Decision tools for complex business models.
Forecasting and predictive tools are continually improving. This progress has accelerated since the widespread adoption of machine learning and statistical techniques. The most significant gains have emerged in near-term forecasts, while strategic prediction capabilities have only begun to attain velocity. Among these, business simulations and Digital Twins of Organization (DTO) are emerging methods to understand the evolution of complex systems and predict outcomes over one to five-year horizons.
DTOs encode supply chains, distribution networks, construction projects, and other complex business systems. They model non-deterministic, at times irrational, behaviors with behavioral economics, psychology, mathematics, and a dose of physics.
These strategic decision tools are of paramount importance. The growing complexity of business models handicaps the evaluation of which investments deliver decent long-term returns. It obscures the danger of decisions that send a company down the wrong track and often result in astronomical costs. DTOs lift the veil of uncertainty.
They do so with business simulations, the mathematical techniques that evolve the business over one to five years. Starting from the current state described by the DTO, business simulations predict a future outcome. When computed tens of thousands of times with slight variations of inputs, the simulations paint a picture of what is likely to happen.
So companies can test new scenarios and watch how the future plays out. The concept is a bit esoteric, but an example will help clarify the method. I promise it is all practical. But first, let’s answer a challenge.
If DTOs and business simulations are there to predict the future, doesn’t machine learning already accomplish this? And since machine learning is so successful, why complicate matters with yet another thing?
Machine learning is excellent. It is good at finding patterns where people cannot see them. It is good at predicting what will happen next: what people will buy, which machine will break down on the assembly line, and if someone is stealing your money.
But machine learning needs a lot of data. It operates on the idea that the future resembles the past. It is good at predicting in the short term — the next minute, the next day, maybe the next few months.
But what if your data is poor or there is not enough? Or what if you want to know what happens in five years? What if you want to explore black swan scenarios — markets crashing, your key supplier suddenly going out of business, or new tariffs wreaking havoc with your cost structures? The standard tools of data science may falter, but business simulations excel. Now an example.
Imagine you are a large tier-one automotive supplier. You build primary components for vehicles from smaller parts and raw materials. The pandemic rocked your worldwide supply chain, and you are considering a change.
You want to evaluate re-shoring a portion of production and shortening supply chains. Which investments will the effort require? How long will the transition to the new model take? Will the work yield robust results and operations? Or will the changes you implement break the flow of goods to your customers, or perhaps result in prohibitively expensive parts? Unwinding ill-advised changes will cost money, time, and lost revenue.
So, you choose to mitigate the risk by simulating the impact of the proposed changes on your business.
Firstly, you use existing platforms (there are a couple) to build a digital twin of your supply chain. The digital twin includes factories, warehouses, mines, transportation networks of all freight modalities, labor pools, regulatory nodes, customs, etc. The model is comprehensive, terrifyingly so, but tractable with modern computational power.
Secondly, you calibrate your DTO using existing data and machine learning techniques so that it matches what is happening in your business right now. If the model can predict what happens today, it can forecast what may happen in the future.
Thirdly, you run simulations thousands of times. Why so many? Simulations are stochastic, which, in our case, means that given the same initial conditions, their outcomes will vary. Upon completion, the predictions will settle on the most likely range of results. That’s the law of large numbers for you.
Finally, you can ask the model questions. Say you restructure your supply chain: move a factory from China to the US, cut out ocean shipping, change labor practices to match the US laws, and introduce a volatile trade tariff policy on raw materials. Repeat the simulations. You may see improvements in inventory availability but higher costs of production. You may identify a risk to raw material costs from higher tariffs but a reduced likelihood of shipment delays at ports. If your objective is supply chain robustness, then the new design might be a win.
Posing endless questions to the model can take time and effort. But allow the simulation freedom to explore a wide range of scenarios. Then group the results into categories relevant to your business’s priorities. You will be surprised by what you discover.
In the end, the DTOs inform perspective. The decisions are still yours. But with added clarity, decision-making carries less risk. Simulations function as radar in the fog of future uncertainty — you begin to see the obstacles that will sink the ship and have the consequent capability to change course.