by Carolyn Chen and Kendall Wheaton
In all areas of the military, defence planning is an essential activity for achieving short- and long-term objectives. From a strategic outlook, the military needs to determine what capabilities, or projects, to fund today in order to build and maintain the force of the future. This can be a complex process due to multiple conflicting objectives, resource constraints (financial, personnel), scheduling and cost uncertainties, interdependencies between projects, and so forth.
Traditionally, project approval decisions have employed a ranked list approach. This approach evaluates each project individually to determine its value according to a scheme (e.g., a Multi-Criteria Decision Analysis method). Projects are then funded from most to least valuable until the budget limit is met.
While a ranked list is easy to implement, it misses out on the opportunities of examining all projects holistically as a portfolio. For example, a portfolio can be designed to include multiple low-value, low-cost projects which can result in a higher aggregate value than funding one more high-value, high-cost project. Furthermore, the ranked list approach is not well suited for complexities such as project scheduling or interdependencies.
A portfolio approach exploits optimization methods to create more valuable portfolios, which can also accommodate such complexities. The disadvantage to portfolio optimization is that it can be more difficult to implement compared to a ranked list. Therefore, portfolio decision support systems have been developed for analysts and decision makers to facilitate the analysis of portfolio problems to improve the decision quality.
Within National Defence (DND), major capital projects are managed under the Project Approval Process. The first step in this management process is the Chief of Force Development’s (CFD) Capital Investment Program Plan Review (CIPPR), which reviews major capital project submissions to determine which projects should be funded, given budgetary limitations. CIPPR is a complex process due to the hundreds of projects being considered over a 20-year planning horizon. Additionally, the scale and scope of these major capital projects varies considerably as they include new vehicles and weapons for the Army, Navy and Air Force as well as equipment for command and control, Cyber and Space, and various other departmental projects.
To address the complexities of managing such portfolios, CFD sought the creation of a portfolio optimization decision support tool to improve their process. The Visual Investment Planning Optimization & Revision (VIPOR) tool was developed by Defence Research and Development Canada’s (DRDC) Center for Operational Research and Analysis (CORA) to support decision making for the CIPPR process.
VIPOR combines portfolio optimization with interactive visualizations that allow the decision maker to analyze and explore the portfolio options. This tool uses information on the candidate major capital projects to determine the portfolio, or combination of selected projects, that maximizes the total value of the portfolio without exceeding the allotted budget. VIPOR presents the results through interactive visualizations which allow the user to manually adjust and re-optimize the portfolio in an iterative process. This lets the user explore decision trade-offs and what-if scenarios.
Upon VIPOR’s success in CFD, it was decided to create a generic portfolio optimization tool which can be used in other areas of DND and the Canadian Armed Forces. The Strategic Portfolio Analyzer with Re-configurable Components (SPARC) is the generic version of CFD’s VIPOR tool. It was developed by DRDC CORA in collaboration with the National Research Council of Canada. SPARC was initially designed to match VIPOR’s capabilities, but its continued development has resulted in improvements and expansions to the features available. This includes alternative optimization algorithms, a method for handling interdependencies, a method for handling multiple objectives, and an expanded set of interactive visualizations.
Soldiers from 5 Canadian Division on Ex Maroon Raider in February 2024. Photo: Cpl Brad Upshall
For SPARC to be used across a wide range of portfolio-type problems, a customizable data input was designed. This allows SPARC to ingest any list of items for selection, given some type of resource constraint. The items are defined by the user and can include, but are not limited to, major capital projects, minor capital projects, research projects, experiments, and tasks for the staff in a unit. The resource constraints can be any limiting factor, such as cost, personnel, time, etc. Each resource can also be defined using different types, such as different funding sources or different personnel qualifications.
SPARC can also include information external to what is needed for the optimization, which can be used to expand on the visualization capabilities. This can be categorical information, such as the geographical region, sponsor, program element, etc., which allows for items to be grouped and then the results can be visualized through the lens of the various categories.
