Digital Twins: Why the Confusion, Why It Depends and How to Make Sense of It

If you’ve ever dived into any sort of digital twin discussion, you’ve probably noticed a sea of buzzwords, conflicting definitions and, let’s be honest, a fair bit of vendor hype.

One person says, “It’s just a 3D model.” Another insists, “Nope, it’s AI-powered and predictive.” Yet another argues, “It’s not a real digital twin unless it’s synced in real time.” Or you hear that “it also has to be an immersive experience”.

Now, as many of ye know, I’ve been down the digital twin rabbit hole for many years. But the fact of the matter is that a lot of folks out there today have a good dose of digital twin fatigue.

So, why all the confusion? And how on earth do you make sense of it all?

Well, let’s start with the three main challenges I keep seeing folks wrestling with:

• Why digital twins are so hard to define
• Why it all depends on the capabilities you need
• How to use capability and maturity models to make smarter decisions

Why is there so much confusion around digital twins?

1. It’s a catch-all term

As I’ve said for years, a digital twin is simply a digital representation of something. But as that could mean anything, let’s step back a bit.

The origin of who said “digital twin” first is up for debate, but NASA was one of the first to use the term in a practical sense back in the early 2000s. Now the term covers everything from simple 3D models to AI-driven simulations of entire cities.

It’s like calling both a bicycle and a space shuttle a “vehicle”. Technically true, but not exactly helpful when you’re trying to make business decisions.

2. Different industries, different expectations

Depending on what industry you are in, everyone has their own industry-related expectations, and they are all different.

• Energy and utilities: Used for things such as grid management, power plant monitoring and asset optimisation.
• Manufacturing: Might focus on factory operations and predictive maintenance.
• Healthcare: Used to build patient digital twins to test treatments before going live.

Each and every industry has its own set of definitions, which leads to very different expectations about what a digital twin should do.

3. Vendor marketing unintentionally makes it worse

A CAD company might say:

A digital twin is a 3D model.

An AI company might say:

It’s only a digital twin if it predicts future performance via AI.

An IoT provider might claim:

No, it’s only a twin if it streams real-time sensor data.

A simulation company may say:

It has to have physics-based simulation capabilities.

An XR company may say:

It has to be an immersive experience.

Some may say:

It has to be photorealistic to train the robot AI.

The list goes on, and on, and on.

And let’s not forget the part a company’s marketing strategy can play. I’m sure many of us have been involved in various marketing or go-to-market strategies where the focus was:

We need to have our own unique term that makes us stand out.

Thus, we have so many companies out there today with their own unique digital twin-related terms.

Why it all depends on capabilities, not just labels

Instead of getting caught up in whether something qualifies as a digital twin or not, focus on what your twin needs to do. Start by thinking in levels of capability.

For example, a Descriptive Twin might be as simple as a 3D model, such as a 3D CAD model or a reality-capture-derived 3D model created using lidar. It’s static, but it can still provide useful information.

Then you have an Informative Twin, which takes it up a notch by integrating real-time sensor data, like a digital twin of a factory with live performance metrics. This gives you more context to understand what’s happening in the here and now.

Next up is a Predictive Twin. This is where AI starts to play a role, forecasting future outcomes. Think of a power grid twin predicting outages before they happen, giving you the chance to prevent issues before they escalate.

If you’re looking for more advanced capabilities, you’ll need a Prescriptive Twin. These models go beyond predictions by simulating different what-if scenarios to optimise operations. For instance, a self-adjusting industrial system could automatically tweak settings to maximise efficiency on the factory floor.

At the highest level, we could have an Autonomous Twin: a self-learning system that adapts with minimal human intervention. A great example is a city traffic twin that continuously optimises traffic flow in real time, learning and adjusting as conditions change.

The real challenge is that people often mix up these different levels without realising it, so things can get very confusing, fast.

Enter a maturity model

But the conversation doesn’t stop there. The next piece of the puzzle is the maturity model. This is where you evaluate not just the capabilities, but also how ready the technology is to scale and how it aligns with your organisation’s needs and market trends.

In simple terms, a maturity model helps you assess where you are on the curve of adoption and how far you still need to go to fully leverage the digital twin’s capabilities.

You’ll also want to think about how the technology fits with your organisation’s existing infrastructure. Does it integrate seamlessly with your current systems, or will it require significant changes?

