Getty Images

Tip

Simulation and predictive analytics boost forecast capabilities

Simulation and prediction analytics cover two different ways to forecast data. Together, they can boost capabilities, but organizations must craft them carefully to trust the results.

Analytics is essential for data-driven organizations to make decisions, but predictive analytics and simulations are key for understanding future trends and determining optimal responses.

Enterprise leaders are using analytics at an increasing rate, with one recent survey saying 84% of organizations have already deployed or are planning data-driven projects. Organizations plan to spend an average of $12.3 million on data-driven initiatives in the upcoming year, Foundry's "2022 Data & Analytics Study" reported.

Modern enterprise analytics programs encompass a range of capabilities. Many organizations are focusing on descriptive and diagnostic analytics, both of which are on the lower end of the complexity scale. The more mature enterprises, though, have implemented the more advanced predictive analytics and simulation modeling.

"Simulation and predictive analysis are both techniques to understand outcomes," said Faisal Malik, national lead of the data and analytics practice at Centric Consulting.

What is predictive analytics?

Predictive analytics is the use of historical and current data to make predictions about what will happen in the future and the likelihood of those predictions happening. It uses statistical analysis and machine learning algorithms to create predictive models and then analyzes data sets to offer potential future outcomes.

"It's a statistical guess. You're using statistics to guess what you think the future will be," said Allison Hartsoe, CEO of technical consulting firm Ambition Data, author of The Age of Customer Equity and member of the Digital Analytics Association.

However, unlike manually calculated statistical models, predictive analytics uses modern information technology, which enables organizations to run queries and analyses on more data with more variables to produce more fine-tuned predictions.

"The fact that I can do that in a more complicated fashion makes predictive analytics much more powerful than sheer statistics," Hartsoe said.

Flow chart outlining the predictive analytics steps in the data management process
The predictive analytics data management process

What is simulation?

Simulations are often used by organizations trying to understand outcomes in high-risk scenarios or in unique situations for which there's no historical equivalent, according to Malik and other experts.

"Simulation is a way of doing a lot of experimentation without investing a lot of time, money and effort," Malik said. "You can simulate environments and learn from them what is the optimal course of action that I want to take, and once I learn that, I can program it."

For example, NASA and other space exploration entities would use simulations to study how to land a spacecraft on Mars.

"You can simulate that at a very reasonable cost and at low risk before landing the shuttle and finding out you made an error," Malik said.

Automakers use simulations to create driving scenarios in various environments that teach self-driving cars. Aircraft manufacturers can simulate aircraft performance in various conditions. Similarly, elevator companies can use simulations to understand the performance of their machines in variable circumstances -- even those that are highly improbable and, therefore, both costly or impossible to replicate in the real world by testing actual physical equipment.

Simulations versus predictive modeling

Predictive analytics is a type of simulation, but simulation and predictive analytics differ in what kind of data each one uses.

"Predictive analytics and simulations are both models; they're both done with computers," said Tom Coughlin, Life Fellow with the professional association IEEE. "But predictive analytics is constrained by use of actual historical data, [whereas] you can create new types of data to help you in your predictions by using simulations."

Simulations can make data sets that don't exist in the real world or go beyond what has happened, Coughlin said. Combining the simulated data with real data can make predictive analysis more powerful or deal with situations that haven't happened yet.

"In a simulation, you have a bit more freedom for the things you want to explore," he said.

Examples of simulations in action are the models used to determine what could happen -- and what are the most likely outcomes -- if sea levels rise a foot worldwide.

"There's no historical information on that, but a simulation can help you get an understanding of that," Coughlin said.

Another difference between predictive analytics and simulations is the timing of their use, with predictive analytics tending to be more of a point-in-time decision-making tool and simulations being persistent to enable ongoing analysis.

With simulations, "we're building a data model of something, and we want to subject it to different digital analyses," said Igor Ikonnikov, research and advisory director in the data and analytics practice at Info-Tech Research Group.

Consequently, organizations run predictive analytics using in-memory databases, Ikonnikov said, which supports the speedier response times typically required by organizations when using predictive analytics to make decisions.

Simulations, however, are typically stored in traditional database systems, which store data on persistent media, Ikonnikov said. This enables organizations to compare simulations and perform "cross-trajectory analysis."

"We can track trajectories in all details. We can tweak tiny details. We can tweak variables at certain points of time," he said.

Benefits of simulations in predictive analytics

Like all analytics, simulations help organizations better understand potential scenarios and possible outcomes. Simulations can also support predictive analysis.

"The variables you're getting from simulation can feed into predictive analytics," Malik said.

For example, financial firms can use simulations to understand the effect of a war on the stock market, Malik said. They can use the simulation within their predictive analysis to predict the probability of specific outcomes given a specific action in the simulated market condition.

Simulations enable exploration of more possibilities and organizations to analyze circumstances that go beyond what they historically have seen.

Additionally, simulations also let executives exercise their decision-making muscle in a somewhat risk-free environment, as they enable enterprise leaders to see how choices will play out in something of a sandbox environment, experts said.

"They can learn the outcome in the simulation, but they don't have the obligation to then go and use that decision," Hartsoe said.

Furthermore, simulations enable exploration of more possibilities and organizations to analyze circumstances beyond what they typically experienced, experts said.

In turn, this enables enterprise leaders to delve into and identify what factors and decisions affect outcomes, something that helps them to be better prepared to make decisions if the simulated circumstances become reality.

"It helps the executive frame business decisions [by understanding] what levers they can pull, to run through and practice different ways to get to a specific outcome," Hartsoe said. "So, as they make choices, they know there are things they can add in and not add in."

There are other benefits, too. Organizations can use simulations to check for biases in their analytics programs, as well as validate results, Coughlin said.

Because they're persistent, simulations help create auditability and transparency within analytics and decision-making, enabling organizations to see "what scenarios, what assumptions were used to make this conclusion," Ikonnikov said. Their persistent nature also enables repeatability.

"If you create certain simulations and you persist those scenarios in the databases, you can reuse the same data or come back to the same question and continue to tweak it and analyze the results," he said. "It shortens the time to answer the question."

Challenges with simulations

Organizations face some challenges when creating simulations. Each organization must ensure it has the quantity and quality of data required for their models, Coughlin said, as falling short in either area could produce poor-quality and untrustworthy results.

Results can be faulty if the modeling of something in the simulations is incorrect, a typical challenge with any type of analytics, Ikonnikov said.

Organizations must be diligent in their treatment of the synthetic data sets used in simulations to ensure "they're using them for analytics only -- they're not for operational reports," he said.

Growing use of simulations

Despite such challenges, demand for simulation capabilities is quickly growing. It's estimated the simulation and analysis software market share will increase by $7.98 billion from 2021 to 2026, with year-over-year growth of 11.92% for 2022, in a recent report by Technavio, a London-based market research and advisory company.

Experts expect more companies to use simulations as part of their analytics capabilities as they start or increase their investments in simulation and analysis software and as they advance their enterprise data and analytics programs.

As that happens, enterprise executives must be clear about which analytical capabilities to apply when, Malik said.

"These techniques are very useful, but you have to be savvy enough to know whether to use predictive analysis or simulation to solve your problem," he said. "You have to ask what is the best tool to apply to get the results."

Dig Deeper on Data science and analytics

Data Management
SearchAWS
Content Management
SearchOracle
SearchSAP
Close