Business Problem Framing
Domain 1: Business Problem Framing
The CAP exam tests whether you can turn an ambiguous business situation into a scoped, decision-ready analytics problem before any data or model work begins. Expect questions on problem statements, stakeholder alignment, success criteria, constraints, and value framing.
Core rules and concepts
- Separate the business question ('How do we reduce customer churn?') from the analytic question ('What is the 90-day probability that a customer will churn?'). The analytic question must support a specific decision, owner, and action.
- Identify stakeholders: executive sponsor, decision-maker, subject-matter experts, data and IT owners, legal and privacy, and end users. Elicit requirements through structured interviews, not assumptions.
- Define success at three levels: business KPI (churn rate, margin), operational acceptance (campaign capacity, threshold), and model performance (AUC, lift, calibration). State baseline and target.
- Write scope explicitly: in-scope populations, time windows, systems, and out-of-scope items. Record constraints (budget, schedule, data access, regulation), assumptions, and risks.
- Frame value: cost-benefit, expected value of better decisions, ROI, and decision latency. A model that cannot change a decision has no business value.
- Capture the result in a charter or problem statement: background, decision, analytic question, success metrics, scope, stakeholders, assumptions, milestones.
Worked example
A subscription retailer wants to 'reduce churn.' The analyst reframes it as: which active consumer subscribers are most likely to cancel in the next 90 days so the retention team can target a limited incentive budget? Success is a 3-point reduction in quarterly churn versus baseline, with campaign capacity of 20,000 contacts and a model that ranks the top 20% by lift. Scope covers consumer accounts with 6+ months tenure; business accounts are excluded. Stakeholders include the CMO (sponsor), retention manager (decision owner), CRM IT (data and execution), and legal (incentive compliance). Assumptions: churn is defined as no renewal within 30 days of expiry; incentive budget is fixed. This framing determines the label, population, threshold, and deployment mode.
Traps that cost marks
- Jumping to method or data before agreeing on the decision and success metric.
- Accepting a vague goal such as 'improve insights' with no owner or action.
- Ignoring operational constraints (campaign capacity, latency, budget) that determine the useful model form.
- Treating stakeholder requirements as fixed without validating assumptions or documenting scope creep.