Why many digital maintenance initiatives struggle in practice and the practical challenge of aligning assets, teams, and technology to realize value?
Digital transformation in maintenance is often seen as a technology overhaul or the adoption of new digital tools. However, the real difficulties are usually less visible. Who drives the decision? Does the monitoring fit the actual machine behavior? How do separate systems and vendors work together? And do emerging digital tools support plant expertise or bypass it? This article looks at some of the practical challenges plants face in turning digital transformation ambition into reality.
Introduction: The vision is attractive, but the real barriers are less visible.
Across industrial segments, digital transformation in maintenance and asset performance management has become a familiar ambition. Plants are exploring connected sensors, online condition monitoring, predictive maintenance, cloud platforms, analytics, and AI-driven diagnostics to improve visibility, reduce downtime, and enable better maintenance decisions.
Yet in practice, this journey is rarely straightforward. The challenge is often not the availability of technology, but how to make it work in the realities of an operating plant. The people who recognize the technical need may not be the ones who can secure budget or drive implementation. Different aspects of machine health may sit with different teams, vendors, and systems. Pilots may show promise but struggle to scale. And as AI-based tools become more common, plants still need to make sure these tools support field expertise rather than replace it.
For these reasons, digital transformation in maintenance is not just about adding new technology. It works only when the technology fits the machine, the plant, and the people using it.
A common reality: The people who see the need are often not the ones who can drive the project forward.
In many plants, the first people to feel the need for digital monitoring are maintenance or reliability engineers working close to the equipment. They see repeated failures and missed warning signs and understand the limits of manual inspection. They often know where better data or online monitoring could help.
But these teams often do not control the budget, plant priorities, or cross-functional execution. Management may ask for a clear ROI or wait until the initiative fits a target business priority. As a result, there is often a gap between technical need and organizational action. In some cases, the decision is made top-down, with a solution selected before field team members are fully involved. When that happens, the final system may not fit the realities of the field.
Plants do not buy “digital transformation” out–of–the-box. They end up inheriting a fragmented ecosystem along their digital transformation journey.
Plants rarely buy one complete digital transformation or even a digital maintenance solution that covers everything end to end. Instead, they often deal with separate vendors for vibration monitoring, power quality or electrical monitoring, IIoT integration, and front-end software. Each may solve one part of the problem well, but the plant still has to make all of it work together.
This creates a fragmented ecosystem. The front-end platform may present the data but may not fully understand the field realities of vibration or electrical diagnostics. Domain specialists may understand the machine signals well, while broader plant integration may be handled by others. At the same time, electrical and mechanical teams inside the plant may also work separately. The result is that the plant is trying to solve one combined asset-health problem through multiple disconnected systems, vendors, and teams.
A big part of the problem sits in the hidden layer: protocols, gateways, networking, and data access.
What we see in the dashboard is only one part of the system. A large part of the real complexity sits in the hidden layer between the sensor and the software - communication protocols, gateways, networking, and how data is actually made available.
In practice, industrial sensors differ widely in what they expose. Some send data only to their own cloud platform. Some provide 4-20 mA outputs. Some support MQTT (Message Queuing Telemetry Transport), often through a vendor gateway. A smaller number support interfaces such as Modbus (over RS-485) or Modbus TCP (Transmission Control Protocol) that can be more easily integrated into broader plant systems. Because of this, plants may find themselves pulled into one vendor ecosystem or working with an integrator who chooses devices mainly based on interface compatibility and the integrator’s engineering resource constraints. This hidden layer often determines whether a monitoring deployment becomes part of a larger digital workflow or remains an isolated system.
Monitoring strategy must match the machine. For this, decision-makers must be technically educated on the options and not get swayed by a specific product’s marketing claims.
Choosing a monitoring strategy is not simply a matter of selecting the newest or most heavily marketed product. Different monitoring approaches - portable systems, battery-based online systems, and wired continuous systems - each have a place. The right choice depends on the machine, its operating pattern, the fault behavior, the criticality of the machine, and the type of visibility the plant actually needs. Of course, the price of the solution, post-sales support, and ease of deployment are practical considerations too.
