Why Digital Twins are set to transform products and businesses
2016 has been pivotal year for IoT. Most manufacturers of physical products are actively evaluating the option of injecting their products, factories and value chains with sensor-enabled intelligence. As products generate/collect data, it could be used to model a cloud-based software doppelganger of the physical product. This is the concept behind the Digital Twin, virtual twin, device shadow, or whatever name the industry eventually settles on. As the product moves from design to factory to the field, tracking and capturing its entire lifecycle, and piping that data back to the design stage is called the Digital Thread. In other words, Digital Thread is the end-to-end lifecycle data and analytics on the product. It provides the linkage between the manufactured physical product and its Digital Twin.
The term Digital Twin was coined by Dr. Michael Grieves. In his vision, he presented how these virtual representations could form tighter loop between design and execution.
By mid 2016, Digital Twins have become an important aspect of smart product discussions, and rightfully so. If we were to pause and think about the lifecycle of a traditional product — be it a toy or a motorbike or a pump — once the product is sold, it is usually very difficult for the manufacturer to get information about the product usage and performance in the field, and also to know how the product functional characteristics morph over time. Companies typically try to scrape this information through the customer service channels or by conducting surveys or through user interviews. None of these approaches are truly effective for the product makers to get the complete, comprehensive picture. Second challenge for product makers is that we are getting into an age of customer experience. Companies like Nike and Starbucks have shown how the product is a gateway for creating great experiences for the customer and to earn their loyalty. Clay Christiansen characterized this as a job-to-be-done by the product for the customer in his milk-shake marketing paper. With reliable data, arriving at the right customer experience design and product-market fit could be much easier for the manufacturer.
With a Digital Twin, each physical product will have its own software version of itself (the twin). If the product is instrumented right, the data collected over time could give very precise information about its usage, performance, and other characteristics. Product Designers can now have accurate statistics of how each component or sub-system behaves. Most importantly Digital Twins (along with Digital Threads) facilitate continuous learning, the ability to dynamically recalibrate, and improve end-to-end traceability.
As of this writing, there is no consensus yet in the industry as to what features Digital Twins and Digital Threads should have, how they should function, or what paradigms should they be supporting. In reality, a Digital Twin might end up being the overarching concept — just like Client-Server, Service Oriented Architecture and Cloud computing are. Each vendor will have their interpretation of the Digital Twin concept and implement it with features and capabilities that best fits their customers’ needs and their business models. For example, vendors like PTC and Siemens have been experimenting with the use of AR to augment their PLM platform and IoT management cloud. Vendors like GE, AWS and Oracle focus more on the operational data and its virtual representation in the cloud.
Regardless of the vendor philosophy and implementation, the core capabilities of Digital Twins might be projected into 5 facets — service, design & engineering, manufacturing, usage, and risk. These facets are not necessarily mutually exclusive.
Now that we have this background and understanding, let’s dive in and explore each of the view areas in more detail.
Product Design — As each physical product out in the field relays data, analysis from this data can be used to gain insights into various affordances of the product. This data is invaluable in designing various aspects of the product, including usability, materials, performance, serviceability, backward compatibility, and more. Armed with this info, designers can mitigate/avoid future problems and deliver more value for customers in the right areas.
By projecting the inferences from the data into the CAD models, the delta between as-designed and as-performed can be reduced considerably. Secondly, with reliable data at hand, designers will be able to bring more rigorous A/B testing, what-if analysis and simulations. Assumptions made during design can be more empirical and drawn from the real world — versus the current approach of theoretical modeling.
Digital Twins for product design have been in use for almost a decade now for highly capital intensive products such as jet engines and heavy machinery. However, as Digital Twins get more ubiquitous, democratized, and accessible, their benefits can be leveraged by every product manufacturer and product user (for self-service).
Product Manufacturing — Despite the best design and engineering, product defects could be introduced during the manufacturing or in the supply chain (especially with OEM components). Digital twins help identify the root cause of pesky production problems — such as the machine, assembly line, factory, supplier from where the component came from. By correlating smart factory data with Digital twins, defects from manufacturing stage can be alleviated quickly.
