TL;DR

#

A Digital Twin needs accurate data, not perfect data — start small and iterate. Break down silos to integrate different views; analytics complements physics and experts. Not everything needs real-time; measure value pragmatically and avoid overpromising.

🧠 Data Modelling, Analytics and Digital Twins

#

Some key ideas from experience (not from the trend cycle)

#

When we talk about Data Modelling, Advanced Analytics and Digital Twins (DT), the conversation often focuses on technology. However, experience shows success rarely depends on technology alone.

I want to share three key ideas that often make the difference between a pilot that stays in PowerPoint and a solution that truly delivers value.

🔹 1. Good data for good models

#

It’s well known: no good models without good data. But it’s important to clarify one fundamental point 👇

👉 Good data doesn’t mean all the data.

• In industry, most sensors were designed before IIoT and even before the Internet
• Today we work with adaptations, bandwidth limitations and transmission costs
• In many cases, only a fraction (5–10%) of what is measured can be sent to the model

That’s why waiting for the perfect dataset is usually a recipe for never starting. A much more effective strategy is to adopt fast cycles of:

🔁 test – fail – fix

Don’t wait too long to validate ideas. There will be time to fail… and to improve.

🔹 2. Variety breaks silos

#

One of the biggest values of a Digital Twin is that it enables multiple disciplines to view the same asset from different perspectives.

Think of a well:

• A geologist describes it one way
• A production engineer, another
• An operations specialist, another

And they’re all right.

✨ That diversity is exactly what makes a DT powerful.

To build it, you need to:

• Break data silos
• Also break silos in requirements and processes

Here a key idea appears:

It doesn’t matter who owns the data, what matters is the problem we want to solve

As someone once told me (and I never forgot): ❤️ “Don’t fall in love with the technology, fall in love with the problem.”

🔹 3. Analytics is not the new physics

#

This point is critical and sometimes uncomfortable.

• We have over 100 years testing and validating physical fundamentals
• We know what works and what doesn’t
• That’s why experts exist, inside and outside companies

Advanced analytics doesn’t come to replace that knowledge. It comes to:

📈 extend it, complement it and scale it

Analytics:

• Uses large volumes of data to support hypotheses
• Helps describe complex behaviors
• Challenges experts to rethink known solutions
• Even allows formulating questions we hadn’t imagined before

It’s not “models vs. experts”, it’s models + experts.

⏱️ Does a Digital Twin only work with real-time data?

#

No.

A DT can be fed with both:

• 🔴 Real-time data
• 📦 Batch (historical) data

Not all functionalities require real-time.

Examples where real-time adds value:

#

• Online monitoring and recommendations
• ROP optimization
• Prediction of tripping and casing speed
• Predictive maintenance

⚠️ Main challenges to implement advanced analytics

#

In practice, the biggest obstacles are often not technical:

• 💰 Technology and TCO (Total Cost of Ownership)
• 🏢 Organizational alignment
• 📉 Lack of digitized information
• ❓ Unclear business value
• 🧠 Over-designing solutions nobody needs

💵 Can the value be measured in USD or barrels?

#

Sometimes yes, sometimes not directly.

A useful analogy:

📱 Can you measure the value of having a smartphone in your pocket?

Maybe not exactly, but you can measure the capabilities it enables.

Typical impacts:

• Reduction in OPEX and CAPEX
• Predictive maintenance vs calendar-based maintenance
• Automation of forecasts and prediction processes
• Less dependence on on-site specialists (Excellence Centers)
• More simulation on computers, fewer field tests

📌 Practical rule:

20% of the variety of data usually generates 80% of the value.

🚧 What to do if you still don’t have the right information?

#

The reality is simple:

nothing is perfectly aligned from day one.

There are always pending items:

• Change management
• Processes
• Budget
• Data availability and quality

Also, many sensors require:

• Hubs
• Aggregators
• Third-party networks
• Encryption and adaptations

The key is:

• Use models only for what the data allows
• Don’t over-extrapolate
• Don’t promise more than you can deliver

👉 We’re not doing academic research. 👉 We’re accelerating decisions and solving real problems.

📌 In short

#

A Digital Twin is a digital copy of a real asset that uses data to understand its behavior.

It doesn’t need to be perfect or have all the data from day one.

• ✔️ Start small
• ✔️ Integrate different perspectives
• ✔️ Rely on physics and existing knowledge
• ✔️ Use analytics to expand what you already know

And above all:

🚀 fewer promises, more real value.