenges and eventually leads to simplified models to keep a satisfying level of performance but compromises quality, as this model can’t use today’s data abundance to the fullest to obtain actionable insights.

Enter KGs – representing Value Chain complexity without exploding complexity of VC representation.

Knowledge Graphs represent knowledge not just by aggregating data of multiple underlying transactional systems (for example, ERP systems of pre-merger companies) but also by considering relationships across numerous entities (products, parts, brands, product platforms, markets, customers, plants, suppliers, etc.).

This provides the right level of complexity for each horizon and functional stakeholder. This fundamentally breaks down functional data silos because it allows for a ‘function-specific view of one truth’ yet is consistent with the real-time propagation of each change. Moreover, it can do this quickly and is flexible and extensible to adjust to changing business rules and configurations, new data, etc.

KGs break the functional data silos by providing ‘function relevant views’ that are consistent.

KG-based supply chain planning can be particularly impactful where different functional views of a product increase the risk of functional silos, which is the case for most configurable products.

For example, in the automotive industry, a car has different functional descriptions: customers buy sales features (like models and options), so this is sales-speak, whereas supply-speak would be about parts, manufacturing capacities, and supplier capacities to produce these parts. These different views are typically linked with usage statements describing relationships between these terms.

This is useful, as selling features such as park pilot or in-car entertainment do not require the sales representative to know how these are connected to different wiring harnesses, are installed, or the capacities for hundreds of other parts related to this many other features. Similarly, the manufacturing plant does not need to know which channel or market a car is sold to, as its focus is to build on time and meet quality standards. While both sides need their view—without redundant details—these views must be consistent.

Consider the current semiconductor shortage requiring companies to find optimal capacity-compliant production planning. Optimal concerning business objectives which can be dynamic and differentiated. For example, goals like margin, revenue, market share, and serving strategic customers can have different weights for brands in the same group. A decision in sales-speak (i.e., what to build for what customers) may be subject to restrictions in supply-speak (i.e., available capacities).

Many features use semiconductors. For example, a park pilot feature uses a camera, which includes chips that are shared with displays used in entertainment features. Making optimal use of available semiconductor supply requires considering all relevant relationships.

Knowledge Graphs can connect data at different levels (e.g., considering revenues for all models in a particular market using a certain part related to a specific supplier tool capacity, supporting rapidly building scenarios, and identifying incompliance and enabling prescriptive analytics proposing supply and demand actions) and show these to each function in their language. Sales can be restricted to sales-speak while the supply chain works on a capacity level, and the Knowledge Graph’s real-time propagation will keep them synchronized to enable collaboration.

Today, companies can close assembly plants for days or weeks, strip certain models of specific features, or replace them with analog solutions. These actions can be decided quickly without much analytics support. However, these are rather harsh measures that help manage the shortage but are unlikely to achieve optimal business use of a scarce semiconductor supply.

Right view also means an appropriate level of granularity for each function and horizon.

This is enabled by the telescopic capability of KG-based supply chain planning.

Let’s continue with the semiconductor shortage to illustrate this. Driven by industry volatility, automotive requirements collapsed and rebounded earlier than expected. Working from home also went increased requirements for consumer electronics. An immediate solution is not in sight, and demand will further increase driven by moves to electric vehicles, infrastructure electrification, 5G, etc.

This makes capacity planning a challenge. Particularly for high investment and extended lead time foundry capacity based on consolidated demand trends for the many products and industries sharing requirements for this capacity. Requiring consistency with other value chain steps like semiconductor assembly and package planning working on a more granular level (e.g., allocation of supply between automotive customers).

With its ‘telescoping capability,’ Knowledge Graphs-based supply chain planning systems provide the ability for semiconductor players to have one single integrated plan across horizons for their supply planning, enabling companies to make decisions at the right level depending on the horizon but with consistency in the decisions. They will also allow companies to leverage relevant external information on trends in diverse end-customer requirements to create and evaluate different scenarios quickly.

Technology goes beyond KGs, and future-proofing supply chains go beyond Technology

Knowledge Graphs modeling of supply chains allows us to get from data to insights quickly. However, increasing the full organizational clock speed also requires getting insights into decision-making, from decisions to actions. To achieve this, standalone Knowledge Graphs are not enough and should be embedded in the core of organizational capability leveraging technology, people, and process:

• On the technology side, to ensure speed and consistency of decision-making, Knowledge Graphs modeling should be the backbone of a planning platform integrating end-to-end processes from supply chain design (e.g., Capacity planning) over financial and commercial planning to S&OP and S&OE (e.g., Control Tower)
• On the people and process side, a new man and machine workflows will require people to change roles and reskill on all levels of the organization. Organizational formats will have to break functional organization silos to fully benefit from breaking knowledge silos.

There is no one-size-fits-all approach, and each company needs its discovery journey. It should be a close partnership between business (providing an understanding of problems, processes, and people) and technology experts (knowing what technology can enable).