Digital twin or simulation? The difference is the wire
A simulation model is a snapshot built for a decision: assembled from assumptions, computed, discarded. A digital twin is the same model with a live data connection to real operations. It updates its own parameters and mirrors the current state. The difference is not computational depth but connection: it costs interfaces, data quality, operations and maintenance. For a one-off question (is this automation worth it?) a model is right and a twin is expensive ballast. For continuous steering with daily decisions, as in manufacturing or logistics, the twin is the right form. In administrative processes it usually fails on the data foundation, not on the idea.
Table of contents
The definitions, briefly and without excitement
Simulation model: a representation of a flow, fed with assumptions, computed for one question. Lifespan: weeks. After the decision it is archived.
Digital twin: the same model, but permanently connected to real operations. It pulls volumes, processing times, capacities and states continuously from source systems and mirrors reality instead of assuming it. Lifespan: years. It has an owner, a budget and an on-call rota.
The difference is the wire, not the model.
Three levels of connection
| Level | What happens | Effort |
|---|---|---|
| 1: Offline model | Parameters entered by hand, once | Hours |
| 2: Digital shadow | Data flows into the model automatically, one way | Weeks to months |
| 3: Digital twin | Data flows in, decisions flow back into operations | Months, ongoing |
Most initiatives calling themselves "digital twin" are level 2. That is not a criticism. Level 2 is often exactly right. But the return channel that defines level 3 is the one that brings governance, liability and operational responsibility with it.
What the twin additionally gives you
- It is always current. No "the model is from last spring".
- It detects divergence. When reality and model drift apart, either reality changed or an assumption was wrong. Both are valuable.
- It answers questions repeatedly. "What happens if two people are absent tomorrow?" becomes a query rather than a study.
- It can forecast, because it knows the current backlog, not just a volume structure.
What it additionally costs
- Interfaces to ERP, MES, ticket system and time recording, each built and maintained individually.
- Data quality as a permanent topic. A twin pulling wrong numbers is more dangerous than no twin, because it is believed more.
- Operations: who looks when an interface fails overnight?
- Model maintenance. When the process changes, the model changes. Otherwise the twin mirrors a flow that no longer exists.
- Data protection and employee representation, as soon as personal performance data flows.
Practical rule of thumb: the connection costs a multiple of the model. Not knowing that ratio means planning the smaller part and being surprised.
When each is right
A simulation model suffices when:
- the question is one-off ("is this automation worth it?"),
- the decision falls this quarter,
- the process runs outside systems (email, phone, paper),
- nobody intends to steer with it afterwards.
A twin pays off when:
- decisions are made daily or weekly (dispatch, staffing, sequencing),
- a reliable data source already exists,
- the process is expensive enough that an operations team pays for itself,
- reality changes quickly and yesterday's model is worthless.
Typical environments for level 3 are manufacturing, intralogistics and network or plant operations. Typical environments for level 1 are administrative, application and approval processes, not because they are simpler but because the data trail is missing there.
The "digital twin of an organisation"
The term comes from the process mining world and means a continuously log-fed representation of the flows. Two things about it, both sober:
What is right about it: where a continuous log exists, automatic updating is a genuine advantage. The weakest part of any simulation is estimated inputs; a log replaces them with measured ones.
What to check: a representation of the past is not yet a model of the future. Ask the vendor the same questions as of any simulation tool: does it represent queues in front of busy steps or sum durations along paths, and does the result come as a range or as a single number? The distinction between measuring and computing persists even when both sit in one interface; it is covered in detail in Process simulation vs. process mining.
The pragmatic route
Almost nobody needs level 3 at the start.
- Build a model, fed by hand. Half a day. It answers the pending question.
- When the same question arrives a third time, automate the two or three inputs that change most often, usually volume and capacity. That is the move to level 2 and costs a fraction of a full integration.
- Only once decisions are made daily should you consider level 3, with an owner, a budget and clear responsibility.
Follow that order and you have an answer after step 1. Start at level 3 and after six months you have an interface.
Frequently asked
Is a digital twin the same as a simulation?
No. A simulation model computes assumptions for one question and is archived after the decision. A digital twin is the same model with a permanent data connection to real operations: it updates its own parameters and mirrors the current state. The difference lies in the connection, not in computational depth.
What does a digital twin cost compared with a model?
A multiple, and the difference sits almost entirely outside the model: interfaces to source systems, data quality assurance, operations, model maintenance when the process changes, and where applicable data protection and works council procedures. The model itself is the smaller item in both cases.
Is a digital twin worth it for administrative processes?
Rarely, and usually not because of the idea but because of the data foundation. Approvals by email, queries by phone and waiting inside inboxes leave no usable trail. That is exactly where lead time is created. Without that data, the twin mirrors the part of the process that is fast anyway.
What is a digital shadow?
The intermediate level: data flows automatically from operations into the model, but nothing flows back. The model observes; it does not steer. For most initiatives that is the sensible target level. The return channel that makes a true twin brings governance, liability and operational questions that rarely justify the added benefit.
How do I start if I want a twin eventually?
With a hand-fed model for the question currently pending. When the same question recurs, automate the two or three inputs that change most often, usually volume and capacity. Decide on full integration only once decisions are made daily or weekly on the basis of the model. That order gives you an answer after half a day rather than an interface after six months.
Run the numbers on your own process
FlowVisual turns the figures in this article into a model that runs, with your volumes, your capacities, your range.
Guide: seven steps to the number- Comparison
Process mining or simulation? The past against the future
The two get confused because both are called “process analysis”. They answer opposite questions. Choosing the wrong one only becomes obvious once the answer arrives.
Read - Comparison
Process mapping or process simulation? What each one can do
A process map ends half the misunderstandings in a workshop. The other half starts with the first numeric question, and there no diagram helps, whatever the notation.
Read - Comparison
Process improvement software: seven categories, one misunderstanding
“Process improvement software” is not a product but seven product categories with seven different purposes. Buy the wrong one and you end up automating a step that was never the constraint.
Read