CDDP study plan

Map the full data-governance operating model: decision rights, stewardship roles, policy hierarchy, and escalation paths, and write a one-page diagram of who owns what.

Draft a data-ethics decision checklist covering purpose limitation, consent, bias review, and harm assessment that you can apply to any new data use case.

Classify ten realistic governance scenarios (e.g., a marketing team requesting sensitive customer attributes) by which policy, role, and approval gate applies.

Explain the difference between accountability, ownership, custody, and stewardship, and state who holds each for a customer master dataset in a mid-size retailer.

Build a regulatory comparison table for the major privacy regimes you may face (GDPR, CCPA/CPRA, HIPAA, and one sectoral rule), listing lawful basis, data-subject rights, breach timelines, and penalties.

Trace a data subject access request and an erasure request end to end through a hypothetical architecture, naming every system that must be queried or purged.

Apply privacy-by-design principles to a new analytics pipeline: minimize fields, set retention, define pseudonymization points, and document the lawful basis for each field.

Distinguish anonymization from pseudonymization and de-identification, and identify at least three re-identification risks in a proposed 'anonymized' release.

Draft the access-control section of a data-handling policy: role-based access, least privilege, masking tiers, and periodic recertification.

Compare conceptual, logical, and physical model layers and produce all three for one business process, showing how each layer constrains the next.

Normalize a messy order-and-customer dataset to third normal form, then deliberately denormalize it for a reporting star schema and justify each trade-off.

Design a star schema and a snowflake variant for the same subject area, and state when each is preferable for query performance and maintenance.

Model slowly changing dimensions types 1, 2, and 3 on a customer attribute and write the surrogate-key and effective-dating rules for each.

Write the core data-architecture artifacts for a target state: a context diagram, a layered architecture (source, integration, warehouse, semantic, consumption), and a data-flow narrative.

Select storage engines for four workload profiles (transactional, analytical columnar, document, and time-series) and justify each choice on access pattern, concurrency, and cost.

Compare batch, micro-batch, and streaming integration patterns and pick one for a stated latency requirement, documenting the trade-offs.

Design an idempotent, restartable ingestion job with watermarking, deduplication keys, and late-arriving-record handling.

Specify partitioning, clustering, and file-layout choices for a large fact table and predict their effect on scan volume and query latency.

Write an integration specification covering source contracts, schema evolution rules, error queues, and reconciliation counts.

Define six data-quality dimensions and write a concrete, testable rule for each against a real dataset you can inspect.

Build a data-quality scorecard with thresholds, owners, and alert routing, and run it once to produce actual pass/fail results.

Design a master data management approach for one entity: match and merge rules, survivorship logic, golden-record definition, and stewardship workflow.

Diagnose duplicate customer records and document the matching algorithm, including blocking keys, fuzzy thresholds, and false-positive controls.

Write a root-cause analysis for one recurring quality defect and propose a preventive control at the point of entry rather than downstream cleanup.

Choose the correct statistical test for five stated business questions (difference of means, proportion comparison, correlation, trend, and A/B outcome) and justify each selection.

Interpret p-values, confidence intervals, and effect sizes correctly, and write plain-language explanations a non-technical stakeholder would accept.

Identify confounding, selection bias, survivorship bias, and Simpson's paradox in four short scenarios and state the corrective action for each.

Perform descriptive and inferential analysis on a public dataset, reporting distributions, outliers, and at least one hypothesis test with assumptions checked.

Explain when correlation cannot support a causal claim and describe one design (holdout, matched comparison, or difference-in-differences) that could.

Match chart types to analytical intent (comparison, distribution, composition, relationship, trend) and rebuild three poor charts into effective ones.

Apply perceptual principles: pre-attentive attributes, ordering, color use for categorical versus sequential data, and accessibility contrast requirements.

Design a dashboard for a named audience with a stated decision it supports, defining the metric hierarchy, filters, and drill path.

Critique one dashboard against a checklist covering clutter, misleading axes, missing baselines, and ambiguous labels, and document the fixes.

Write the narrative layer for a report: headline, supporting evidence, and recommended action, keeping it free of chart-reading narration.

Compare warehouse, lakehouse, and data-mesh architectures on governance model, coupling, and cost, and recommend one for a stated organizational context.

Evaluate two emerging capabilities (e.g., vector search, streaming lakehouse, automated lineage, or feature stores) on maturity, risk, and fit for a real use case.

Assemble a consolidated one-page formula sheet spanning all eight domains: governance roles, modeling normal forms, quality dimensions, statistical test selection, visualization rules, and privacy controls.

Complete a timed 60-question mixed practice set and log every miss by domain and root cause (knowledge gap, misread, or guess).

Re-drill your three weakest domains using targeted review and rewrite the corresponding formula-sheet sections from memory.