Many managers struggle to see a clear path towards creating digitally -enabled organizations. It is not clear what digitalization investments should focus on, how they can be justified financially, and how the organization can be steered towards embracing these technologies. The plethora of vendors, system integration and digital transformations consultants adds further confusion to these important questions, leaving many organizational decision makers uncertain about how the digital transformation should be staged.
The shiny visions of highly automated operations where artificial intelligence drives countless decisions and activities, albeit fascinating, generate realistically unattainable expectations that can ultimately paralyze the change process. Rome was not built in a day, and the companies that created the most advanced digital environments took their time to do that. Consider, for instance, the case of Siemens’ Nuremberg factory, which assembles industrial controls equipment. This site is a flagship factory where Siemens showcases a mind-blowing level of automation – human intervention is mostly left to exception reporting and control functions. Monitoring and connecting real-time data about products, processes, environmental conditions, and quality controls results in a whopping 99.9988 percent non-defective end products. When asked how they achieved this “miracle,” plant managers offered a shocking answer: “There was no revolution here. We constantly invested in improving and automating the factory for three decades.” As a result, the plant has multiplied by ten its output without increasing its workforce.
Many companies are far from where digital champions are. How can they get there, step by step? After interviewing hundreds of executives with digitalization responsibilities, my research points to four main stages that must be considered to drive this transformation. I refer to these as groundbreaking, experimenting, piloting, and scaling.
We all too often assume that companies have stores of high-quality data that span years of operations, ripe for being exploited. More often than one would expect, this data is not available, paper-based, incomplete, scattered across different systems and organizational units, or has serious quality problems. Most data lakes are dry lakes, or wetlands at best. For instance, until recently, Audi had only paper-based information about quality checks for some of its automotive assembly lines. A Ford assembly plant had a digital assembly line quality control system but, to save on data storage costs, it deleted process quality data six months after collection.
With inadequate data it is difficult to build effective digital solutions, so the first stage often entails creating a data collection infrastructure that feeds a to-be data lake. How can such an investment pass internal filters for investment decisions? The fundamental idea is that data collection systems are indeed monitoring and control systems – and monitoring brings disciplined and efficient execution. The payback of a state-of-the-art data collection infrastructure is ensured by enhanced control and visibility. But the real outcome is the creation of a potentially invaluable data lake.
Having a data lake is a good starting point but what to do with it is not clear to many managers. No wonder a frequent complaint is that “I do not know how to argue for the effect of digital initiatives on Return On Investment.” Besides this, investments aimed at exploiting data lakes raise expectations that may hurt the reputation of the proponent if the results turn out to be disappointing. For instance, Gartner deemed 90 percent of data lakes created through 2018 to be useless, as the organization is unable to generate significant value from the data stored. When uncertainty about the business potential of a data lake is high, then experiments – not pilots – are the way to go. The fundamental idea of experiments is to investigate cheaply a working hypothesis. Unlike projects, discovering that something we hoped was possible is not possible does not qualify an experiment as a failure. It is valuable learning. Learning and reducing uncertainty is the goal of experiments.
Experiments in the journey towards digitalization should use limited resources for a brief amount of time to investigate essentially two things. First, can any useful decision support tool (e.g. an algorithm, recommendation agents, etc.) be built based on statistical inferences from the data lake? For instance, an application service unit of IBM discovered that a regression-based tool could forecast the time it took to fulfill a customer request with two-digit accuracy improvement compared with human planners. Second, since people may be skeptical and even oppose the creation of these systems, the second experiment was aimed at creating a simulated user interface. This simulation enabled users to grasp the potential benefits of the digital solution, and IT people to understand and assess the kind of changes that were needed to the ticketing system for solution implementation.
Once management knows what value-adding analytics to build from a data lake, and is less uncertain about the associated costs and benefits, then it is time to launch a real pilot, or even to revamp an existing process. At this stage the interest lies in improving hard business metrics, be they related to efficiency, customer service, or financial performance. Through pilots, managers can also explore the appropriate way to organize two distinct groups of stakeholders in digitalization initiatives: algorithm developers (i.e. those who create the algorithms based on data lakes) and algorithm deployers (i.e. those who embed these algorithms in the systems employees use in their everyday work). Pilots provide answers to important questions. Which specialists should be in the two groups? How should their efforts be evaluated? How open should these groups be to non-organizational members? How centralized versus decentralized should they be in different organizational areas, units, and subsidiaries?
An example of a successful digitalization pilot is offered by the Princes Group, a fast-moving consumer goods manufacturer based in the UK. Facing strong demand and capacity-constrained facilities, the CFO was interested in increasing equipment output. This directly translated into ensuring that loss of productive time, for example due to preventive and reactive maintenance, was minimized. The algorithm developers (which mostly included external consultants and production engineers) created algorithms that set time goals for worker activities based on ongoing process parameters. For instance, in case a specific piece of equipment broke, the algorithm could calculate how much time the serviceman had to fix the glitch before the process incurred a production loss. Essential to the success of the pilot was the team of algorithm deployers, which also included workers, supervisors, IT specialists, as well as external industrial IT specialists. This team defined how the algorithm interfaced with the data collection platforms connecting the target production line, the line overhead displays, as well as a specific “app” that was plugged into the line so that maintenance and production engineering personnel could analyze data on line performance. The result was an almost 20 percent increase in the line output, which directly translated into a two-digit revenue increase.
An ever-growing concern of managers, especially those working in sophisticated multinational firms, is the need for scaling “local” initiatives (oftentimes, local “pilots”) so that they can be easily transferred to other processes or facilities within the corporation. Without this final step, independent teams are likely to “rediscover the wheel” and produce limited, local, returns to their efforts in digital innovation, without making a real competitive difference. However, ensuring scalability of solutions is not a trivial problem, because the solution needs to be flexible enough to be adaptable to different processes and facilities.
The key concept to ensure that local pilots turn into global solutions for a large company is the creation of corporate digital innovation platforms. The idea of a platform is to facilitate the connection of solutions developers and solution users, enabling many different users to adopt the solution created by one developer. To this end, companies such as General Electric have created “digital twins,” which are digital representations of an asset or process that can be paired to a “standard” algorithm to support the management of that specific asset or process. When you move from a local to a platform solution, in other words, algorithm development also includes the creation of digital twins. For instance, the digital twin of a wind turbine collects data on the “history” of that specific turbine (failure events, repairs, past configurations, etc.) and couples it with a standard predictive maintenance algorithm from the Predix platform to reduce downtime and increase energy generation. Also, the deployment of digital solutions in a platform system must accommodate local interface requirement. For instance, different electric companies – say Germany’s E-ON versus Spain’s Iberdrola – may require different information displays for their technicians. Compared with a pilot, a platform solution must embed a customizable interface that eases its adaptation to local conditions.
Of course, the development of digital platforms requires central oversight to ensure that safety and quality standard are met by locally developed solutions. Likewise, proper incentives should be provided to local innovators. Developing scalable solutions entails a greater effort than developing local solutions: it is essential that some of the global benefits generated by the scalable solution be shared with the local innovator in order to make this effort attractive.
When an organization wants to assess the readiness of its digitalization roadmap, it must be clear that the maturity of the organization on a certain type of solution may be low – say at the “groundbreaking” stage, while for other types of solutions it may be higher – say at the “pilot” stage. What is important is to understand the overall picture. And that at each stage, the goals pursued by digitalization initiatives are different.
“They always say time changes things, but you actually have to change them yourself.” – Andy Warhol