Decision-makers need enough technical understanding to compare the options properly. Some machines may be well served by periodic portable measurements, others by interval-based automated monitoring, and some may need more continuous visibility. The important question is not which product sounds most advanced, but which monitoring approach best fits the machine and the maintenance objective.
In one precision manufacturing application, continuous monitoring of CNC spindles proved more suitable than infrequent measurements, since high-speed assets, where failures can manifest rapidly, often need greater visibility into machine behavior. In another case, a shipboard CBM architecture combined online monitoring for critical assets with portable monitoring for non-critical assets, with both feeding a common analytics platform. Together, these examples show that the right monitoring strategy depends on the machine and the operating context, not on a one-size-fits-all technology choice.
Even the right technology can fail if it doesn’t take into account the machine’s unique operational states.
Surprisingly, even when the appropriate monitoring technology is chosen, it may still fall short if it is not aligned with how the machine actually operates. In practice, machine behavior often changes across different operating states, and the monitoring system must have that context. Sampling interval, on/off state handling, operating cycle, and measurement location can all affect whether the data is truly useful.
In one steel processing application, measurements taken once every five minutes were not aligned with the actual cutting events where the most relevant vibration behavior occurred. To address this, current or power measurement was explored to identify when cutting was actually taking place, but doing so downstream of a VFD required a more application-specific approach. In a CNC application, machine behavior also had to be understood in terms of operating state, and machine learning-based state detection was used to distinguish four states, including off, idle and two different job states. These examples show that even the right technology must still be adapted to the machine’s real operating conditions.
Machine health is often split across teams and systems, even when the machine is one system.
In many plants, machine health is still viewed through separate technical and organizational silos. The mechanical team may focus on vibration, the electrical team may focus on current or power quality, and another team may look at energy or plant-level software. Each view is valid on its own, but the machine itself doesn’t operate in silos.
This can make digital transformation harder than it appears. A machine problem may involve mechanical, electrical, and process-related behavior at the same time, while the plant’s tools, teams, and vendors remain separated. Recently, some plants have begun explicitly asking for a single platform that can bring together both mechanical and electrical condition data. This reflects a broader reality: while machine health is one combined problem, the teams and even the traditional vendors serving it are often separate. Digital transformation becomes more effective when these data streams, teams, and monitoring systems are brought together rather than managed separately.
AI and analytics must support plant expertise, not replace it.
As AI-based diagnostics and analytics become more common, plants need to use them in a way that strengthens, rather than sidelines, field expertise. These tools can help identify patterns, flag anomalies, and bring attention to assets that need closer inspection. But the reality is that they operate within the limits of the data they receive and the logic they are built on.
Field technicians and engineers bring something unique and critical: plant-specific context. They understand how a machine normally behaves, what recent maintenance has been done, how operating conditions vary, and which alerts truly matter. The most effective use of AI and analytics is therefore not to replace this knowledge, but to support it by helping plant teams focus on the right issues, investigate faster, and make better decisions.
In practice, the workable path is often hybrid and incremental.
In many plants, digital transformation cannot and does not happen all at once. Budget, operating constraints, and the scale of the plant often make a gradual approach more practical. Instead of waiting for a full plant-wide rollout, it is often more realistic to start with a narrow, focused scope and build from there.
This is also why hybrid monitoring strategies are common in practice. Critical assets may justify online monitoring, while for other assets, periodic checks are sufficient. In one shipboard CBM architecture, critical assets were planned for online monitoring, while non-critical assets were to be monitored using portable routes, with both feeding a common analytics platform. This is a practical example of a phased and hybrid approach shaped by asset importance and operating realities.
Conclusion: The challenge is not just adoption of technology, but alignment across decisions, systems, and plant reality.
Digital transformation in maintenance is often discussed in terms of technologies - connected sensors, software platforms, and AI-based tools. But in practice, the bigger challenge is often alignment - between decision-makers and field teams, between different monitoring systems, and between the technology and the way machines actually operate.
Plants are more likely to see value when digital transformation is treated not as a one-time technology upgrade, but as a practical effort to align monitoring strategy, system integration, team ownership, and field knowledge. The goal is not just to collect more data, but to help plants make better maintenance decisions based on how their machines actually operate.
