A couple of decades ago, leveraging the strategy of delayed differentiation in supply chains was the rage. Delayed differentiation offered a relatively higher level of personalization without creating inefficiencies like excess inventory in the supply chain. For example, a manufacturer would color a lot of polo shirts only after demand trends started trickling in on which colors are in high demand. While this may seem banal by the current standard of personalization expectations, it was considered an innovative idea back then. Hence, case studies like these landed in supply chain texts in schools and colleges. Today, we live in the era of the hyper-personalized scenario of the “segment of one.” This is the scenario where a customer can order a customized product, and that product is manufactured to address the demand of that segment-that one specific customer. As you can imagine, this means inserting planning and efficiency challenges into the supply chain. This article will discuss why AI capabilities are critical to designing planning processes for the "segment of one."

Challenges of "Segment of One" and The Role of AI

SAP, in its digital supply chain literature, aptly summarizes the challenge of building the segment of one process: “To support the drive to service the segment of one, design processes must deliver all practical configurations and combinations, planning and manufacturing processes geared to support “lot size of one” production and delivery processes that profitable support smaller and more frequent shipments.” As you can imagine, based on the definition above, the challenges that need to be addressed to build this capability are end-to-end, from design to last-mile fulfillment. Fortunately, AI-enabled analytics can help address those end-to-end challenges. This is a high-level overview and examples of areas where this capability can introduce complexity. Each of the example areas can be broken into more granular challenges. There are additional challenge areas as well that have not been highlighted. Product design complexity: What level of personalization is optimal? While this may seem like a product design question, the fact is that it needs to consider all the supply chain complexities. The level of personalization offered needs to balance supply chain complexity and customer experience enhancement from personalization. Product attribute analysis is an example of an AI algorithm that can be leveraged here. While attribute analysis has conventionally been leveraged to answer the question, “What product attributes, like the features, quality, and design of a product, work together to get the customers to make a purchase.” This algorithm can be modified to incorporate a factor of supply chain complexity. Demand planning complexity: The percentage of demand volume that currently comes from these hyper-personalized segments of one product is very low for most companies. This volume will increase in the future, and as competition intensifies, this increase may become significant. Demand planning and forecasting must factor in this specific demand volume. If you are a forecaster, you can imagine the challenge of forecasting for these segments of one. The good news is that AI algorithms can help address this challenge. You can use AI algorithms to build clusters of these segments of one. By identifying and aggregating these into larger clusters, you can reduce forecasting complexity and help ease supply chain and manufacturing planning. But no matter how much you can consolidate, the complexity of your forecasting process will still increase. This is where more advanced forecasting methods like deep learning can help. While I do not favor leveraging these advanced methods for conventional demand forecasting, this is one scenario where deep learning-based algorithms can impact. The key here will be “learning” from the millions of "segment of one" orders over time to cluster and predict them accurately to the most granular level possible. Manufacturing complexity: The manufacturing planning hinges on the product design, i.e., what level of personalization is offered. The golden rule is that you do not want a dedicated personalization line or any form of ideal capacity that waits for these personalized orders. Designing and planning a manufacturing setup like this will not be easy, specifically when the percentage of volume for these personalized products increases in the future. Note that the complexity here is in two different areas. One is designing a manufacturing process (a one-time thing) for the segment of one. The second is designing planning processes around it. Fortunately, AI algorithms can help you in both areas. This article will limit ourselves to planning the optimal “delayed differentiation” in manufacturing planning. A quick note on designing the manufacturing setup before we jump into the planning aspects: While there will always be some additional resources needed to finalize that personalized product, the customer is generally already willing to pay an additional price for the personalization. The first area where AI algorithms can play a role is designing an optimized manufacturing process. You can determine the most optimal configuration with delayed differentiation by combining AI and simulation tools. Let us use an example here to understand the planning implications of the segment of one and the role AI can play- The example of personalized M&Ms. This is an actual product and is an easy example. I am unaware of how this personalization is currently baked into the process. Let us explore how these personalized M&Ms can use the same manufacturing line as the mass-produced ones. The personalization here is that you can print your initials or a picture on M&Ms. You can select your custom colors too! You order one pack of Red M&Ms with the initial “D” online. The order gets consolidated with millions of other orders online and, based on the delivery date of your product, gets consolidated with a particular manufacturing batch. At a very high level, the manufacturing process of M&Ms is as shown in illustration 1. One observation is that while my product is hyper-personalized for me, the actual personalization step happens much later in the manufacturing process.

Figure 1: High level manufacturing process flow

The role AI algorithms can play in the planning process are:

• Help consolidate and schedule an omnichannel segment of one order with an optimal mass-production manufacturing schedule.
• Propagate the impact of this volume on manufacturing, warehousing, and transportation planning (This will eventually be across the three levels, strategic, operational, and tactical). As an example, while optimization algorithms can help plan manufacturing schedules, deep learning algorithms can keep learning from those schedules so that, eventually, you may not need to run optimizers. The AI algorithm would have learned the segment of one volume order patterns, how the flow by day, month, season, etc., and can help address many tactical planning challenges. This will sit in your supply chain control tower setup.

Transportation complexity: Revisiting figure 1, you can see that the transportation planning for these personalized orders needs to happen separately. Unlike the packs for retail and distributors, which are pallets and cases, these personalized orders will be eaches and packs. The packages will be parcel volume, and the nature of the transportation route will be different (handled by a logistics services provider, though). You need to leverage analytics to keep the transportation cost low, and AI algorithms can also help here.

SAP's Solutions

SAP's solutions can allow you to design and build processes for your segment of one journey. Leveraging its suite of products, SAP customers can build a platform ready to address the challenges of fulfilling the segment of one demand. Figure 2 is another example of the "segment of one" demand in consumer goods. This example is of personalized sneakers. As Figure 2 highlights, with SAP S/4HANA as the digital core, SAP BTP, and many other SAP offerings like SAP IBP, SAP EAM, and SAP Business Networks, you can build solutions that address the imperatives of the digital economy.

Figure 2: SAP example of "Segment of one" in the consumer goods industry

 

Source: SAP White Paper: Digitizing The Extended Supply Chain-How to Survive and Thrive in Digital Economy