How to Build a Digital Twin Business Case

Put a price on getting it wrong

The value of a simulation is the cost of the mistakes it helps you avoid, plus the better option it helps you find. Start with the mistakes, because your plant already knows their price from earlier projects.

Typical costs of a wrong layout or capacity decision:

• moving machines twice, with rigging, installation, electricians and lost production for the second move
• a slow ramp-up, with weeks of lower output while the team finds where parts wait
• buffer stock that grows to hide a flow problem and ties up working capital in racks
• overtime and extra shifts to make up for a line that does not reach its rate
• an automation investment that waits for parts because the flow around it was not checked

Look at your last two or three changes. What did the rework, overtime and lost output cost? Those are real numbers from your own plant, and finance can check them.

What goes into the case, and how to label each line

Each line of the case needs a source for its number and a label that tells finance how far to trust it.

The fifth line is where most cases go wrong. Before the study, nobody knows how much better the best option is. Writing a fixed percentage of extra output into the case at this stage means using a made-up number, and a careful CFO will spot it.

Use a break-even threshold instead. Suppose, for illustration, that a line earns a margin of 20,000 per working day and the plant runs 250 days a year. If the best layout adds just 1% of output, that is 200 per day, or 50,000 per year. Compare that with the cost of the study. With your own figures in place of these illustrative ones, the question stays the same: how small a gain pays for the work? If the answer is "less than anyone in the room expects", the case is easy. If the gain would have to be large, either pick a bigger decision or reduce the scope.

Real case: what a layout project returned

A medium-sized manufacturer of high-quality wooden furniture, in business for more than 20 years, found that its production and warehouse layout no longer fit a changing production environment. A wide product range slowed the flow of material and information. A lack of buffering for materials and finished goods reduced flexibility and caused delays.

DBR77 built a digital twin of the current production and warehouse layout and simulated the material flow. AI identified the key areas and generated layout proposals. Specialists selected the best option, and the chosen layout was simulated. The results:

• Working capital reduced by PLN 300,000 after the buffer was eliminated.
• Departmental costs reduced by 14%.
• Direct labor demand reduced by 1.5 full-time positions.
• Mapping the production area took several days. Further AI-driven optimization took seconds instead of days.

These numbers belong to one plant. They show which lines of a business case a layout project can move: working capital, departmental cost and labor. Use them to choose which lines to measure in your case, and take the figures themselves from your own plant.

Show the downside next to the upside

An approval committee trusts a case that shows what could go wrong. Add three things:

• What the result depends on. If the best option wins only when cycle times stay at their best, say so. We cover how to test this in What a Sensitivity Analysis Should Show Before Approval.
• How the options compare under the same disruptions, such as a machine stop, a late delivery or a demand peak. See How to Compare CAPEX Options When Every Scenario Looks Plausible.
• What the study will not answer. Scope limits stated in advance prevent disappointment later.

Results from a stochastic simulation come as ranges, which suit a case that is open about its downside.

Keep the first case narrow

The first case does not have to justify a plant-wide program. One decision, one area, a model built from data you already have: CAD layout, routings, cycle times and the order mix. Live machine data from DBR77 IoT can be connected later, when the model is used often.

The model then stays. The second decision, whether a robot, a new product or another relocation, starts from a model that already matches the plant, so it costs less and takes less time. Mention that in the case, but do not count it as a saving yet. When the result is ready, present it as described in How to Present Simulation Results to the Board.

FAQ

What does a digital twin project cost?

It depends on the scope: the size of the area, the number of variants and the data available. Ask for a quote for one specific decision. That number goes into the case as a fact.

Should we put an expected saving into the business case?

Not as a single number before the study. Use a break-even threshold, and add the measured result after the simulation. Past project costs from your own plant can go in as estimates with a range.

How quickly does a first study deliver a result?

In the furniture case, mapping the production area took several days, and further AI-driven optimization took seconds. Your timing depends on scope and data.

Who should own the business case, finance or operations?

Both. Operations names the decision and the options. Finance checks the numbers and the labels. A case signed by both is much harder to reject.

Conclusion

Before you write the case, collect what your last two or three layout or capacity changes cost in rework, overtime and lost output, and agree with the controller which of those figures finance accepts. Then work out the break-even gain for the decision you want to test. If that gain is smaller than anyone in the room would guess, the case is ready to write.

Sources

• DBR77, Case study: Layout optimization
• DBR77 Digital Twin, Deterministic vs stochastic simulation