Retailers, suppliers and distributors are collecting more data than ever in an attempt to bring greater visibility into supply chains. Yet data scientists say collecting data simply for the sake of it yields little benefit if there aren't standards and a definitive end use for that data.
Organizations in the supply chain that put a stronger focus on obtaining clean and accurate data, and on making that data more actionable, can boost their productivity and efficiency. Experts say it starts with identifying end uses for data then creating a strategy to collect, manage and process it into end results.
Too much data, too little direction
Through new sensors, IoT devices and SaaS solutions, organizations are capturing more data than ever. According to a report by IBM, 90 percent of the data in the world has been collected in the past two years, and businesses and consumers now generate 2.5 quintillion bytes of data per day.
While companies in the supply chain are investing millions into new data technologies, many are “just dumping data into databases” without questioning the validity of that data or how it is being used, said Robert Handfield, Executive Director of the Supply Chain Resource Cooperative at North Carolina State University.
As a result, many organizations are not “closing the loop” on their data and converting it into actionable insights, said Vivek Soneja, Global Head of Supply Chain Line of Business at Anaplan.
“Theoretically, you should be able to discern insights from that data, predict what’s going to happen, and then you need to be able to respond to those predictions,” Soneja said.
Many organizations in the supply chain are operating in siloed environments, fragmented by departments and region. They often have multiple procurement teams across the globe, each of which may be tracking the same supplier with variations of names and limited knowledge that other areas of the business are interacting with that supplier. Some old ERP systems also don’t fully account for what goes on in other nodes of the network, such as distribution centers, stores and shopping locations.
“We really didn’t have the ability to bring all this network data into one place, but we do now. It changes the game on how we plan across the network and the entire supply chain,” Soneja said.
Cleaner data can offer greater accuracy, visibility and efficiency
The impacts of bad data are easy to spot at the retail level. This can be a big problem in an omnichannel environment where consumers and companies expect accurate real-time visibility into inventory.
One example is when stated inventory in a retailer’s system doesn’t align with what’s on the shelf.
“It’s a common problem. Somehow the data has been polluted or inaccurately entered into the system and there’s a mismatch between what has been sold, what should be there, and what is actually there,” Handfield said.
Just how clean data is can depend on how it’s being captured, Handfield said. While barcode readers, sensors and IoT devices have a high level of accuracy, anything that involves humans, manual entry or subjective decision-making could have errors, he says. Incompatible back end systems can also spur the need for redundant data entry, which can lead to duplication and mistakes.
Clean data is simply data with a high level of accuracy “that you can make more sense out of,” Soneja said. In addition to enhanced productivity and efficiency, more reliable data can boost customer satisfaction, produce better on time fulfillment and optimal shelf stocking rates, and trigger a “domino effect” of positive benefits throughout the organization.
“Think about all the data that is used for decision making. If it’s only 60 percent accurate, that’s a lot of errors that might be in your financial reporting, forecasting and demand planning,” Handfield said.
Improving data standards and practices can also make that data more actionable. Evaluating data collection and procession can also help make more sense of it. For example, Soneja said many companies collect data on social media about what customers are saying about products or fulfillment. But many of these companies don’t properly format data with keywords or associated tags to truly turn it into actionable insight. That data still needs to be converted into formats that both machines and people can understand.
“People have lots of data, but no one is really doing enough work to discern the right information from it,” Soneja says.
Create clean data by identifying end uses and implementing standards
As artificial intelligence and machine learning brings new capabilities to the supply chain over the coming years, organizations will need new infrastructure and standards to perform in increasingly complex data environments.
Cleaning up data starts by asking what problem is trying to be solved and what the data is being used for, said Suresh Acharya, Head of JDA Labs at JDA Software. Once that end use is identified, the organization can then determine the right data to collect and the format in which to process it.
Organizations should also create high standards and procedures for entering information into the system, Handfield said. They should create strategies around how they want to govern and manage data, and then create a long-term view of the purposes they want that data for.
“Focus on the data you use on a day-to-day basis for KPIs and decision making. Focus on that and start on those elements first to ensure you get it right. That’s the stuff you use most often for business decisions,” Handfield said.
While there isn’t a silver bullet or stand-alone solution to instantly create clean data, data science starts with a greater focus on the science and not just data, Acharya says. If organizations can become more collectively disciplined in collecting the right data, storing it, cleaning it at the right time and at the right place, it may be able to accomplish amazing efficiencies in the years through machine learning.
“We can do that, and it’s starting to happen in supply chain, but the data journey ends up being the longest one,” he said.