At what project size does a digital twin pay off?

“That must be for big projects.” It's the first thing mid-sized developers say when we bring up digital twins. It's understandable, but it's backwards: what changes with size isn't whether it's worth it, but what for.

Want to see this on your project? Book a free demo

First, why the big ones do need it

McKinsey reviewed more than 300 projects with contracts above one billion dollars and found average cost overruns near 80% and delays around 50%. Broadly, large projects take about 20% longer than scheduled.

The reason isn't technical; it's coordination. Many actors, many interfaces, and nowhere for everyone to see the same thing at the same time.

What the model contributes at each scale

Small project — one tower, under a hundred units

Here the risk is commercial, not constructive. Every month of absorption weighs on a tight budget, and there's no room for a model apartment that gets built early and sold last.

The model earns its keep on sales: it replaces the physical unit, travels over WhatsApp and reaches the buyer living in another city.

Mid-sized project — several towers or phases

Here the risk of repetition appears. A detail poorly resolved in phase 1 gets built identically in 2, in 3 and in 4. The error doesn't add up: it multiplies.

And because the unit types repeat, the model pays for itself in the first phase and gets reused in the following ones with minor changes. It's where the investment performs best.

Large scale — mixed use, megaprojects, public works

Here the risk is pure coordination. Several designers, several contractors, chained schedules and decisions that depend on one another.

The Construction Industry Institute found that design changes, errors and omissions explain 79% of deviation costs — 9.5% of total project cost. On a job with twenty interfaces between disciplines, that number stops being a statistic and becomes the schedule.

The real problem: what nobody accounted for

No project blows its budget because of what was planned. It blows it on what wasn't accounted for: the duct nobody coordinated, the headroom nobody verified, the finish nobody defined until the invoice arrived.

Autodesk and FMI quantified that gap. Decisions made on bad data cost the industry roughly US$1.85 trillion in 2020 and explained 14% of all rework that year. “Bad data” means exactly that: information that is incomplete, inaccessible, inconsistent or late.

A model doesn't guess what's missing. It makes it visible, which is all you need to be able to ask about it in time.

How to start without overbuilding

You don't have to model the entire project in detail from day one. The order that works best is this:

• The commercial model first. Facade, common areas and the unit types that carry the most sales weight. It pays for itself through the sales cycle.
• Then the technical depth on the same file: MEP, structure and the critical coordination points.
• And finally, keep it alive. Update it with the real changes so it works as a reference at handover and in after-sales.

That's the logic behind every digital twin we build: start where the investment comes back and grow from there. To see how it would apply to your case, take a look at the full range of solutions or get in touch.

Sources

McKinsey — Imagining construction’s digital future · Construction Industry Institute — Costs of Quality Deviations · Autodesk + FMI — Harnessing the Data Advantage in Construction