One of the biggest trends to emerge in the industry 4.0 and IoT space is around optimization. So, what exactly is “The Digital Twin”, how is it different than forecasts and simulations, and more importantly, what happens to companies that get left behind?

Here’s what I believe:

Digital Twins are a concept, not a single product or a piece of technology. Multiple technologies—2D/3D simulation, IoT, 4G/5G, big data, blockchain, edge computing, hardware, cloud computing and AI/M—might all come together to make the concept a reality. However, I also do not believe that there is a requirement for a physical entity or an asset. While it may have hardware, I believe human decisioning can also have a digital equivalent; healthcare decisioning is a great example of this.

Forecasting vs Simulations vs Digital Twins

Data comes with challenges, and analytics leaders are constantly looking to evaluate and obtain the right tools, techniques, and platforms that will give them an advantage in business. One question frequently asked is, “Are Digital Twins and simulations and forecasting all the same?” While similar, these are actually three different problem-solving techniques. Below, I’ll go over them and explain what they’re best suited for.

Forecasting & Predictive Modeling

Forecasting is a process of predicting or estimating future events based on past and present data by carrying out an analysis of trends. Gut feel decision-making doesn’t cut it. For example, we could use forecasting to predict how many prospect calls Ryan (one of our sales executives), is likely to receive the next day. Or how many meetings he’ll lead over the next week. The data from previous years is already available in our CRM, and it can help us accurately predict and anticipate future sales and client opportunities where Ryan may be needed, or whether we need to hire more sales execs. A forecast, unlike a prediction, must have logic to it. It must be defendable, but is based on historical data (# of calls & meetings per sales exec) and new trends (new product launch or market changes).
Predictive modeling, on the other hand, is a form of artificial intelligence (AI) that uses data mining and probability to forecast or estimate more granular, specific outcomes. For example, predictive modeling could help identify actual customers who are likely to purchase supply chain optimization software over the next 90 days. To do so, we could indicate a desired outcome (a purchase of our analytics software solution) and work backwards to identify traits in customer data that have previously indicated they are ready to make a purchase soon. For example, they might be the executive sponsor, have an established budget for the project, come to one of our events, and found Ryan likeable and helpful. Predictive modeling would run the data and establish which of these prospective customers will close the sale.


After a forecast has been generated, analysts then have the ability to simulate changes to key metrics–we call these “what if” simulations. Editing cells within the software allows you to visualize periodic changes to your data. The effect of each change is reflected across all dependent metrics and global KPIs. As it relates to manufacturing optimization, it’s often used with physical elements such as assembly lines, robotics or ovens. One of the most challenging decisions facing supply chain organizations is deciding on the right product mix. While some products may offer high profits but aren’t sold very often and/or take up a lot of space in the warehouse, others have a much greater turnover. Some products may align with the organization’s long-term strategy, while others don’t but are profitable. In most instances, these decisions are extremely complex, and it’s difficult to evaluate how they will affect overall revenue to profitability. A simulation can provide the right answers based on exploring several “what if” scenarios. A good example of this is Nathalie, once of our consultants, creates “what if” inventory scenarios for our retail clients. We use this to determine the cause & effect of changes in demand or supply planning to determine how SKU changes could affect the sales/basket analysis.

Digital Twins

Digital Twins help us to understand the present and predict the future. Simulation is an important and an integral part of the Digital Twins, but the purpose of the Digital Twins goes beyond simulations; it’s a ubiquitous term used for a wide range of things across a wide range of applications, from high value manufacturing and personalised medicines, to oil refinery management, and risk identification and mitigation for city planning. Within engineering, claims are made regarding the benefits of using Digital Twinning for design, optimisation, process control, virtual testing, predictive maintenance, and lifetime estimation.  By type, the market is segmented into parts twin, process twin, system twin, and product twin.
For some of the people, the reason why “Twin” is used has been forgotten, so let’s look at the three key differences:
• The Past:  Simulations are typically used for design, and in certain cases, offline optimization. Digital Twins, on the contrary, are used for the entire design-execute-change-decommission lifecycle in real-time.
• The Present: Simulations, at best, can help understand what may happen in the real world. Digital Twins not only help understand what may happen, but crucially what is happening now (how the design is behaving in the real-world).
• The Future:  Digital Twins allow us to visualize models before we manufacture a product or execute a process. Examples are gas distribution networks, new airplane models or even COVID social distancing applications.

According to Gartner, 13% of organizations implementing IoT projects already use Digital Twins, while 62% are either in the process of establishing a Digital Twin use or plan to do so. Tesla has a Digital Twin for every car manufactured. Every day, thousands of miles of data from the cars are fed into the simulation models back in their factory. That’s a thousand and more ways to learn and optimize from the real-world which wouldn’t be possible with simulation models alone. In our autonomous future, the vehicle may drive itself into a garage for a check-up or when it feels under the weather, without needing the owner behind the wheel.

