Maybe you’ve heard this type of story before: an industrial company announces the start of a major digital initiative. Time goes by, but nothing seems to happen. After many months, and many meetings, there is little to show for the efforts.
Unfortunately for many industrial companies, the tantalizing potential of the power of IIoT, machine learning, and digitalization is proving difficult to realize.
Why does this happen?
Some might point to the software infrastructure – the lack of unified data management systems, or the challenges of connecting legacy information technology (IT) and operating technology (OT) systems to newer cloud technologies. Others might blame the quality or sufficiency of data. Or, when data is available, perhaps the challenge is the lack of capability and understanding in developing appropriate analytics for this data. Organizationally, there may also be challenges related to who owns the problem (and the budget) – is it the IT group, the operating units, the control and instrumentation team, or a corporate digital transformation group?
In fact, the challenges often lie in these areas. However, another problem is also a major factor: a clear and detailed understanding of who the users are and how a new industrial internet of things (IIoT) application will help them make better decisions.
In other words: how will an IIoT investment inform a decision process in a way that’s different and better than the status quo?
It’s a combination of problems
IIoT implementation is, in fact, a software problem. It’s also a data and/or a data science problem. At its root, however, this is a business problem, related to the complex orchestration among business needs and stakeholders, the data analytics approach, and software requirements.
You need a unified approach
The dirty secret of stalled or failed IIoT projects isn´t any specific discipline, department, or infrastructure need. It’s the unified approach multiple areas must take, and the joint problem-solving and complex orchestration they must agree upon. This is especially challenging with the legacy systems and highly engineered physical environments at core of heavy industries.
Someone needs to take action and make a decision
An IIoT application typically involves a continuous analysis of sensor data from one or more physical assets, sometimes combined with other types of historical or real-time data (such as physical, financial, operational, or other types of information). The output of this analysis informs a business decision, most commonly a human operator’s decision.
This type of approach may cross many organizational and information silos in a company, and potentially involve supplier or third-party data and organizations (such as GPS, AIS or weather) as well. Ultimately, the decision-maker wants to take actions that will tangibly improve the business – by reducing costs, reducing risks, or improving revenue. The process of making better decisions requires complex orchestration across many stakeholders and systems.
This involves explicit coordination
For instance, sensors and data historians may be tied to existing SCADA or control systems and managed by control and instrumentation specialists or third-party vendors. Additional data may reside in separate data stores, ERP systems, third party databases, or even PDFs or spreadsheets in various business units.
The responsibility for owning and managing data in a centralized way typically falls under IT or information systems (IS) teams, with clear lines of demarcation for transferring data. Many of these teams are highly focused on major infrastructure projects for master data management systems, often including cloud-based data lake projects.
Meanwhile, the analysis of this data may fall under a data science or analytics team. The typical data scientist in industrial companies today, is relatively new to the company and the industry, or they may have moved from other quantitative disciplines, such as engineering design, simulation, optimization, or highly technical physics-based analysis, into a newer role in data science. As a result, most industrial companies are still in the early stages of understanding what data science means for their company, and how best to deploy these skill sets.
However, there’s not much data scientists can do without sufficient amounts of data in the right structure, format, and cleanliness. The challenge of data engineering – getting data unified, cleaned, and prepared to run through complex data science models, is often a hurdle. The consistent data handoff among OT, IT and data science teams is often an even bigger hurdle.
Data scientists are very good at building models and analyzing complex data sets, but ultimately, they’re not closest to the business problem. They have limited ability to advise a decision-maker on which problems are most valuable and important. Instead, decision-makers are often left on their own in deciding how and when to incorporate data analytics into decision processes. This challenge of selecting the right problems for IIoT analytics is perhaps the most significant hurdle for many industrial companies.
Make sure you understand the risks of falling behind
A final challenge is the threshold of both time and value by which most companies measure themselves. If the business believes that IIoT projects must be on hold due to infrastructure concerns, or a lack of data science capability or direction, then it’s easy to maintain the current state without regard to the risks of falling behind in an evolving market.
How should you do it?
Start with the end user in mind. When companies start thinking of IIoT projects in terms of cloud-based software development projects with short sprint cycles, rapid prototyping, hypothesis testing, and so forth, things start to change radically. Starting with the end user in mind – their needs, actions, and decision parameters – the business process, data infrastructure, and analytical approach can be rapidly trialed and tested, often in weeks rather than months.
The enabling technologies for IIoT are increasingly available and affordable to many industrial companies. Similarly, industrial data analytics techniques are maturing by the day. All of these are tools that industrial companies can use to improve their businesses. But the companies that succeed with IIoT start with the business problems they need to solve and back into testing rapid, lower-cost solutions, rather than moving in a straight path based on the constraints of the existing IT, OT, and legacy business understanding.
Want to learn more about the Industrial Internet of Things, how it will disrupt traditional thinking and its barriers? Have a look at our whitepaper: The Executive Guide to the Industrial Internet of Things.