5 best practices for creating a data-driven supply chain

What is a data-driven supply chain?

A data-driven approach to supply chain management is the strategic use of data to better predict production and inventory changes closer to real time. This in turn empowers faster decision-making. A data-driven approach typically uses newer data sources, such as from technologies like AI, to make predictions.

"Data-driven supply chains can bring huge advantages to companies by giving them a complete picture of their supply chain performance," said Jason Bergstrom, smart factory leader at Deloitte.

Here are five steps that can help create a data-driven supply chain.

Data-driven supply chains can bring huge advantages to companies by giving them a complete picture of their supply chain performance.

1. Improve internal data use

Using data to improve a supply chain begins with taking full advantage of the supply chain data a company already possesses.

Leaders typically start by looking outside the organization for data when they begin to improve their supply chains, said Balika Sonthalia, a partner in the strategic operations practice of Kearney, a management consulting firm located in Chicago. Most organizations are unaware of their internally available data and don't use it well.

Supply chain leaders should ask other departments for their supply chain-related data, Sonthalia said.

Sonthalia recently worked with a tech retailer whose supply chain team didn't incorporate data from promotions and consumer reaction to those promos when they created their replenishment algorithm.

If the supply chain team had asked the product marketing teams for that data, the company could have improved their operations, she said.

2. Incorporate external data

After improving internal data use, supply chain leaders should then bring in outside information like consumer purchase data, which can also potentially help improve their supply chain.

Consumer behavior has grown far more unpredictable in the wake of the pandemic, making demand planning harder, Sonthalia said.

She worked with a high-value products brick-and-mortar retailer that experienced a spike in direct-to-consumer (D2C) business because of COVID-19. However, different COVID-19 variants soon led to fluctuations in the D2C business, and the retailer found itself with under-utilized contracts and excess inventory.

Data can help improve the supply chain in situations like these if leaders look at the latest data -- in this case, consumer buying behavior during a new COVID-19 variant -- as soon as it becomes available so they can alter supply chain operations in response to it, Sonthalia said.

This more agile approach Sonthalia refers to is demand sensing, a forecasting method that uses newer technology such as AI to analyze real-time data. It can improve customer demand forecasts and help guide planning around partnerships and inventory.

Customer data isn't the only information that can help boost supply chain agility. Supply chain partners' data is also important.

Supply chain leaders can improve their operations even further when they incorporate distributors' and suppliers' data, said Bret Greenstein, data analytics and insights partner at PwC, an accounting firm headquartered in London. Distributors' data can give supply chain leaders better insight into consumer demand, while suppliers' data can give insight into lead times, among other information.

3. Automate data capture, but focus on meaning

Automation can help to put data at supply chain leaders' fingertips. To create insight, they may need to look to new methods and tools.

Supply chain leaders and other stakeholders should identify all the data that can help them make important supply chain decisions, Greenstein said. They can then use software to automate data acquisition, management and analytics processes.

A number of tools can feed data into a cloud data warehouse but extracting meaning from that data is far more difficult, Greenstein said. Deciding which data matters requires understanding data sources' causality versus their correlation. Causality indicates that a change in one thing created a change in another, like a supply chain bottleneck causing further delays. Correlation occurs when there is a statistical relationship between data, but one may not have caused the other.

4. Improve understanding of unstructured data

Gaining insight from unstructured data is difficult but can often deliver valuable supply chain insights.

Structured data is information that has been assembled into a database so a user can gain insight from it. By contrast, users can't put unstructured data in a normal database because it doesn't conform to the usual data models. Data from customer emails is one type of unstructured data that could provide powerful insights.

Supply chain leaders should create a data foundation that can handle this unstructured data, Greenstein said.

"Companies that integrated their [unstructured data] into a cloud data foundation ahead of the pandemic were the ones who were able to adjust to the wild fluctuations in supply and demand [the] best," Greenstein said.

For example, a large convenience store Greenstein worked with had moved their store-level data into an enterprise data fabric on the cloud just ahead of the pandemic, he said. Because of this, they were able to identify the local variations in buying patterns and store performance, then make better decisions about how to distribute supply.

5. Implement AI/ML models

AI and machine learning (ML) models can also help improve the supply chain by finding data patterns, then forecasting potential outcomes.

Companies should develop AI and ML models that are powered by supply chain data, log data and third-party sources, Greenstein said. AI/ML models create forecasts with different confidence levels, which tell supply chain leaders how likely the forecast is.

Companies can customize their models for different conditions because one AI/ML model may not be accurate for all situations, he said.

For example, a company that sells sunblock might set up one model for sales during a warm, sunny summer and another for a cold and rainy summer.

AI and ML models can quickly change forecasts based on new data, Greenstein said.

This ability is key for supply chain optimization.

Forecasts need to adapt to changing conditions, Greenstein said. Referencing only an unchanging quarterly forecast model isn't enough for supply chain success. Leaders must be able to respond quickly to new supply chain data instead of only basing decisions on historical patterns.