What a Digital Twin in Manufacturing Really Is (and What It Is Not)
A digital twin in manufacturing is a working model of your plant that runs production over time: machines with their cycle times, buffers, transport, people, shifts and the real order mix. Its job is to test a change before you make it, such as a new layout, a robot or an extra shift, and to show which option gives more output at lower risk. A 3D model, a CAD drawing or a dashboard of live machine data can be part of a twin. On their own, they are not one.
Why "digital twin" means different things to different people
McKinsey defines a digital twin as a digital replica of a physical object, person, system, or process. The definition is broad on purpose, and the same research found that almost 75 percent of companies in advanced industries already use digital-twin technologies of at least medium complexity. With adoption that wide, the term is used for very different things.
In factories, it covers a 3D visualization of the hall, a dashboard of machine states or a CAD layout. Trouble starts when a board funds a 3D picture and expects answers to operational questions. Where will parts wait? How many forklifts does the new hall need? Without a twin that can answer them, the factory itself becomes the test bench. ERP, MES and spreadsheets record events and add up averages. A twin runs the flow through time with your cycle times, routings, shifts and waiting, so the decision is tested on the model first and the risk of a wrong one drops.
A plant manager needs a narrower working definition: a digital twin is a model of your factory that behaves like the factory over time, so you can try changes on the model first.
What a manufacturing digital twin contains
Most plants already have the data for these layers in some form.
- Geometry: the plant layout (a CAD or PDF drawing) as the base, with 3D models of machines, stations, racks, buffer areas, vehicles and people placed on it.
- Process logic: the sequence of operations, cycle times and changeovers, entered as nominal values.
- Resources: operators, shifts, forklifts and autonomous carts, with their speeds and working rules.
- Demand: orders and product mix for a typical day or week, plus the peaks.
- Rules: how work is released, batch sizes, priorities, the order in which material is picked.
When these layers run together, the model produces numbers: units per day, minutes a part spends waiting, utilization of each station and vehicle, the queue in front of each buffer. DBR77 Digital Twin runs a deterministic simulation with fixed times, so the same input always gives the same result and a difference between two variants comes from the change itself (see Averages vs simulation over time).
Many useful twins begin with ERP data, routing sheets and a stopwatch on the shop floor. If you are starting, use the data you have, even nominal cycle times. The first results will surprise you anyway: comparing the model with what the floor really produces shows the gap between plan and reality, and that gap is where the real bottleneck shows up. Live machine data, for example from DBR77 IIoT, is the next stage, once the model is used regularly and you want it to follow the plant.
Is it a digital twin? A quick test
Ask one question: can you change something in it and see what happens to output and waiting time? If you can't, you have a drawing, a picture or a dashboard, each useful in its own way.
| What you have | Is it a digital twin? | Why |
|---|---|---|
| 3D walk-through of the planned hall | Not on its own | Shows how the hall looks, not how it runs |
| CAD layout with cycle times in a spreadsheet | No | Works on averages; no queues, no waiting, no interaction between stations |
| Dashboard with live machine status | No | Shows what is happening now; cannot test a change before it is made |
| Model that runs a week of production with your order mix and compares layout variants | Yes | Tests changes before they happen, with numbers you can compare |
| The same model, fed with live machine data | Yes, a more mature stage | Compares the model with the plant every day; a later stage, not required to start |
What a twin answers that a drawing cannot
A car lamp manufacturer in southern Poland, with 270 people on two shifts, needed a new production line layout with the highest output and the best use of people. The hard part was time: how long material spends moving between workstations and intermediate warehouses, and how long it waits to be picked up.
DBR77 built a digital twin of the material flow, with standard transport speeds and even the acceleration and braking of the trucks, and compared three layout variants. The result shows where drawings mislead:
- The winning layout placed a buffer warehouse in the central area, between production and assembly.
- Its average material flow time was not the shortest of the three.
