The INFORMS Analytics Framework™: A Road Map for Success with Analytics

The INFORMS Analytics Framework™ is a comprehensive structure designed to guide practitioners at any level through the entire life cycle of analytics initiatives, from initial business concept to sustained, value-delivering operations. It stands as a critical reminder that a successful analytics solution is much more than a well-written mathematical model or creative software solutions. The Framework provides repeatable practices and methods to avoid pitfalls and realize successful achievement. Exceptional, reliable and trustworthy analytics efforts require the incorporation of excellent communication across all team elements – engagement strategies, common understandable objectives, disciplined use of analytics and most importantly, clear agreed-upon problem (question) framing. A common theme throughout the Framework is the constant connection to the business objectives in activities within each domain. 

Developed from the original 2013 Certified Analytics Professional (CAP) Job Task Analysis, the Framework was substantially updated and rebranded in 2024 to support the full spectrum of modern analytics, from simple dashboards and machine learning (ML) to complex optimization and artificial intelligence (AI) initiatives. This Framework now serves as the foundation for the new three tiers of the CAP exam (Essentials, Pro and Expert) and is widely used in industry, public sector and academic settings to ensure that applications of analytics deliver genuine business impact value.  

The Framework comprises seven sequential yet highly iterative domains. Uncovering issues in a later domain (such as the Data domain), or having to start in a different domain by necessity, often requires circling back to refine earlier domains (such as Business Problem Framing). 

The Seven Domains of the INFORMS Analytics Framework

1. Business Problem Framing

Any application of analytics must begin with a clear, concise statement describing the business problem or question, not a technical one. The goal is to ensure that analytics is being applied to solve the right challenge for the organization. 

• Problem Statement and Stakeholder Identification: Develop an initial statement of the problem or question, and identify all stakeholders, sponsors and beneficiaries who will be affected by or responsible for the solution. 
• Amenability Check: Determine whether the problem or opportunity is truly amenable to an analytics solution. Many problems initially presented as analytics challenges are, in fact, solvable only by addressing underlying business process problems. If processes or data collection is inadequate, analytics must pause and be reevaluated until those issues are resolved. 
• Business Case Creation: Refine the problem statement until it is clear and concise; then, develop a business case. This includes defining the cost of the solution (e.g., software licenses, cloud resources, database setup, data acquisition), expected benefits and organizational effects, including any changes required for the solution to be adopted and used. 
• Sponsor and Stakeholder Agreement: Secure full agreement from sponsors and key stakeholders on the problem to be solved before any modeling work begins. 

2. Analytics Problem Framing

This domain translates the agreed-upon business problem into an actionable analytical structure. 

• Reframing: Convert the business problem statement into an analytics problem statement (e.g., “This is an optimization problem” or “This is a classification problem”). This is done at a business level, defining the constraints, decisions and objectives, but without writing any mathematical formulas or developing any software. 
• Drivers, Inputs and Assumptions: Define, conceptually, the key drivers and inputs (e.g., fuel costs, traffic) that will inform the model and the required outputs. Clearly state any necessary assumptions (including realistic constraints) required to make the analytics applicable, because these may limit the scope of the solution. 
• Key Performance Measures and Baseline Performance: Define the primary measures of success from an analytical standpoint (how success is measured with numbers or key indicators). Crucially, establish the baseline performance – the current measurable quality of decisions or processes that the new solution must surpass or present to demonstrate value. 
• Risk and Agreement: Identify initial risks to the analytics effort (e.g., uncertainty about data availability, resource constraints) and secure stakeholder agreement on the analytics approach chosen to solve the business problem. 

3. Data 

Data preparation is the most time-consuming domain and, frequently, most of the effort in leveraging analytics to deliver business value. A data-first approach is essential because data quality issues can invalidate even the best models. 

