Managing under uncertainty is not easy, especially when you’re facing new realities that have never been experienced before, such as rising energy prices, trade disruptions, a pandemic, and inflation.
However, there’s a solution that at its core has been built to help with causality and systems’ intrinsic rules. This approach is simulation. With it, you can make a forecast for unforeseen events, for example, large-scale disruptions due to pandemics or geopolitical shifts.
Read more: How to set randomness in your simulation model.
Four levels of uncertainty
Many managers are bound to make decisions under uncertainty – acting based on often imperfect observations and with unknown outcomes. But uncertainty itself can have different levels. Identifying the right level of uncertainty can help managers and consultants develop actionable strategies that protect a company against threats.
The uncertainty that remains after the best possible analysis undertaken falls into one of the four categories below.
Level 1. A clear enough future
At this level, the environment is so stable and slow-changing that a simple forecast of the future could be precise enough for strategy development.
Level 2. Alternative futures
The future is one of a few alternative discrete scenarios, but you’re unsure which of them will eventually happen. Many businesses facing major regulatory or legislative change confront this level of uncertainty.
With a simulation model, managers can run various what-if scenarios to test and analyze how the modeled system would perform and assess possible risks.
Level 3. A range of futures
To identify a range of potential futures, there are a limited number of key variables, and the outcome may lie somewhere within that range.
For instance, a café owner knows from their observations that the first guests usually arrive at any time from 8:30 to 10 am and there could be any number of customers from 1 to 5 entering at the same time.
When modeling a café, a simulation engineer would need to take these ranges of variables into account.
Level 4. True ambiguity
Multiple dimensions of uncertainty interact to create an environment that is virtually impossible to predict. In contrast to level 3 scenarios, it’s impossible to identify a range of potential outcomes, let alone scenarios within a range. Situations that fall into this category are rare and they tend to migrate to levels 2 or 3 over time.
Nevertheless, they do exist. McKinsey gives an example of a telecommunications company deciding where and how to compete in the emerging consumer multimedia market. The company will confront several uncertainties concerning technology, demand, and relations between hardware and content providers. These uncertainties may interact in ways so unpredictable that no plausible range of scenarios can be found.
Most real-life scenarios that are dynamic in nature usually belong to level 2 or 3 and can be addressed with simulation modeling.
Simulation-based experiments that help with uncertainty
With simulation modeling you can handle time and causal dependencies, explaining why things happen. This helps see the impact of business decisions before implementing them in the real world.
The case of alternative futures
For various levels of uncertainty, there are experiments available in AnyLogic. For situations falling into the level 2 category, there are sensitivity analysis, parameter variation, and compare runs experiments.
Sensitivity analysis runs a simulation model multiple times varying one of the parameters and shows how the simulation output depends on it.
Parameter variation experiment runs a model with different parameters and analyzes how they affect the model behavior. The variation is performed automatically and includes several single model runs. The results of these runs could be displayed on one diagram illustrating different behavior with a certain subset of parameter values.
Compare runs experiment is similar to parameter variation but instead of input values automatically changing according to a predefined algorithm, a user can interactively control the inputs and ultimately compare results.
Monte Carlo for a range of futures
For level 3 situations, AnyLogic offers a Monte Carlo simulation experiment. The experiment shows results for running a stochastic model or a model with stochastically varied inputs sampled from certain distributions.
Monte Carlo simulations need to iterate many times to produce useful results and consequently benefit from fast computer processing. When models are very complex and dynamic, processing requirements can become significant and run times very long. This reduces possibilities for what-if experimentation and may limit the usefulness of a model in decision-making.
Experiments in Cloud
Simulation allows you to experiment with a virtual model of a real-world system. To provide you with representative results, experiments require multiple simulation runs and processing power. With AnyLogic simulations connected to Cloud, you can quickly run experiments on any computer and easily share the results with your colleagues.
Conclusion and case studies
Many companies around the world use simulation modeling as a decision-making support tool. They model potential risks and develop strategies to make their businesses more robust and resilient to unexpected changes in the world.
Download our whitepaper to learn how simulation helps develop and analyze disruptive business strategies.
We’ve gathered inspiring case studies in which our clients share their project details and experiences:
