AWS for Industries
The Autonomous Journey: How CPG Firms Can Transform Supply Chain Demand Planning
Demand disruption has been a huge challenge for consumer packaged goods (CPG) supply chain management in 2020, and it will continue to be a big issue going forward.
As we discussed in an earlier blog post, Data-driven Demand Planning: Managing Disruption in CPG with ML and Demand Sensing, one part of solving supply-chain demand disruption is to improve planning. Sophisticated forecasts and demand sensing, powered by machine learning (ML), are key for making planning more accurate and efficient.
Order allocation, which is critical for optimizing operations and maximizing revenue, is even more important for responding to demand volatility. This is the dynamic process of making decisions about how, when, and where orders should be fulfilled, including order prioritization, sourcing location, and shipment method. Effective order allocation makes it possible for CPG companies to deliver orders on time, in full, in the most cost-effective way.
Legacy systems that fulfill orders based on static sourcing, capacity, and lead time rules lack the agility required to service 24×7 omnichannel ecommerce orders with tight delivery turnarounds. Using best practices and new artificial intelligence (AI) and ML capabilities, CPG companies can make order-allocation decisions in real time, at scale, across an evolving omnichannel.
Order allocation is more important than ever because supply and capacity shortages have created major fulfillment challenges for CPG companies. Traditionally, companies included a limited set of considerations in order allocation, such as economic order quantities and transportation planning constraints. During the COVID-19 disruption, more factors, including labor costs, warehousing capacity, packaging requirements, had to be considered.
In addition to supply shortages, many CPG companies have grown their direct-to-consumer (DTC) channel, which presents new challenges around optimizing service and cost. According to Digital Commerce 360, consumers spent more than $860 billion online with US merchants in 2020, up 44 percent over 2019.
Global DTC growth will continue, with a diverse physical footprint to service customers by country or market. Fulfillment networks will continue to grow more complex, with cross-border lead times and internal and third-party warehouse and logistics capacities and systems. DTC customers’ expectations tend to be time-sensitive. If they can’t get reliable delivery in a short turnaround—same day or next day—they’ll go elsewhere.
Building Advanced Order-Allocation Capabilities
There are three key enablers to delivering effective order-allocation capability:
- Data analytics
- Advanced analytics and simulation
Inventory visibility allows you to know how much stock you have, its status, and where it is (including in transit) so you can accurately determine and predict availability in real time. Beyond inventory, visibility for other aspects of your supply chain—such as warehouse and workforce capacity, weather impact on transportation lanes, and inbound shipments—empowers you to make sophisticated, intelligent decisions to project availability, bottlenecks, and inbound constraints. Being able to see the big picture allows you to evaluate the options and tradeoffs and make measured decisions rather than constantly managing crises.
Visibility depends on having data that goes beyond traditional transaction information. This requires the use of cameras or barcode scanning to track inventory levels and prevent manual counting. Applications such as GPS, geofencing, IoT sensors, and real-time supplier data integration can capture transportation-related data, warehouse throughput, production schedules, supplier shipments, and more.
Unifying siloed data into a single view is critical to streamlining the process and revealing new insights. You can do this through any ERP system, but it’s more effective through a data lake that can flexibly accommodate both internal and external data sources.
After you have the right data, you have to analyze it to make better predictive decisions. This step combines different models for optimization, cost, prioritization, and more.
Optimization is the process of continuously evaluating inputs such as cost, margin, and service to identify the best possible outcome and achieve the desired goal. An optimization engine for order allocation:
- Checks across all sources of supply.
- Identifies the total fulfillment cost for each order.
- Continuously calculates the most cost-effective scenario to deliver against the service promise when any trigger event occurs.
Research suggests that 40 percent of major ecommerce platforms fail to provide any delivery predictions. This can have a direct impact on sales and profitability. Leading platforms like Amazon.com, on the other hand, focus on predicting delivery dates accurately and optimizing their supply chains around that data.
The first level of optimization competency is being able to offer the customer an estimated time of arrival (ETA) and delivering to that ETA in the most cost-effective way using the data you have. As the organization matures, you can add more complex data to the equation for more sophisticated optimization.
Cost to serve takes into account the many variables of cost, including holding, sourcing, transportation, and handling, and incorporates penalties for suboptimal loads, routes, processing, and more. Cost to serve forms the basis for comparison between multiple options from which the optimization engine makes a final choice. That choice should be the least expensive option that fulfills all the goals of the shipment.
For example, although it might generally be cheaper to ship an item from a warehouse in a particular country, that warehouse might not be able to deliver the shipment on time due to lack of resources or capacity. In this case, the company should go with another option to fulfill the delivery date promise—even if it costs more.
Leaders in this space know how much it costs to hold product at a certain location and to pick, pack, and ship it. They assign those costs and feed them to advanced optimization engines.
Prioritization means making allocation decisions based on logic and rules for each type of customer. This requires segmenting customers according to their importance to the business and knowing the cost impact, service impact, penalty impact, and more for each segment. Companies that operate on a first in, first out model (FIFO) and don’t prioritize might find themselves in the unenviable position of running out of supply for their most important customers.
Going beyond segmentation, companies might consider extending prioritization to their larger logistics networks. This could mean using full truckloads and same-day deliveries, factoring in variable fulfillment center productivity, or prioritizing the reduction of carbon emissions.
Today, many organizations work with static rules, but going forward, they’ll need more dynamic prioritization to deal with workforce and supplier limitations and complex situations, as we’ve seen very clearly during the pandemic.
Although data analytics inform decisions for fulfillment options, it’s also used to influence decision-making for planning and executing downstream areas—from routing to transportation planning to labor requirements.
Advanced Analytics and Simulation
AI and ML can take order allocation decision-making to the next level. Rather than using static rules, you can build a feedback loop using machine learning to evolve the process dynamically. For example, if you typically source products from a particular inventory location, ML can show you if it actually costs more or takes longer than you thought. And your rules can change with changing data.
AI and ML can analyze data and create a predictive model from forecasts, historical performance, and more. For example, based on the historical performance of a warehouse, lane, or carrier, you can predict whether they can handle a certain percentage of overcapacity or will fail under certain conditions.
Simulation and modeling are crucial to using the answers gleaned from data analysis to immediately decide which action to take in response to changes. Having the ability to simulate answers allows you to allocate more intelligently and change allocation as your supply and demand change.
Bringing It All Together
Many organizations make a one-time, static decision for how best to fulfill an order and are unable to revisit the decision as their supply chain changes. Leading organizations, on the other hand, automatically respond to changes in factors such as supply, demand, weather, network, stock, and much more, to create the best outcome.
By incorporating a range of data, building a cost model, and using advanced AI/ML and optimization techniques, these organizations can make order-allocation decisions in seconds to provide cost-optimized, accurate delivery predictions and satisfy customers.
For profitability, it’s vital to invest in these technologies. CPG companies that use them to make effective order-allocation decisions quickly and efficiently differentiate themselves from their competitors, even as they continue to face the challenges of ongoing demand disruption.
If you’d like to learn more about how AWS for CPG can help your organization, contact your account team today to get started.