• Needs, Sources and Management: Determine the specific data needed to represent the conceptual drivers and inputs. Identify sources and their structure (tables, streaming, etc.). Develop a management plan for how data will evolve, whether it will be resident or transported for use, and how it will be maintained over time, as well as a plan for data accessibility controls. 
• Acquisition and Cleaning: Acquire the data, including a robust test set. The bulk of the work involves cleaning, harmonizing, validating and ensuring consistency across all data points and tables. 
• Documentation: Create a comprehensive data document listing all tables, columns, relationships, use constraints, owners and data assumptions. This serves as a vital reference for the entire team. 
• Iteration: The findings from data analysis often expose initial flawed assumptions, requiring the team to loop back to update and validate the business and analytics problem statements. 

4. Methodology Approach

With the problem defined and the data understood, the next step is selecting the appropriate analytics methodology. 

• Method Selection: Determine the appropriate analytical techniques (e.g., linear programming, heuristics, specific machine learning algorithms) and select the best one based on available resources, team capabilities, policies and data. 
• Architecture and Technology Stack: Define the solution technical architecture: how data will link to the model, where the solution will be deployed (e.g., cloud) and how updates will occur. Consider the environments in which the analytics solution will be developed, tested, operationalized (production) and backed up for disaster recovery. What kind of user interface is required? Then, select the specific technology stack. 

5. Model Building and Evaluation

Although technically the core, this domain is often only 10%-15% of the total time spent when applying analytics. 

• Design and Structure: Design and build one or more models. Sometimes, multiple models interact or are being compared. One best practice is to write out the mathematical representation of a model before developing any part of a software solution. 
• Validation and Trust: Run the model, evaluate its performance against the analytics measures and get user feedback to ensure the solutions make sense. To establish trust, build user interfaces that allow business users to evaluate the solutions and provide feedback.  
• Documentation: Write thorough documentation about the model’s performance, assumptions, limitations and value. 

6. Deployment

Deployment is a two-pronged effort including technical implementation and organizational adoption. 

• Technical Deployment: Stand up the technical application, which may be an operational system running 24/7 (e.g., solving a problem every five minutes), real-time action indicators or a strategic report generator. 
• Business Deployment and Validation: Deploy the new decision-making process into the business. This involves ensuring that people are trained on various aspects of the solution, which helps build trust and institutional knowledge in that solution. Obtain a business validation report and stakeholder agreement that the solution is meeting the needs of the business and is ready for use. 
• Implementation Support: Analytics professionals must support the final implementation and testing to verify that the deployed solution is working correctly and that data is flowing reliably once it enters live production. 

7. Solution Life Cycle Management

A successful solution requires perpetual care to continue delivering value. 

• Performance Tracking and Recalibration: Continually track the solution’s performance and whether it continues to deliver business value. Be ready to recalibrate models to prevent model drift as business constraints, data streams or external factors change over time. 
• Side Effects and Organizational Change: Continuously validate the business case and manage the solution’s side effects. Dynamic adaptability is critical to successful analytics. 
• Training and Documentation: Support training activities for end users and ensure thorough knowledge transfer to internal teams. Maintain complete, up-to-date documentation so the solution remains supportable and evolves with the business. 

Tasks of the INFORMS Analytics Framework 

The team that created the new INFORMS Analytics Framework was also responsible for defining testing objectives for each of the three new levels of the CAP certification. These testing objectives are termed the “blueprints” for each level of the exam. Each of the testing objectives is associated with specific tasks within each of the domains. Whether as part of a specific analytics project or any other investigation into whether analytics is applicable to a business problem, these tasks, summarized above, should be considered to ensure analytics success. The Framework can become an essential component of operational excellence and the structure by which successful analytics efforts are approached and maintained. We recommend downloading the complete INFORMS Analytics Framework for a deeper dive into the specific details. 

More information about the INFORMS Analytics Framework can be found at informs.org/analyticsframework. 

Acknowledgment 

We would like to acknowledge the work of our colleagues who spent many hours in 2024 to define the detailed examination criteria in the blueprints that led to the new framework development. These colleagues are Shannon Browning, Arnie Greenland, Mehran Hojati, Thor Osborn, Zohar Strinka and Nick Ulmer (all CAP-X certified).