A key part of the maturity model is assessing technology readiness. Is the digital twin technology mature enough to meet your needs today, or are you betting on something that’s still in the development phase?

And don’t forget market acceptance. How widely accepted is this technology across your industry? If it’s something cutting-edge with limited market adoption, you might face hurdles in terms of both integration and scalability.

Then there’s the elephant in the room: cybersecurity and compliance.

As with any new or advanced technology, there are risks, especially when sensitive data is being handled. The more complex and interconnected your digital twin becomes, the more critical it is to ensure compliance with industry regulations and standards, whether that means data privacy laws, cybersecurity protocols or other requirements.

Cost is also a big factor to consider. How much is all of this going to cost, both in terms of initial investment and ongoing maintenance?

Digital twins, especially at higher levels of capability, can get pricey. It’s important to weigh those costs against the value they deliver. Factor in expenses such as software development, integration with existing systems, training and ongoing updates.

Another consideration is deployment. Should the digital twin live on-premises, in a private cloud or in a SaaS environment?

Each option has its pros and cons. On-premises deployment offers more control but can be more costly and resource-intensive to manage. Cloud-based solutions, on the other hand, offer flexibility and scalability, but they come with questions around data security and compliance.

Finally, integration with existing systems is crucial. A digital twin that doesn’t work well with your current infrastructure is like a square peg in a round hole.

A successful digital twin solution should not only integrate with your existing data and operations, but should also be able to scale over time as your needs evolve. That’s where some sort of maturity model helps you gauge how well the technology can adapt to your business as it grows.

My two cents: how to cut through the noise

Instead of focusing on what a digital twin is or is not, ask yourself:

• What capabilities does it need?
• How mature is the technology in terms of readiness, scalability and alignment with your company’s goals?

A few practical points:

• Forget the hype and focus on what you actually need. Digital twins range from basic models to AI-powered systems, so define your use case first.
• Think in capabilities, not labels. Descriptive, predictive and autonomous twins each have different uses. What do you need yours to do?
• Use maturity and capability models to guide decisions. These frameworks help benchmark progress and prevent zombie projects or unnecessary technology.
• Avoid vendor confusion by asking the right questions. Instead of asking, “Is this a digital twin?”, ask questions related to the capabilities you need. Does this twin have real-time updates? AI? Predictive capabilities?

The bottom line is that digital twins aren’t one-size-fits-all.

Now, you might be thinking:

Ah Kev, you are oversimplifying all this.

Fair point. But the reality is, if we want to move away from the “my digital twin is better than your digital twin” debates, we’ve got to level-set on the capabilities we’re actually talking about when it comes to digital twins. It’s about aligning expectations with reality.

And here’s the kicker: be honest with yourself about what you actually want and, just as importantly, where you are today.

I’ve worked on a few projects where management had their hearts set on a Predictive Twin. But when I asked about their progress on integrating IoT data into their existing workflows, the response was:

Not great…

If the foundations are not there, it’s hard to build something new on top.

A digital twin is just a tool. And like any toolbox, there are different tools, each with its own level of intelligence, interactivity and automation. The trick is choosing the right one for the job at hand.

Also, if anyone is looking for some hands-on guidance, feel free to reach out for a workshop or advisory engagement. I’d be happy to help ye bridge the gap between all the hype and what ye are actually looking to achieve.

All comments and feedback welcome.

Kev.

For reference

And before you ask, no, I didn’t just make all of this up over the weekend.

The thoughts I have shared above are basically how I strive to explain all this to folks, based on various models I have come across over the years. Here are but a few:

• The Digital Twin Consortium Digital Twin Capabilities Periodic Table
• Digital Twin Consortium: Infrastructure Digital Twin Maturity: A Model for Measuring Progress
• Digital Twin Maturity Model, a white paper by Malte Heithoff, Judith Michael and Bernhard Rumpe, Software Engineering, RWTH Aachen, August 2024
• Digital twin maturity levels: a theoretical framework for defining capabilities and goals in the life and environmental sciences, by Brett Metcalfe
• The 250 classifications of a Digital Twin, via IoT Analytics

And there are so many others from various companies such as Siemens, Unity, Dassault Systèmes, ABB, Bentley, Accenture and Gartner.

View the original post or source →