Building Better Supply Chain Data
“Data is the new oil,” The Economist proclaimed in a widely quoted 2017 article. In an information economy and a digital age, the companies that have the best data and the most sophisticated tools for parsing it will win.
Since that story appeared, there has been lots of discussion about information-driven companies like Google, Facebook and Amazon. But the metaphor works just as powerfully for supply chain. Getting the right goods to all the right places at exactly the right time—or not—amounts to billions of dollars in cost savings or waste. And, the difference is in the data.
So, against that urgent backdrop, forgive us for asking an impertinent question: Do you really know what data is?
The dictionary defines it simply as a piece of information, but that assumes all information is equal. Our experience suggests that there are many important distinctions to be made both in the types and quality of data. Much like the 50 words in the Inuit language language for snow, we need a vocabulary to show how various pieces of information should be valued and correlated.
Think for a minute about just one example of supply chain information:
A shipment is marked as “in transit” by one of your carriers. That may have been true on the day the status was entered if the shipper physically scanned the box as it was loaded onto a vessel. It’s somewhat less reliable if several pallets were loaded but not all of them individually scanned. Or, maybe that status got the facts right but you didn’t get visibility to the data for 48 hours. So, it was already outdated when it first appeared.
Data is even less trustworthy when its creation is triggered by an event in an ERP system. Say, the factory finished production and switched your order’s flag from “in production” to “in transit.” Is the product actually moving, or is it still sitting at the dock? Maybe it’s at destination and stuck in customs.
Unless you know how the data has been gathered and updated, it’s hard to know what “in transit” actually means. If your big data analysis makes the wrong assumption, your finely oiled operations may end up more like a gooey, toxic mess.
----------
So, how do we classify data so that we can work with it usefully, change data types and sources where needed and correlate information across a supply chain? We propose a set of definitions to refine our crude data, much in the same way that data scientists categorize data types for statistical purposes.
Physical Versus Synthetic Data: Physical data refers to any notifications that result from actual interaction with goods. Scanning a box is physical data. Even though the level of reliability is quite different, a worker who enters damage information into a spreadsheet after inspecting a pallet is reporting physical information. Some types of condition monitors and Internet of Things (IoT) sensor devices yield physical data. Think about a tip-and-tell placard, a GPS tracker attached to a pallet or a light/humidity/temperature sensor.
Synthetic data, by contrast, results from the application of rules, predictions and correlations. Inventory auto-replenishment triggers are synthetic in the absence of a warehouse worker to count how many goods actually remain on the shelf. Many ERP events are synthetic: A delivery notification to the distribution center might trigger those goods to be marked “on hand and available” at the warehouse regardless of whether they’ve actually been put away on the shelf yet. And, forecasts depend on synthesizing other information to make a prediction.
Captive Versus Dynamic Data: While the physical / synthetic pairing helps us assess data at the time that information is created, it doesn’t assure ongoing accuracy. For that, we need to explore two more opposing categories.
Think of captive data as information snapshots frozen in time. The cells of an offline spreadsheet may or may not have been accurate when they were populated. But, they maintain the same value once they’re entered unless manually updated. Similarly, a status update of “in transit—departed origin” reflects the moment that event occurred. A shock indicator tells you that goods were jolted violently, but it doesn’t say when or if additional upsets occurred subsequent to the first one.
Dynamic data is more like the frames of a movie, showing its subject over time. A GPS tracker pings out a continuous stream of locations. An IoT condition monitor likewise tells you exactly when—and how many times—goods suffered shocks, temperature excursions, humidity, light exposure, etc.
All of this is not to denigrate any particular data type. But it’s critical that supply chain managers understand the unique characteristics and limitations of each. Unfortunately, many supply chain control and visibility systems treat different data types similarly, which can lead to poor decision-making.
We advocate mapping a supply chain’s data sources against these data types. And, we engineer systems like our own ChronosCloud that connect disparate data types, enable multi-party analytics and break free of traditional enterprise computing rules and restrictions.
----------
Let's take a closer look at data quality. Traditionally, supply chain practitioners have thought in terms of good versus bad data. Right or wrong. In fact, quality isn’t that simple.
Wrong data is easy. It’s just wrong. Someone scanned the wrong barcode, transposed a number, associated the wrong carrier with a shipment, sent the wrong file. While that may not always be easy to identify, it’s simple to see how an inaccurate entry hurts the overall information chain. In one supply chain we assessed, those inaccuracies caused more than half of all shipments to be untrackable.
Likewise, missing data is a straightforward concept. But it’s a little trickier to flag in practice As former U.S. Secretary of Defense Donald Rumsfeld might have said it, there are known unknowns like an empty cell. “But there are also unknown unknowns—the ones we don’t know we don’t know.” That latter category includes unconnected suppliers who have information that’s valuable but not being shared digitally.
Delayed data goes a step further. How much does it affect overall decision-making quality if you discover—days late—that a sensitive instrument received a shock in transit? Compare that with an Internet of Things (IoT) sensor that reports the same condition on a live basis, allowing managers to establish responsibility and to plan for replacements, repairs or other measures.
Delayed data can also be as simple as status updates that arrive late. In a recent data audit with a Fortune 100 client, we found that more than half of all tracking data lagged behind the actual event by more than 2 hours—and sometimes was not reported for 48 hours or more. That lack of visibility significantly limited the client’s ability to see and manage inefficiencies.
Inaccessible data are those nuggets of information that could provide valuable insights if only you were able to see them. Often these details are locked in the systems of partners or suppliers. Or, they’re trapped in systems that aren’t connected to overall visibility or don’t fit in existing data architecture. If your enterprise has multiple ERPs, you know what we mean.
If you’re working to digitize your supply chain, you should carefully consider how proposed solutions account for different data types and data quality. Traditional integrations are a good but limited starting point. Our own ChronosCloud platform adds the capability to see all partner data in one place. And, it helps you dig deeper into your supply chain and make sense of the torrents of information coming from IoT devices. Finally, ChronosCloud’s mobile tools help gather information from unconnected sources and offline interactions.
With all these data types working together, you get high-definition visibility and powerful insights. We’d love to demonstrate the power of ChronosCloud for you—or learn more about your supply chain challenges.
