Applying Agile Methodologies to Manage Work and Drive Value
By Carolyn McArthur
The Stepping Stone, January 2024
Actuaries work in a constantly changing environment, characterized by fluctuations in market trends, technological advances, and regulatory changes. The risk is continually shifting, and yesterday's solutions may not address today's challenges. So, actuaries need to innovate and experiment with new models, methods, and approaches.
In the November 2023 issue of The Stepping Stone,[1] we discussed how to apply an agile mindset and embrace soft skills to address that ever-changing environment. Now let’s look at how agile methodologies can be used to help manage work and drive value in the organization.
Agile methodology is a collaborative and iterative approach that emphasizes flexibility, adaptability, and continuous improvement. It can enable actuarial teams to break down complex projects into smaller, manageable tasks, allowing for faster delivery, enhanced collaboration, and increased customer feedback. Although these methodologies originated in the software development industry, they have been adapted to many industries including insurance.
For actuaries, agile methods can create a workflow that is more adaptive, responsive and efficient. There are many agile frameworks. Three of the most common are Scrum, Kanban and Lean. Here we will show how these might apply to a predictive modeling initiative.
Scrum
The Scrum[2] framework (most widely used) involves breaking down complex work and organizing it into short, time-boxed development iterations that are referred to as “Sprints,” generally two to four weeks in which a selected group of tasks is completed. It involves three primary roles: product owner (representative of the business and stakeholders); Scrum master (manager who ensures the team is following the Scrum process and removes impediments); and the development team (group responsible for delivering the product or increment of work).
There are five regular ceremonies (meetings) held, described below: daily stand-up, sprint planning, sprint review, backlog refinement (grooming), and sprint retrospective. Teams plan the work at the beginning of each sprint and review at the end of the time-box.
The Scrum method can be applied to predictive modeling, enabling actuaries to set goals, prioritize tasks to develop, and refine models iteratively. By breaking down complex models into smaller components, actuarial teams can validate assumptions, add new data, and adjust models based on trends that emerge.
Within a Scrum framework:
Sprint planning would be the first collaboration of the team to outline the goals for the sprint (typically a two to four-week period). The full backlog or list of tasks is reviewed (backlog refinement). Tasks that align with the specific goal of the sprint are extracted from the list (backlog) and prioritized by the team to create the sprint backlog (a subset of tasks which can be completed during the time allotted).
The team meets in a daily stand-up or Scrum meeting during the sprint to stay aligned and focused on the sprint goal. The team responds to three basic questions: What did you do yesterday? What will you do today? What obstacles or impediments are blocking your progress?
At the end of each sprint, they conduct a sprint review with the stakeholders to review the tasks completed, show progress, ensure alignment with the predictive model goal, and gain feedback.
During the sprint, the leader organizes a backlog refinement (or grooming) meeting as needed (typically one per sprint) to update the backlog with prioritization, estimate the level of effort, add new tasks, update requirements, adjust scope, estimate effort, and include changes based on stakeholder input.
After the sprint is complete, the team holds a sprint retrospective to reflect on the success of the sprint and identify any processes that could be improved going forward. The team decides on any actionable items that can be incorporated into the next sprint to increase efficiency and enhance the predictive modeling process.
This iterative approach can help ensure the team remains agile, adapts to changes, techniques or requirements, and can incrementally improve the models, where often times many rounds of refinement and validation are needed prior to releasing to the end user.
Kanban
Kanban[3] is a visual framework to manage workflow by prioritizing and moving tasks through various stages. The three main stages are: “To Do,” “In Progress,” and “Done.” Actuarial teams can use Kanban visual boards to track tasks, identify impediments and optimize the workflow.
Within a Kanban framework:
Visualizing the workflow would be the first effort, where a Kanban board with columns outlining the stages of predictive modeling and task cards created related to the process are placed in the appropriate column. For example, the columns might be Data Gathering, Data Cleansing, Development, Model Training, Validation, Deployment, and Complete. The tasks would be specific actions or steps required for each process, placed under the appropriate column on the board.
