Project Context
NovaBank is preparing to launch a new fraud detection model for its small-business credit card product. The model improves fraud recall by 18%, but its logic relies on ensemble features, confidence thresholds, and false-positive trade-offs that most business stakeholders do not understand. You are the program manager leading the final 8-week rollout across Risk, Engineering, Operations, Legal, and Sales.
The delivery team includes 10 people: 4 ML/Backend engineers, 2 product managers, 1 data scientist, 1 designer, 1 operations lead, and you. The COO wants launch approval at the next executive review in 3 weeks, but Legal will not sign off unless customer impact and explainability are clearly documented.
Key Stakeholders
The Chief Risk Officer wants maximum fraud reduction, even if more transactions are reviewed manually. The Head of Customer Operations wants to minimize false positives because review queues are already at capacity. The VP of Sales is pushing for a fast launch before quarter-end to support a pipeline of 6 enterprise prospects. Legal and Compliance want simple, defensible messaging for regulators and customer support teams.
Constraints
- Total remaining budget: $95,000
- Timeline: 8 weeks to production launch; executive go/no-go in Week 3
- Manual review team can absorb only 12% more cases
- No additional engineering headcount
- Customer-facing training materials must be finalized 10 days before launch
- Model monitoring dashboard depends on a data platform update due in Week 5
Complications
- Early pilot results show a 22% increase in false-positive alerts for low-volume merchants.
- The lead data scientist will be on vacation during Weeks 4-5, just as executive feedback is expected to be incorporated.
- The COO has asked for a one-page summary, while Risk leadership wants a detailed technical appendix.
Deliverables
- Create an 8-week execution plan for launch, including communication checkpoints for technical and non-technical audiences.
- Propose how you would explain the model's trade-offs, risks, and customer impact to executives in plain language.
- Define the launch recommendation, including whether to launch broadly, phase rollout, or delay.
- Identify the top execution risks and mitigation actions.
- Specify success metrics and a post-launch monitoring plan for the first 30 days.