- Its daily output was still on average 6% higher than a linear layout without an intermediate warehouse.
- In the other variants, semi-finished parts lost time waiting for pickup between processes.
On a drawing, the shortest route looks best. A model that runs time shows that waiting for a forklift costs more than a few extra meters of transport.
Buffers, carts, robots and shifts follow the same logic; see 5 Factory Decisions You Should Simulate Before You Commit.
See which decisions a digital twin supports
Layout changes, intralogistics, warehouse picking, line balancing and shifts: see how DBR77 Digital Twin is used in each case.
What a digital twin is not
A 3D render is a different product. DBR77 built a virtual showroom for Fanuc, with a replica of the exhibition space, VR presentations and 3D product models, so customers understand complex products faster. Its job is sales, so be clear which one you are buying.
CAD is the authority on geometry, while a twin is the authority on behavior over time. You need both, as we explain in Digital Twin vs CAD vs Simulation Software.
Monitoring answers a different question. A dashboard tells you the state of a machine right now. A twin tells you what would happen to the whole line if you added a buffer in front of it.
A twin also comes with no guarantee: it is as good as its inputs and the people who own it and update the base model whenever the plant changes. Most failed twin projects fail for organizational reasons, covered in Why Most Digital Twin Projects Fail, and How to Avoid It.
When a digital twin is worth building
A twin pays off when three conditions meet:
- The change is expensive or hard to reverse: a new hall, a layout change, a robot cell, a capacity investment.
- There are several plausible options, and the discussion runs on opinion because nobody can show numbers.
- Things interact: shared forklifts, buffers, shifts and product mix make the result hard to predict in a spreadsheet.
If the change is small, reversible and limited by one visible constraint, try it on the floor. A model would cost more than the mistake.
DBR77 Digital Twin models the material flow from your layout and process data, compares variants on the same measures, shows results in 3D and VR for review meetings, and starts from static data.
FAQ
Is a digital twin the same as a simulation?
They are close. A simulation is a run of a model. A digital twin is a base model of one specific plant as it runs today. Each new decision is tested as variants on a copy of it, so the base does not change while you test. Once a change is made on the floor, the base model is updated to match and becomes the starting point for the next decision.
Do we need live data from machines to start?
No. Most first models use ERP data, routings and measured cycle times. Live data is the next stage, once the model is used often.
What does a first model need to start?
Nominal times are enough to start. Enter routings, stations and times by hand, or bring them in by import (import scope to be confirmed). The scope is set by one decision and one area, and the model can grow with the next decision.
Is a digital twin only for large or new plants?
No. The DBR77 case studies come from existing mid-size plants, for example a car lamp maker with 270 people and an HVAC manufacturer with over 150. Brownfield plants often gain the most, because every change happens around running production.
Conclusion
Before you call anything a twin, check that you can change one input and get a new output figure back. For a first project, pick the next decision that will commit real money, such as a layout change or a robot cell, and list the variants on the table. That list, plus your routings and order mix, is enough to brief a first model.
Test your next change before you make it
The demo uses a sample plant to show where parts queue and how layout variants differ in daily output. Bring the layout change you are weighing, and we will talk through how it could be modeled.
Sources
- McKinsey & Company, What is digital-twin technology?
- ISO, ISO 23247-1:2021 Digital twin framework for manufacturing, Part 1: Overview and general principles
- DBR77, Case study: Production line design (summary on this site)
- DBR77, Case study: Showroom (summary on this site)
- DBR77 Digital Twin, Averages vs simulation over time
Want to test a decision from your plant?
Describe the decision to a DBR77 expert and we will show how DBR77 Digital Twin compares the options on output, waiting time and transport load.
Questions
Have a question about this topic?
Tell us what it is about. We will reply within one business day.
Digital Twin newsletter
Short notes on testing layout and flow decisions on a model, and news from DBR77 Digital Twin. You can unsubscribe at any time.