SPARC is also designed to be used with both free optimization software (for initial testing and smaller portfolio problems) or with commercial optimization software for faster solutions of complex problems. Fast optimization is important as it allows SPARC to be used interactively to generate new portfolio results to explore alternatives within a matter of minutes, without needing to wait overnight for solutions. This improves the efficiency of the decision-making process and facilitates the usage of SPARC with decision makers, whose time is often limited.
Optimization algorithms for the coverage of different objectives and for project variants have been included in SPARC. The coverage algorithm is best applied to problems where there are multiple requirements that need to be fulfilled by the items selected for the portfolio. The algorithm will select items that maximize the coverage of all requirements, without exceeding any resource constraints. For example, a set of candidate projects may contribute to the operational functions of command, sense, act, shield, sustain and generate. The decision maker would want to ensure that the set of projects selected for the portfolio provides coverage to all these functions.
Additionally, project variants can be included when there are multiple versions of a project, but only one should be selected in the portfolio. This can be useful when a project has multiple options. For example, each project could include an option that provides the minimum capability for the lowest cost, an option for medium capability and medium cost, and third option with the highest capability at the highest cost. The algorithm will pick the variants of the different projects which best fit in the portfolio.
SPARC also includes methods for handling interdependencies between items. Interdependencies in SPARC are defined by a dependent relationship between items, where Item A must be selected with Item B, and, optionally, a synergistic relationship, where Item A is improved when selected with Item B.
Interdependencies are visualized using a force-directed network graph to allow the user and decision maker to easily view the multi-level interconnectivities of the items being considered. The interdependencies algorithm has been designed to account for these relationships within the optimization. SPARC also allows the user to selectively form or break interdependencies using three levels of dependency – weak, medium, strong – to better understand the effects of these interdependencies.
Lastly, SPARC includes a method for handling multiple objectives by facilitating the manual exploration of the relationship between objectives and constraints. This sequential satisficing method allows users to easily convert objectives into constraints and vice versa. The user can then explore portfolio options by maximizing one objective while maintaining some minimum acceptable level for the other objectives.
This exploration allows the user to see the effects of competing objectives which better informs decision makers of the trade-offs between objectives. Consider an example where a decision maker needs to select projects which have been evaluated based on their project value as well as their risk level. Sequential satisficing can be used to determine what level of risk is required to achieve the minimum level of total portfolio value or determine what value can be obtained for the overall portfolio without exceeding a specified maximum risk level.
SPARC is now being integrated and used within the planning processes for major and minor capital infrastructure projects.
For major capital infrastructure projects, SPARC is being used as the basis for their Capability-based Infrastructure Requirements Tool (CaBIR-T), which supports decisions on which new infrastructure requirements to fund given their on-going set of projects and budgetary limitations. Additionally, SPARC is being integrated in the planning process for minor capital projects for the West region. In both cases, SPARC uses a customized value calculator that follows the scoring scheme for each process.
SPARC has also been explored for use in project selection for experimentation portfolios, as well as for applications in materiel management, personnel task management, and capability analysis. For example, SPARC could be used to select the combination of equipment to buy for training kits, to provide the highest impact within the available budget. It could also be used to determine tasks to complete within a headquarters given limited personnel capacity. SPARC can be used for a wide range of problems as long as they fit the portfolio type, where there is a list of items to be selected and some factor limiting the selection.
Dr. Kendall Wheaton ([email protected]) is a Defence Scientist with DRDC in the Centre for Operational Research and Analysis (CORA). As an operational research analyst, he has supported the Navy, Air Force, Joint Staff, Canadian Joint Operations Command, and currently, the Chief of Force Development. In this posting, he develops decision support tools to help strategic planners responsible for capability development.
Carolyn Chen ([email protected]) has previously worked at Natural Resources Canada and the Canada Border Services Agency on projects related to radiation detection and unmanned aerial systems. In 2021, she joined DRDC CORA as a member of the Strategic Planning Operational Research Team. During this time, she worked on the SPARC project where she supported capital investment planning processes for both the CFD and the Assistant Deputy Minister (Infrastructure and Environment). She has recently moved to the Operational Research team supporting the Canadian Army.
