A whopping 73% of all digital transformation initiatives fail to deliver sustained ROI, according to the Everest Group. With all the cash flowing into digitization, that means billions upon billions of dollars are going to waste. What’s behind this dismal success rate? For one, the process isn’t a bed of roses--the harsh reality is that companies are still facing a lot of trials and tribulations on their digital transformation journeys. Another is that they simply try to bite off too much at once.
Broken down, digital transformation happens in stages. Each stage is an opportunity for operators to get more visibility into their business, and more control of their own activities, giving them the performance improvements they set forth to achieve.
To illustrate these stages and the questions that should be asked at every one, let’s evaluate one company--a factory that is a node in a global supply chain. In this example, we’ll look specifically at a factory operator, where items get unloaded from trucks at the loading docks, processed by several machines, and then shipped out on other trucks to be delivered. It has a mix of people, processing machines, autonomous robots and cranes moving the items from dock to processing to shipment.
As this operator undertakes a digital transformation, here are nine questions they should be putting forth.
Question 1 - Data visibility: What am I tracking?
The first questions any company needs to ask about their operations are: Do I know what’s going on? Can I track what I do in my physical plant? For a lot of operators, the answer is no. Their team may be able to explain what they are doing at any given moment, but they have no way of tracking all the activity in their plant, no central source of truth, no store of historical data. In many cases, they have cursory information about what came in and what went out, but they are not gathering data that is granular enough to show which activities improve their outcomes and which activities hurt them.
The lack of visibility into supply chains and manufacturing can lead to catastrophic breakdowns, shortages and costly inefficiencies, especially when conditions change quickly, as they did in 2020.
Question 2 - Goal definition: What do I want to achieve?
An often-skipped, but critical step is taking the time to sit down and define the goals of your business and physical plant before you begin collecting data to track your performance. You can have multiple goals. Many factory operators will care about maximizing metrics such as product quality, throughput and personnel safety, while curtailing cost. Some may have the additional goal of minimizing carbon emissions. Those goals will point you to the priority data that you want to collect.
Question 3 - Data collection: How are you planning to approach measurement?
Data collection means wiring up your operations to gather data about what people and machines do at the factory. Data should be tied to your goals to reflect critical aspects of their performance. For example, if emissions matter to you, how can you measure the energy consumption of various machines and processes, as well as their relative efficiency? If throughput matters, how can you measure how long a process takes, and how likely a machine is to complete a process without flaws. When people talk about the Internet of Things (IoT), very often they are talking about hooking machines up to sensors to track those metrics at the data collection stage.
Question 4 - Data storage: Is a relational or noSQL database best for my initiative?
Without data, there is no digital. Data is at the center of digital transformation. That’s because data is how we replicate real-world operations digitally. We capture traces of them, and record those traces in bits.
One way to understand data is with the analogy of water. Data is a liquid. It has a source, it flows in streams, it pools in lakes, and it can be pure or impure, which will make it more or less useful when it is consumed. So, when you are working with data, or imagining a digital transformation, you should be asking yourself about its source, channel, tank and purity. You should be imagining a waterworks, in which the decisions that you drive with the data are the gears that transfer energy to the rest of your organization.
Building proper data storage will depend on the types of data you need to store, the volume of data you have to handle, and what you plan to do with it. You’ll make choices between relational databases and NoSQL databases.
Question 5 - Data queries and business analytics: What just happened?
Once you start collecting data, you’ll want to look back on it, or analyze it retrospectively. You want to find out what happened and why, seeking insights by applying various algorithms to mine the data. Your data needs to be queryable, so that you can derive insights from it. If you have collected and stored data, but have no way to extract insights from it, you don’t have a data solution, you simply have an additional problem on top of your physical operations. The insights you might extract from data about a factory might show you which items take the longest to process, which machines are the most likely to break down and what times of day your machines are most occupied--and these descriptive analytics can serve as the basis for new decisions about how to operate the factory. For example, which machines should particular items be routed to, and what times of day are they most likely to be processed quickly?
Question 6 - Predictive analytics: What’s about to happen?
One step beyond the descriptive analytics of business intelligence is predictive analytics. Rather than looking at what did happen, you try to predict what will happen. Predictive maintenance, demand forecasting, and capacity and production planning are all part of extrapolating from the data you have collected to try to guess what will happen next.
Question 7 - Data streaming: Can I make better predictions if I know what’s happening now?
Predictive analytics can be most useful when applied to streaming data. Rather than making predictions based on large historical data sets that you have stored, you might take data streaming straight from your factory sensors in real time, and attempt to understand what will happen in the very near future. This can be key to preventing catastrophic equipment breakdowns, for example, or anticipating bottlenecks on an assembly line.
Question 8 - Simulations and digital twins - If I do this, then what?
Simulations are virtual copies of physical operations. While data is a digital copy of events, simulations are digital copies of the logic of a system. For example, you might want to model a factory. To do that, you need to know the location of each machine, the parameters that can be adjusted on that machine, its position relative to other processing equipment and its staffing requirements. We call that the logic or the physics of your system.
A valid simulation of a factory will mimic the behavior of that factory, and also allow you to perform thought experiments with software that would be too expensive to enact in real life. For example, what would our throughput be if we doubled the processing equipment? What would happen to our factory if it was overwhelmed by orders, or damaged in a fire? Simulations can be combined with real, historical data to see what would have happened if the same events were handled in a different way by your machines and personnel.
Question 9 - Prescriptive analytics and autonomous control: What should I do next?
By exploring your choices in simulations, you may find new strategies that will help you meet your business goals. Maybe there is a way to coordinate the work of your processing equipment, or schedule the jobs you send to it or allocate your staff that would improve your performance metrics. By surfacing those strategies with analytics and simulations--sometimes using AI tools such as deep reinforcement learning--you can identify behaviors that will have an impact on your bottom line. Those decisions can be made in response to real-time data and fed into your ERP, WMS or SCADA system to alter the actual performance of your operating plant.
The world is still shaking from the changes caused by COVID and the economic lockdowns. Disruptions in supply chain and manufacturing have been widespread. Some companies have paused new investments in technology while hoping for more certainty in the near future; others have had to directly address the challenges of the last year by accelerating their digital transformation. When change comes, organizational conventions and culture don’t always suffice for a company to adapt. They also need data to know what’s happening, what changed and what the impact of their responses might be. The current shifts in supply chain brought on by COVID-19 are forcing many organizations to get serious about understanding their physical operations better, and making them more responsive.