Having said that, getting to this level of insight is more complex from an implementation point. The kind of problems that we just talked about are usually one-off, and don’t subscribe to a pattern where supervised machine learning techniques can easily be applied. Never the less, the available data can help with faster troubleshooting of the root cause than before. Many times, simple instrumentation and basic analytics go a long way in delivering value.
Product Service — Servicing is a critical, yet expensive proposition for most product manufacturers. Digital Twins could help with product services in two key areas: (a) Manufacturers will be able to predict when a part or component is likely to fail, and proactively initiate a service call. This prevents costly unplanned down times for their customers. (b) With the power of technologies like AR, technician training and performing repairs onsite would be much more efficient as they have clear 3D virtual models and step by step instructions to quickly and efficiently perform the repair. When repairs are done incorrectly, it causes additional unplanned downtime. With Digital Twins, all these issues could be mitigated significantly.
Product Usage — Many products (such as pumps) are used in multiple verticals, and in diverse environments. For example a pump might be pumping oil in frigid Alaska, while the same make and model of the pump might be pumping water in arid part of Australia in dusty conditions. These operating conditions lead to very different degradation factors over time. Secondly, as it happens often, customers use products in ingenious ways that the product designer has not intended or designed the product for. By understanding the usage characteristics, conditions, workloads, degradation, unit economics and other factors, product manufacturers can have accurate insights to make actionable decisions. They may choose to change the design of certain components, tailor the product offerings for different markets, or innovate on the business model and pricing based on consumption.
Product Risk — Recalls are cost prohibitive for any manufacturer. Until now, risk management has been a black art drawn from experience and history. As we have seen with Takata airbags and other auto recalls, many complex calculations play into the recall decision. With Digital Twins, for the first time, manufacturers have much more reliable ways of isolating and acting on the risk parameters. With quick root cause insight and proactive customer support, costly and embarrassing situations can be avoided.
In the above 5 facets, we discussed the direct value that can be extracted with the adoption of Digital Twins. While this in itself is very beneficial and has substantial ROI, there are many secondary (or derivative) benefits that companies will harness. Unfortunately some of these ‘benefits’ could be at the expense of customer privacy and ethics.
Insights from data will be leveraged to innovate on exciting business models. As more granular information is available, new economic models can be applied. We are seeing examples of this in the world of Cloud Computing. For example, just about a decade ago, companies were happy to have 80% utilization of their servers consistently. Today, serverless cloud services are billed for sub-second usage. We can see similar unit economics coming into play as we move from capital expense models to consumption driven models. For self-driving cars, it is predicted that we will move from person-to-car, to group-to-car, to car-to-car ownership models. As we learn and mature in harnessing consumption driven models, it is fair to assume that these economics and ownership models will be applied to many physical goods.
Usage data collected from smart products is very useful for retailers, advertisers, marketers, product planners, financial analysts, government agencies. Clean and validated data is monetizable asset. There are startups emerging that treat IoT data as a commodity that can be traded — just like gold, copper and oil. It is safe to assume that companies would like to leverage these data markets for driving additional revenue sources. Data trade is also a gift that keeps on giving to the data collectors. Same information can be repackaged and sold in innovative ways. There is gold in the digital exhaust.
Challenges and Path Ahead
While there are many benefits to Digital Twins across the product lifecycle and value chain, like most major initiatives, it is not without its challenges. First, Digital technologies work best when there is deep vertical integration across the value chain. Making these changes (culture, process and systems) is much easier said than done. Secondly, it is easier to have digital technologies embedded into new factories and buildings than to retrofit existing infrastructure. Third, the sheer complexity and the diverse of possible use cases that come out of massive data collection is a big challenge. It takes gradual adoption and baby steps to gain the value from these technologies.
It is fair to say that we will see tactical adoption in areas where there is unique, distinctive advantage without disrupting the current value chain. Therefore areas like Service and Design may be the early adopters of Digital Twins versus the other areas.
If you like what you read, let me know by clicking the ❤ icon, or by leaving a comment. It also helps others find this blog post!
Find me on Twitter: @akadiyala