Digital Twins are best when an outcome is changing over time, therefore the initial model becomes invalid, and when the data that can be correlated with this change can be captured/traced. For instance, these changes could be undesirable, such as fatigue in metal components like bearings or variations in supplied material qualities or they can be desirable, such as improvements in bottlenecks. If the data does not change much over time, or if data associated with that change cannot be captured, then a Digital Twin isn’t likely to be useful. Using a Digital Twin means more effective R&D, greater efficiencies and a more comprehensive approach to product end-of-life.

Beyond manufacturing, the Digital Twins offers significant benefits in healthcare. Think of the ability to tailor drug characteristics or dosage for personalized medicine or even adapt radiation therapy to account for unique internal anatomical changes and patient response. Other healthcare examples could be monitoring prosthetics to detect damage and wear in order to improve prosthetic implant designs (e.g. knee replacements). Creating a Digital Twin will lead to more efficient and effective treatment with fewer side effects, and provide improved health outcomes.

Other Examples:

Digital Twins help supply chain business users and data engineers accomplish a great deal, like:
• Testing supply chain design changes and development
• Monitoring risk and testing contingencies
• Transportation planning
• Inventory optimization
• Cash-to-serve and cost-to-serve analysis
• Forecasting and testing operations over the coming days and weeks
• Visualizing products in use, by real users, in real-time
• Building a digital thread to connect disparate systems and promoting traceability
• Troubleshooting equipment that is far away
• Real-time virtual models of physical assets help avoid bottlenecks
• Improving warehouse layouts and increase productivity
• Analyzing product and packaging data to safeguard shipments
• Understanding product performance

What to consider when determining if a Digital Twin is right for your organization

The most likely applications of Digital Twins are all either high-value or safety-critical. Beyond the hype of Industry 4.0 and Digital Twins, it is important that companies accurately assess their maturity-readiness and security before undertaking this type of complex initiative.

The best way to embark on a Digital Twins initiative is to identify the assets and processes with the highest potential for value creation, and then begin a pilot implementation. Organizations usually start their Digital Twin journey with simulations of critical assets. Such simulations are good for playing out the what-if scenario for the asset. A Digital Twin should be a work-in-progress that continuously evolves and scales – as your IT capacity expands and matures. The Digital Twin of a physical object is dependent on the digital thread lowest level design and specification for a Digital Twin—and the “Twin” is dependent on the digital thread to maintain accuracy.

A single Digital Twin project can later be interconnected to form a large, series of Twins with a highly complex machine or process. By continuously reviewing aggregate analysis (Digital Twin Aggregates or DTAs) across all real-time Digital Twin instances will provide a better idea where the most tangible benefits can be realized.

Creating your first Digital Twin usually requires different elements, including: • Data sources such as IoT Sensors capturing operational behaviors of assets and processes (vibration, temperature, pressure, etc.), alongside their functioning environments (air temperature, humidity, etc.).
• Data networks providing secure and reliable data transfer from physical devices to the digital world.
• A modern platform that serves as a modern data repository pooling and storing shop floor sensor data with high-level business data (e.g. MES, ERP).
• Software Application(s) By combining these data sources into a software solution that uses advanced AI/machine learning algorithms, we can create actionable insights for data-driven decision-making.

Challenges & Mitigating Risks

Dirty Data

The low cost of IoT sensors and the easy access to cloud storage has led to collecting extremely large datasets. This frequently consists of data from multiple sensors of varying types gathered at short time intervals. The challenge with using such data sets in Digital Twins is to identify which measurements at which locations or times have the most effect on the parameters to be updated within the Twin. It is therefore very important to be able to trust the predictions of the Digital Twin, which means that you must also trust the data, the models, and strong governance.

Lack of buy-in

Combating nay-sayers who are comfortable with status quo works better when you start with a strong business case (“let’s experiment” just won’t cut it). Ensure you are factoring in all the tangible benefits related to profitability, improvements in process, reductions in manpower, reductions in mistakes/gut feel decision making etc. Speaking of nay-sayers, don’t forget to create champions of the users/recipients of the Digital Twin. Users of Digital Twin Digital Twin technology must adopt new ways of working, which can potentially lead to problems in building new technical capabilities. Education and involvement will help them embrace this new change before it’s rolled out (or it won’t make it past the pilot stage without a lot of barriers).

Lack of funding

Innovation is usually expensive and there are definitely challenges in obtaining budget for something that has an unknown outcome before it’s started. That said, doing your homework and calculating the ROI goes a long way. Calculate the ROI before/after and don’t forget to include the calculation for Opportunity Loss. To augment your budget and increase your ROI, there are many non-repayable grants available and tax credits available for you here in Canada. Things to consider: their financial contribution can represent at least 50-75% of Total Project Costs, (as calculated by Salary Costs + Expenses + Contractor Costs) and coverage will fluctuate based on grant organization. In-kind contribution is not accepted in this calculation and grant organizations will not reimburse any costs incurred prior to the start date of an approved contribution agreement. SR&ED tax credit refunds can cover anywhere from 40% to 100% of your R&D. If you have questions about your options for non-repayable funding or tax credits, please reach out to me.

Cutting-edge innovation has big benefits but also has big risks. Reach out to those that are blazing a trail for support before you embark on a complex solution like Digital Twins.