![port-of-baltimore-iStock-1299521189.jpg Cranes and shipping containers along the water at the Port of Baltimore.](https://www.supplychainbrain.com/ext/resources/2024/04/08/iStock-1299521189.jpg?height=100&t=1723060077&width=150)
What’s the single most important development shaping the future of how companies drive business value from their data and analytics capabilities?
According to MIT Sloan Management Review, it’s the convergence of big data with artificial intelligence. Yet many companies, when presented with the resulting business benefits, voice the same hesitation: “We don’t have the data for that. It’s scattered and disorganized. Our data isn’t clean.”
It’s understandable that people would feel this apprehension. Scattered, siloed, and voluminous data remains a common challenge for companies across all industries today. As a result, business leaders might think they aren’t ready to deploy cutting-edge technology like AI. In reality, though, the opposite is true: AI helps clean, integrate, and rationalize data to drive tremendous business value.
The biggest opportunities to leverage AI for big data projects can be seen in four key use cases in supply chain management and business operations.
Use Case 1: Transforming forecasting by incorporating demand drivers and leading indicators.
Typically, companies base statistical forecasting processes on historical sales and shipment data. In today’s increasingly volatile marketplace, however, past events aren’t always the best predictors of future events. Big data and AI-based models create the potential for a future-ready environment, in which companies can move from forecasting driven primarily by historical shipment-based data to that which incorporates various drivers of demand. Such drivers encompass external events, including competitive pricing, market conditions, and competitive assortments, as well as internal drivers related to promotions and pricing.
Companies attempting to incorporate demand drivers in statistical forecasting today without the benefit of machine learning and AI must spend significant effort in normalizing data based on outliers. For example, a dip in sales might have occurred due to a stockout created by a supply-chain constraint. But how would the forecasting algorithm know that this sales dip occurred due to a supply-chain issue and not a demand issue in the market? In a traditional approach, a human effort would have to be deployed to input the “why,” and effectively correct the history before the data is given to statistical models.
All that changes when a company deploys an AI-based platform for demand forecasting. Machine learning (ML) algorithms build models based on patterns in data, without relying on explicit instructions. This means that data inputs that drive demand forecasting can be cleansed, correlated, and given proper attribution to outcomes using ML. In turn, prescriptive demand-based decisions are generated based on the patterns seen over time.
Use Case 2: Driving planning with learning systems instead of tribal knowledge.
Today, the modeling of knowledge remains largely tribal in many organizations. Much of the decision-making for planning today resides in the heads and judgment of individual planners. For example, if a planner received a sales forecast that indicates the supply-chain budget should be directed toward expediting, what will planners do? Do they trust that the demand is reliable? Are they willing to spend hard dollars to meet that demand, or will they hesitate? The decision to expedite or not is often the best guess of the planner based on a personal history with the sales executive or customer.
When AI is applied in this scenario, tribal knowledge becomes institutional knowledge. Historical data on forecast versus sales enables the AI-based system to learn what reliable demand looks like, and who is likely to be accurate (or not) in their forecasting. A decision to expedite or incur additional costs to meet demand will now be based on an intelligent recommendation: Yes, automate this request because it’s extremely reliable. Or, proceed with caution because this customer has been unreliable in the past, and management approval of this decision is required.
In this environment, decision tradeoffs can be made with greater speed, accuracy and cost-effectiveness. Human bias is removed, and continuity in decision-making is ensured, regardless of which planner is manning the system.
Use Case 3: Creating integrated planning and decision-making models by connecting disconnected data.
Virtually every company has disconnected data. It’s a pervasive challenge. A study conducted by Vanson Bourne estimated that U.S. and U.K. organizations are losing an aggregate $140 billion each year due to disconnected data. Data silos exist for a variety of reasons that span the technical, structural, and cultural dynamics of a company.
One of the classic problems related to disconnected enterprise data is rooted in the fact that many companies have grown through mergers and acquisitions. The merged companies may become one entity to the world, but behind the scenes, divisions can remain readily apparent, often for years. There are likely multiple enterprise resource planning (ERP) and other siloed systems for sales, supply chain and product management. Under one corporate roof, a singular product could be known by multiple names across several disparate systems.
This creates significant challenges in constructing a consolidated picture to fuel the decision-making that’s required for planning purposes. Traditional corrective approaches include implementing a single ERP system, or correcting the data in all source systems. Those projects are expensive and time-consuming, leaving many companies to conclude, “We know it’s broken, but we can’t fix it right now.”
An enterprise immersed in disconnected data inevitably has concerns about tackling integrated planning and decision-making initiatives. But with AI and natural language processing, systems can determine that varied data points are, in fact, the same thing. A model can be constructed that correlates all those products so that the sources don’t need to be changed. Inventory visibility, planning and decision-making are now linked, because the system recognizes those products are identical.
Use Case 4: solving the challenges of master data in planning systems.
The power of AI in creating planning systems lies in driving automated and intelligent decision-making more quickly. But another common refrain from corporate leaders is that much of the data required to make those planning decisions is master data that doesn’t reside in any system of record.
For example, a large retailer handles hundreds of thousands of SKUs flowing through the distribution center and store network. That retailer needs to model how much capacity is required across various dimensions, including storage and labor capacity to handle goods in transit, in the DCs, and in the stores. To determine capacity requirements, the retailer must understand what each SKU is consuming on the various resources available. The time it takes a person take to unload a shipment of televisions, which is labor-intensive, would vary significantly from the time required unload a shipment of ibuprofen, which is relatively light.
Data required to drive good, accurate planning decisions should be based on the volume of specific products that are flowing through the DCs, and their corresponding capacity requirements. But who maintains all data? In the past, those details were difficult to model, because they need to be done at aggregate levels, and often no one was capturing and maintaining that data.
Now, with big data and AI, retailers can use transactional sensor internet of things (IoT) data as it’s logging to determine capacity requirements. As workers take a shipment of products from the truck, move it into a DC, and so on, a tremendous amount of transactional data is being recorded. Using AI, retailers can automatically generate the master data that’s required for decision-making. Knowing they have a shipment of televisions arriving, they’re now armed with specific, auto-generated knowledge of how much labor is required to move the product. Here, and in all the use cases outlined, AI enables companies to transform their data into one of their most valuable assets.
Chakri Gottemukkala is CEO of o9 Solutions.