More companies are embracing IIoT to drive better business outcomes, but they still haven’t made the most of the data that they’ve collected. As one might expect, IIoT produces massive amounts of data at record speed. It can be overwhelming, especially to industrial companies, which are often strong in their real-time operations but struggle to translate data produced from those operations into meaningful and systemic process improvements.
While digital technologies could play a key role in unlocking value in the business and enabling growth, 93% of utility executives believe that they are struggling to deliver the benefits of digital transformation, according to Accenture research. Furthermore, as is frequently the case in large industrial companies, different departments often have access to different data resources with little cross visibility.
Data overload occurs because the parameters of the IIoT system, the type of data and the rate of change often occur faster than most companies’ ability to adapt, which can result in technical debt. There are quick fixes you can add to the system to accommodate the near-term needs of the data ecosystem, but that’s not how to build a system that lasts well into the future. To really make the most of IIoT systems, we need to rethink how to manage the entire ecosystem in a way that’s future capable.
Converge data for a streamlined approach
Harboring multiple data environments with disparate perspectives is the challenge to getting companies fully digitally transformed and is often a source of infrastructural cost overruns. The first step towards rectifying these issues is data convergence, which is a critical component in using data to drive better business decisions. A wide breadth of data will provide much needed context for many key decisions or analyses based on multiple perspectives of an event. Depth of data will enable a stronger reinforcement of these decisions based on more examples of similar events.
Because of this, it’s imperative that an organization’s IIoT platform allows the analytical framework to both easily access representative data samples from a wide variety of data sources and be scalable enough to dive deeply once an analysis strategy is identified. However, most industrial organizations have not yet found the solution to making these systems work, and they often experience continual IIoT project failure.
Rethink data management
Despite our best efforts, data coming from millions of nodes can be incomplete and imprecise, and there is often significant transformation that needs to occur prior to this data being consistently used to empower key outcomes for the industrial customer. It’s important for companies to go back to the basics of data governance to ensure they’re not overloading their IIoT platform.
This includes two key components that are commonly associated with root causes of data overload: the data models and the infrastructure that supports them.
Data modeling. The first step to reducing the likelihood of failure is to have a good understanding of the ideal construct for your data. Often, this is called a canonical data model, and they serve as an appropriate end state for incoming source systems. This is often achieved by defining metadata to organize and translate a nebulous system of scattered data into a more systematized construct. In many industrial environments, the data will be organized into multiple canonical models that can be mapped against each other, as opposed to a universal model that envelops all data that is ingested.
Generally, it helps to transform data into multiple canonical models that can be cross-referenced against each other. This federated strategy enables a more rapid transformation of data removed from disparate source systems into models that can be cross-referenced against each other with greater ease.
Another trick to reduce data overload includes having a down-sampling strategy of incoming data for optimal performance. Not all data is created equal and there is often a cost trade-off — in the form of dollars or time — with having to process more data. Understanding the appropriate granularity of data and aligning it with the right storage and orchestration scheme will help reduce compute cycles and cost.
Data infrastructure. From a data management perspective, it’s important to have a solution for storage, management and visualization of data that is not just interconnected and scalable, but also appropriate for the needs of the end user. Mapping an infrastructure that is tailored to the unique qualities of your data — its incoming velocity, the amount and the sophistication of the workloads — as well as what decisions the data will need to support will help you decide how much compute is done in the cloud, at the edge or on premises. Considerations also need to be made for the appropriate level of security and privacy for both your end users and the systems that are collecting your data.
In both the modeling and infrastructure scenarios, understanding your domain and the key outcomes that you want will heavily influence your strategy. These outcomes need to be defined by your end user in a way that will positively influence their ability to do their job. After all, that’s the primary impetus for digital transformation. For almost all industrial companies that are data converged, modeled and orchestrated with these end user-driven outcomes in mind, they are the ones that more rapidly achieve value in IIoT projects.
A successful IIoT project can significantly save money by increasing productivity, reducing waste and optimizing key network operations or maintenance processes. But remember, the state of your data is the basis of everything. Deploying advanced analytics and putting your data to work creates an intelligent system that operates with a strong foundation and an agile, repeatable methodology to avoid IIoT project failure.
All IoT Agenda network contributors are responsible for the content and accuracy of their posts. Opinions are of the writers and do not necessarily convey the thoughts of IoT Agenda.