Asset Data Model Blog Header

Top 10 Reasons Why Digital Transformation Strategies Fail

Industrial organizations are always looking for improvements in operations management. Every so often, a big leap forward kicks off a new epoch for industrial companies. For years now, visionaries have recognized the massive potential digital transformation holds to be the next big leap forward. It has the potential to be the next revolution in operational performance improvement, bringing prognostics, months ahead predictions, and root cause analysis. So, where are all the digital transformations then?

There are many different reasons as to why digital transformation attempts have been stymied for industrial organizations. At Element, we’ve heard about them all. We’ve helped our customers overcome many of these through our software, and we wanted to share some of the most common reasons we see for why digital transformation initiatives stumble.

1. Confusion about digital transformation and its objectives.

Digital transformation seems like it can include a lot of efforts and initiatives, but when you get down to it, it’s about bringing Big Data and machine learning capabilities to operations. By turning advanced analytics and machine learning into capabilities that can support your other strategies and initiatives, you can realize huge improvements in operations performance and management.

2. Punting the strategy over to IT, and failing to bring IT and OT together.

IT takes an IT-focused approach. Digital transformation uses IT capabilities and tools, but is built on operational data and business drivers. Digital transformation is more than just about bringing the technologies together. For it to succeed, digital transformation needs to be a marriage of IT technology with operational business needs. It’s about getting the two owners -- IT and OT -- to align and understand the goals that they want to jointly achieve. Failing to bring the two organizations together results in IT tools deployed that don’t work well with operational data, or don’t end up providing value in the way that operations needs to drive improvements.

3. Avoiding the cloud.

The cloud is a key enabler for digital transformation. In order to be able to realize advanced analytics as a capability, you need to connect and join your operational data in a data lake (can be hundreds of data sources across dozens of locations). That’s extremely tricky to do without involving the cloud. If your data historian is not in the cloud, you’ll end up pushing time-series data into an unstructured database without context, rather than aggregating it together as a holistic enterprise view. It’s like dumping puzzle pieces onto a table instead of actually putting the puzzle together. And with a contextualized data lake, your distributed storage infrastructure works well with the distributed compute infrastructure required for machine learning activities. Those concerned about security will be happy to know that today’s cloud environments more than meet security requirements for industrial organizations. Many digital and analytical tools can be deployed on private cloud environments that are contained solely within your organization - or what’s called “Managed SaaS” - and can meet the strictest encryption and control policies. Don’t make the mistake of assuming up front that the cloud won’t pass muster in your organization.

4. Relying too heavily on OEMs or Systems Integrators to introduce it.

OEMs often sell you rules-based systems because they know how their equipment operates, and rules-based systems are designed to optimize the systems they’re designed for. But out in the field, that equipment doesn’t work in isolation, it works in a larger context of other equipment and processes. Rules-based systems aren’t capable of handling changing systems and contexts for which they weren’t designed. SI’s can be helpful in managing the build portion of digital transformation. But after your data models are built, what happens when the physical environment changes? The better bet is take a data-out approach, starting with systems that can join and normalize your data, turning it into a flexible asset that can support any analytical tool or system.

5. Thinking “more sensors” is the answer.

More sensors can help you get a more complete understanding of your equipment, and it indeed may be the case that you don’t have enough sensing, but either way, you already have enough data to surface high impact insights. The problem isn’t with a lack of data streams; it’s with a lack of ability to join those data streams together in conjunction with relevant data from EAM systems, LIMs systems and more.

6. Engaging a “Seal Team” data science organization.

Data science isn’t something that just works. Data scientists must have the subject matter expertise to understand the context of your data if they’re going to be able to handle operational data. Most data scientists’ time is spent on data wrangling, so if your data scientists don’t have a deep understanding of the data they’re trying to wrangle, there’s no way they’ll be able to properly join different metadata and data streams in the way that correctly builds out a representation of your assets.

7. Trying to prove digital transformation out as a use case.

Digital transformation is wide ranging and on-demand. A project-based or use case-based approach treats it like any of your other applications or projects, and doesn’t let the digital transformation happen. This isn’t another purchase; it’s a reshaping of what’s possible in your organization - turning analytics into a capability. Instead of hand-stitching data together for one-off use cases, “analytics as a capability” requires that your data is ready to be asked any question at anytime, for any individual or any use case. This is possible by creating an aggregated data set that breaks down all the functional system silos, and provides layers of context on the data.

8. Failing to realize that digital transformation requires software to do the heavy lifting.

As you’ve been reading this, you may have realized that on-demand analytics requires your data to be constantly ready to support it. That means that data needs to have a digital representation of the physical environment (a blueprint to help you find the data you need - the context of all the parts). The challenge is maintaining this context when physical environments are constantly changing. This can’t be done with humans - it needs to be automated so it is reflective in real-time AND consistent.

9. Putting all your digital transformation chips on one vendor.

No one company understands your operations and all your use cases, nor should they control your data and sell it back to you. Your digital transformation technology stack should be modular, letting you continue to leverage your existing OT technologies and your newly acquired analytical technologies and bring those together. Those technologies that promise a full-stack solution will likely not provide you the entire stack and expect you to stitch all your data together (or ask you to pay a large SI bill to do so), and also likely not provide you with the flexibility to tackle all the use cases your operators, engineers, and analysts want to go tackle.

10. Being happy to be the first to go second.

Waiting for everyone else to implement digital transformation initiatives before you do means that you’re already too late. We’re seeing various companies utilize our technology to implement digital transformations within their organizations. All of this is possible with integrations and connections to their existing OT stack, connected to a highly contextualized data lake, and using off-the-shelf analytical products that these companies already own, in a cloud environment that they own as well.

  • O&G Upstream companies are using it to predict which ESPs are going to fail 2 months ahead of time.

  • Downstream Chemicals and Refining companies are identifying root causes to persistent problem on process units, and performing what-ifs on corrosions

  • Power Generation companies are using it to meet regulatory compliance on mercury, and perform RUL on key component.

  • Mining companies are using it to predict failures on limestone crushing circuits and perform haul truck optimization.

Digital transformation can lead to a step change in operational efficiency for industrial organizations, but it’s not a simple strategy to implement. With some consideration and a data-first approach, you can be on your way to on-demand analytics in no time.