Do You Need a 'Digital Twin'?

"Digital twins" can allow companies to simulate strategies before implementing them -- but can be misleading if not monitored carefully.


It's become fashionable to talk about how companies need to build a "digital twin" -- essentially, an incredibly detailed digital model of the business so they can simulate a range of possible moves and see the results before deciding what to implement in the physical world.

Should you go along with the fashion?

The simple answer is: Yes. And no.

Accenture recently made the case for using digital twins in insurance. The blog notes that digital twins are being deployed effectively in many industries:

"Outside of the insurance realm, digital twins are being linked together to create living models of whole factories, product lifecycles, supply chains, ports and cities. Companies are using them to understand supply chain predictability, worker safety, maintenance and repair costs, and as a risk-free playground for innovation. For example, Unilever is working with Microsoft to develop intelligent twins of its factories so it can test potential operational changes and improve production efficiency and flexibility."

IBM makes a compelling argument about, for instance, outfitting a wind turbine with sensors producing data about key aspects of the physical object’s performance, such as energy output, temperature and weather conditions. The data can then be relayed to a processing system and applied to the digital copy. 

"Once informed with such data," IBM writes, "the virtual model can be used to run simulations, study performance issues and generate possible improvements, all with the goal of generating valuable insights — which can then be applied back to the original physical object."

Kevin Kelly, a co-founder of Wired, paints an even grander version, as usual. In early 2019, he laid out an almost poetic vision of what he calls a "mirrorworld," which is based on an exact, digital representation of everything in the real world.

"The mirrorworld doesn’t yet fully exist," he writes, "but it is coming. Someday soon, every place and thing in the real world—every street, lamppost, building and room—will have its full-size digital twin in the mirrorworld." 

All those possibilities sound great, right? So, what's the problem with digital twins?

The problem is that no model is a perfect representation of its physical counterpart. It's easy to think otherwise, especially once you've become accustomed to using a model for a time, and confusing a model with reality can be disastrous.

Look at Zillow, which developed a sort of digital twin of the housing market and which bought billions of dollars of houses, expecting to be able to flip them quickly -- only to find that its model didn't quite match reality. Zillow lost $380 million in its latest quarter and said it will take a writedown of half a billion dollars on its remaining inventory of homes. The Wall Street Journal says, "Zillow ran into some of the limits of technology in a business still informed by emotional attachments, personal tastes and other intangible factors."

Or, look at the models that led to the Great Recession in 2007-09. Financial services giants, including AIG, created derivatives based on incredibly precise models -- that ignored the possibility that housing prices could drop. Long-Term Capital Management likewise relied on incredibly elaborate models of financial markets -- and needed a $3.6 billion bailout in 1998.

A friend and colleague, Vince Barabba, taught me long ago: "Never say, 'The model says.'"

A model is simply not adequate justification for any decision that matters. You have to always be able to justify a claim or a decision based on actual evidence and logic, not just on a model that was likely developed long ago, based on assumptions that have become obscured.

Vince's track record gives him plenty of credibility on models. He held any number of senior corporate positions, including as SVP of strategy at General Motors, where he gave the world OnStar, and was twice the director of the Census Bureau. He has written numerous books on strategic decision-making.

I've also seen up close and personal, based on some consulting work we've done together, how he uses models but doesn't entirely trust them. A key tool is what he calls "decision records." Any time you are making an important decision, including those that go into elaborate models like digital twins, you record the assumptions you're making. You then revisit those assumptions from time to time to see how they've changed and to see if you need to adjust or even throw out your decision -- as Long-Term Capital Management, AIG, Zillow and many others should have done.

My recommendation on digital twins: Be like Vince.

Take advantage of the increased digitization of the world to build the best models you can and use them to simulate decisions as much as possible. But don't trust them too far and regularly revisit the assumptions that went into building them.



Paul Carroll

Profile picture for user PaulCarroll

Paul Carroll

Paul Carroll is the editor-in-chief of Insurance Thought Leadership.

He is also co-author of A Brief History of a Perfect Future: Inventing the Future We Can Proudly Leave Our Kids by 2050 and Billion Dollar Lessons: What You Can Learn From the Most Inexcusable Business Failures of the Last 25 Years and the author of a best-seller on IBM, published in 1993.

Carroll spent 17 years at the Wall Street Journal as an editor and reporter; he was nominated twice for the Pulitzer Prize. He later was a finalist for a National Magazine Award.


Read More