Types & maturity of digital twins · Digital Twin primer

Question 1 — how the data flows

Digital model → digital shadow → digital twin

This is the most useful distinction, from academic work by Kritzinger and colleagues. It is defined entirely by how the data moves between the real object and its virtual copy. Only the third one is a true twin.

Level 1

Digital model

↔ updated by hand

A virtual model with no automatic link. If the real object changes, a person must manually update the model. A CAD file or an offline simulation sits here.

Level 2

Digital shadow

→ one-way, automatic

Data flows automatically from the object to the model, so the model always reflects reality. But changes in the model do not flow back. A live monitoring dashboard is a shadow.

Level 3 · the real thing

Digital twin

⇄ two-way, automatic

Data flows both ways, automatically. The model mirrors the object and can change it — a recommendation acted on, or a command sent straight back to the machine.

Question 2 — how big is the thing

The four scopes

Twins nest inside each other, like Russian dolls. A twin can be a single part, a whole machine, a system of machines, or an entire process.

SCOPE 1

Component twin

The smallest unit — a single critical part, like a bearing, a battery cell, or a valve.

e.g. a turbine blade

SCOPE 2

Asset twin

Two or more components working together as one product — a whole pump, engine, or vehicle.

e.g. a jet engine

SCOPE 3

System / unit twin

Many assets working together — a full production line, a wind farm, a building's systems.

e.g. a factory line

SCOPE 4

Process twin

The widest view — how whole systems interact to run an operation, a supply chain, or a city.

e.g. a whole city

Question 3 — how smart is it

The maturity ladder

Twins get more capable as you climb. Each rung answers a harder question — and most real deployments today sit on rungs 1–3, reaching for 4.

1

Descriptive

Mirrors the current state. A live, accurate picture of what the object is doing right now.

Answers: “what is happening?”

2

Informative / diagnostic

Adds context and analytics — surfaces anomalies and explains why something is off.

Answers: “why is it happening?”

3

Predictive

Uses history and simulation to forecast the future — when a part will fail, when to service.

Answers: “what will happen next?”

4

Comprehensive / prescriptive

A "living" twin that not only predicts but recommends the best action to take.

Answers: “what should we do?”

5

Autonomous

Closes the loop itself — decides and acts on the real object without a human in the middle.

Answers: “act on it, automatically.”

These three questions are independent. You could have a predictive asset twin with two-way data (a self-optimising jet engine) or a descriptive process twin with one-way data (a live city dashboard). See how industries mix these →