Before they had Prism, Frontier consultants relied on complex spreadsheets to perform their analysis. These spreadsheets could not be deployed to clients via the web, a solution that the team preferred because it was secure, accessible, and easy to update. The team considered switching to an open-source statistics package, but they had concerns about the availability of technical support and documentation, and were not convinced that an open-source package would be suitable for developing commercial-quality products.

In creating Prism, Frontier wanted to separate the development of the quantitative analytics engine from the development of the web front end, which they intended to outsource to a third party. Because other components of their Partners Platform had already been built with .NET, they wanted to deploy the analytics using that framework.

They needed the engine to perform calculations and return results quickly, to ensure the overall responsiveness of the web application. In addition, they wanted to complete initial development of Prism and its core modules in just three months so that it could be demonstrated to clients, prospects, and industry partners at the company’s annual client conference.

Frontier used MATLAB and MATLAB Compiler SDK™ to build and deploy its portfolio analytics engine and to establish a platform for delivering frequent updates with new capabilities.

Working in MATLAB with Financial Toolbox™ and Statistics and Machine Learning Toolbox™, the team developed a strategic asset allocation module that provides 10-year forecasts for portfolio risk and return, as well as a dynamic asset allocation module that provides analytics and scenario analysis capabilities.

They used Database Toolbox™ functions in each module to access market data, expected returns, correlations, and other internal data stored in a Microsoft^® SQL Server^® database.

To calculate the probability of meeting specific targets preselected by the user, the team used functions from Statistics and Machine Learning Toolbox.

After performing unit tests on the new modules, the team used MATLAB Compiler SDK to package and deploy their MATLAB code as .NET components.

For each module, the team implemented a well-defined interface, which Frontier shared with the company developing its web front end. The front end uses the interface to pass user input to the module and receive a table of results, which it displays in charts, graphs, and tables.

Following integration testing and formal quality assurance processes, Frontier deployed the modules into production on schedule.

The company has continued to develop and deploy new modules in MATLAB, including tools for risk exposure analysis, liquidity stress testing, econometric forecasting, and portfolio comparison.

• Quantitative development decoupled from interface development. “With MATLAB Compiler SDK, we packaged our MATLAB analytics for use in a production .NET environment,” says Eriera. “This capability streamlined the entire development and deployment process by decoupling quantitative development from front end development.”
• Stable, responsive system deployed. “We’re very happy with the performance of the MATLAB code,” notes Eriera. “Any bottlenecks we see are due to the turnaround time of the HTTP request or data access, never to MATLAB.”
• Rapid delivery of new features enabled. “With MATLAB our small team is developing or enhancing multiple modules at once and delivering new releases every month or so,” says Eriera. “We can easily repurpose and customize analytics, allowing us to be more responsive to our clients’ needs.”