For each stage of the workflow, a work in progress (WIP) limit is established to ensure the team is focused on completing tasks before starting new ones and to prevent bottlenecks. Along with the WIP, a “done” is defined to provide a clear understanding of what “done” means at each stage. For predictive modeling this could mean a model has passed validation before moving to the next stage in the process.
The tasks are constantly monitored to balance the flow. If a column on the board hits its WIP, the team needs to address the root cause and address the bottleneck.
Progress tracking is done for tasks as they transition from one stage to the next along with any stage where work is being delayed throughout the workflow. Corrective actions are identified and taken throughout the process to alleviate issues.
Team meetings are held to discuss progress, obstacles and process improvements. Unlike Scrum, there are no prescribed meetings or cadence. Kanban uses cycle time, lead time, throughout, and flow efficiency as measures to analyze the efficiency and areas for improvement within the process.
Throughout the Kanban environment, a culture of continuous improvement, regular reflection, feedback loops, and incremental changes are promoted to improve efficiency and quality of the delivery to the end user.
Actuaries are always juggling many tasks, and this method promotes transparency, efficiency, and prioritization of work based on current demand. Using Kanban can help ensure that tasks are planned and well-organized so that the team can quickly adapt to changes, and continuously deliver value not only in predictive modeling but also efforts like assumption setting or data validation.
Lean
The Lean[4] methodology focuses on maximizing the value brought to the customer while minimizing non-value-added activities. This promotes a continuous improvement culture where everyone is involved in identifying and improving the process by optimizing the steps throughout the process. Actuarial teams can apply this method to streamline the predictive modeling, data ingestion, data extraction, data analysis, model development, and reporting processes.
Within a Lean framework:
Identifying and defining what comprises value from the stakeholder or end-user perspective is the first effort under this method. The entire process would be mapped to identify the steps that add value (value stream mapping).
For example, using predictive modeling, the value would be identified from data gathering to model deployment. The current state is outlined by documenting the steps, potential impediments, delays, communication, and sub-processes needed to deliver the valued initiative—in this case the predictive model. Then the future state is outlined by removing any non-value added steps from the process. Create flow through eliminating the non-value add steps, sequencing the work, and standardizing the workflow to ensure unnecessary steps are eliminated. The proper flow maintained consistently delivers value to the end-user that is critical to the success of using this framework.
Like Scrum and Kanban, an environment of continuous improvement, regular reflection, feedback loops, and incremental changes are promoted to improve efficiency and quality of the delivery to the end user. Lean also incorporates a strong element of experimentation to find better ways of creating additional value to the end-user. In predictive modeling this could be identifying consolidated data sources or new algorithms.
Throughout the Lean method the focus is on encouraging cross-functional collaboration, team engagement, process improvement and upskilling the entire team as a whole to solve problems and break down silos. The objective is to avoid sub-optimization and ensure meeting overall goals. Applying this methodology to predictive modeling means focusing on delivering models that add value, providing actionable insights with minimal waste throughout the process. By continuously reviewing, refining, and eliminating waste, a process like this can become more efficient, delivering high-quality models quicker and at potentially lower cost.
These methods and others have revolutionized the way work is done, projects are executed, and solutions delivered. With frameworks like Scrum, Kanban and Lean, actuaries can quickly adjust to changes, ensure solutions align with stakeholder requirements, promote cross-functional collaboration, and increase efficiency in delivering high-quality products.
Statements of fact and opinions expressed herein are those of the individual authors and are not necessarily those of the Society of Actuaries, the editors, or the respective authors’ employers.
Carolyn McArthur, MBA, MSITM, ALMI, is an assistant vice president at SCOR Global Life, a seasoned insurance industry expert with a wealth of business and technology experience maintaining certifications in insurance, agile leadership, six sigma and project management methodologies. She can be reached at CMCARTHUR@scor.com or via LinkedIn.
Endnotes
[1] “Beyond Technical Proficiency: The Role of Soft Skills in Embracing Agility,” https://www.soa.org/sections/leadership-development/leadership-development-newsletter/2023/november/ss-2023-11-mcarthur
[2] https://Scrumalliance.org
[3] https://kanban.university/resources/
[4] https://www.lean.org/explore-lean