3. Techniques
3.1 Business Simulation
Guide to Business Data Analytics
3.1.1 Purpose
Business Simulation is a set of techniques used to model an outcome or a real-world scenario where the degree of uncertainty is high. Simulation produces reasonably good decision aids for stakeholders where outcomes dynamically change within the given context.
Business Simulation is a set of techniques used to model an outcome or a real-world scenario where the degree of uncertainty is high. Simulation produces reasonably good decision aids for stakeholders where outcomes dynamically change within the given context.
3.1.2 Description
Business Simulation uses a model-first approach where a representative model of a real-world interaction is created. The representation is based on domain experience, expert knowledge, knowledge of business processes, and by observing how actors interact and make decisions in that environment. Data can be fed into the model and the results of the simulation can be used to identify or predict optimal actions.
The following characteristics describe an effective business simulation:
Business Simulation uses a model-first approach where a representative model of a real-world interaction is created. The representation is based on domain experience, expert knowledge, knowledge of business processes, and by observing how actors interact and make decisions in that environment. Data can be fed into the model and the results of the simulation can be used to identify or predict optimal actions.
The following characteristics describe an effective business simulation:
- Uses many business hypotheses to create a model of the real-world process or the environment.
- Requires extensive analysis to derive the root causes for some business actions.
- May use real-time information versus historical data to prescribe actions.
- Models hard to predict outcomes with many influencing causes. The outcomes can be demonstrated through simulation tools.
- Where data is sparse. For example, for a new software product launch.
- Where intangible variables are involved. For example, modelling the of goodwill on product sales, social experience, or brand value.
- Where what-if scenarios and business test cases are involved. For example, what happens to customer adoption of a product if the price is increased by 10%?
- When live experimentation is not possible or feasible. For example, the impact of a merger between two organizations.
3.1.3 Elements
.1 Scale and Scope of Business Simulation
The scale and scope of the simulation involves determining the business scenarios and extent of the simulation that needs to be attempted. For example, simulation can be used for complete business operations, specific market and economic changes, or it can be applied to support simple business decisions. The scale of the simulation determines the extent of analysis required for building the model.
Depending on the business scenario, the following types of simulations can be attempted:
.2 Variables and Their Distribution
For a simulation experiment, input variables play a critical role. Analysts determine what variables influence the outcome, based on their experience and knowledge of the real-world scenario being modelled. A simulation model can be used when there isn't enough real-world data available. A simulation model can use synthetic data or use perceived distribution of the data. Often missing data can be modelled through distributions.
.3 Domain Knowledge, Processes, and Business Constraints
Business rules, systems knowledge, and events are key information that analysts uncover using other techniques. These are then modelled mathematically or heuristically to pre-configure the simulation model. Complex interactions between input variables are then modelled using the process and business knowledge embedded into the simulation model.
.4 Model Outcome and Orchestration
When a business scenario is simulated, outcomes can be measured and demonstrated. Unlike other predictive analytics approaches, simulation can handle different business scenario inputs and present the changed outcome. Analysts then identify and interpret the outcomes of simulations for key stakeholders. These models are often self-contained and produce visual results that business stakeholders could use via a self-service mode by providing different business test cases (for example, by changing the input variable values) to see the effect.
.1 Scale and Scope of Business Simulation
The scale and scope of the simulation involves determining the business scenarios and extent of the simulation that needs to be attempted. For example, simulation can be used for complete business operations, specific market and economic changes, or it can be applied to support simple business decisions. The scale of the simulation determines the extent of analysis required for building the model.
Depending on the business scenario, the following types of simulations can be attempted:
- Risk simulation: Use a multitude of factors to simulate outcomes when large scale changes are made to the business process or economic conditions. For example, market crashes, sovereign risks, or business risks such as mergers and acquisitions.
- Event-based simulation: It is sometimes difficult to anticipate what changes will occur if some variables are changed. The prominent examples are what-if and sensitivity analysis where a new event is injected into the model and the results are studied.
- Dynamic simulation: Involves modelling both actions and reactions in a dynamic business environment.
For a simulation experiment, input variables play a critical role. Analysts determine what variables influence the outcome, based on their experience and knowledge of the real-world scenario being modelled. A simulation model can be used when there isn't enough real-world data available. A simulation model can use synthetic data or use perceived distribution of the data. Often missing data can be modelled through distributions.
.3 Domain Knowledge, Processes, and Business Constraints
Business rules, systems knowledge, and events are key information that analysts uncover using other techniques. These are then modelled mathematically or heuristically to pre-configure the simulation model. Complex interactions between input variables are then modelled using the process and business knowledge embedded into the simulation model.
.4 Model Outcome and Orchestration
When a business scenario is simulated, outcomes can be measured and demonstrated. Unlike other predictive analytics approaches, simulation can handle different business scenario inputs and present the changed outcome. Analysts then identify and interpret the outcomes of simulations for key stakeholders. These models are often self-contained and produce visual results that business stakeholders could use via a self-service mode by providing different business test cases (for example, by changing the input variable values) to see the effect.
3.1.4 Usage Considerations
.1 Strengths
.1 Strengths
- Cause-action-reaction chains can be modelled without disrupting the business.
- Complex business situations can be modelled with accurate inputs.
- Simulations are computationally efficient and involve lower data acquisition cost.
- They are accurate for business scenarios with many contributing factors and a low amount of data.
- Simulations can be used in modelling prescriptive actions and predictions under business constraints.
- Creating effective simulations requires expert knowledge of the system being simulated.
- The outcome of a simulation experiment can be difficult to explain due to the many variables involved.
- Other types of modelling techniques are considered more effective (for example, reinforcement learning), and neural networks are becoming more accurate in driving simulation outcomes.
