INFORMS CAP study plan

Draft a one-page problem-framing template that converts a vague stakeholder request into a decision statement, success metric, and constraint list

Practice reframing three short business scenarios into analytics problem types (descriptive, predictive, prescriptive) and justify each choice in writing

Map the CAP methodology domain into a personal checklist covering problem type, method selection, assumptions, and validation strategy

Compare at least four method families (regression, classification, optimization, simulation) on data requirements and interpretability trade-offs

Write decision statements for five practice cases and identify the stakeholder, decision cadence, and cost of a wrong call for each

Build a method-selection decision tree that routes a framed problem to a candidate technique with stated assumptions

Solve two small optimization and two small classification exercises by hand or in a spreadsheet to internalize mechanics

List the failure modes of each method family and the diagnostic you would run to detect each one

Inventory the data lifecycle from sourcing through retention and write the control you would put at each stage

Audit a public dataset for completeness, leakage risk, class imbalance, and sampling bias, documenting each finding

Draft a data-quality scorecard with concrete thresholds for missingness, duplication, timeliness, and drift

Write five questions you would ask a data owner before trusting a new source

Split a dataset into train, validation, and test sets and document the split rationale and leakage checks

Train a baseline and two improved models, tracking metrics, hyperparameters, and random seeds in a run log

Write a validation plan that separates offline metrics from the business metric and states the acceptance threshold

Diagnose one deliberately overfit model and record the interventions that reduced the gap

Design a deployment architecture diagram showing batch or real-time serving, feature pipeline, monitoring hooks, and rollback path

Define the pre-launch checklist covering performance, latency, security review, and stakeholder sign-off

Write a shadow-mode and canary rollout plan with explicit promotion and abort criteria

Specify the monitoring dashboard: inputs, outputs, drift signals, and alert thresholds

Draft a model risk register listing at least six risks with likelihood, impact, owner, and mitigation

Write a retraining policy that states triggers, cadence, ownership, and validation gates before promotion

Design a versioning scheme for data, code, and models and document how to reproduce any past prediction

Outline a model retirement and decommissioning procedure including downstream dependency checks

Write a one-page executive summary translating a technical result into a decision recommendation with quantified impact

Prepare a ten-minute stakeholder briefing and rehearse it aloud, timing yourself and cutting jargon

Build a RACI for a cross-functional analytics project spanning business, data engineering, and IT

Draft responses to three likely pushback questions about model uncertainty and cost

Apply an ethical decision framework to three scenarios covering privacy, fairness, and disclosure of limitations

Document how you would test a model for disparate impact and what you would do if the test failed

Review your weeks 1-7 deliverables and close the three weakest gaps with targeted practice

Complete two timed mixed-domain practice sets and log every missed concept with a corrective action