GCP Data Practitioner study plan

Map the four exam domains to your current experience and write a one-page gap list ranking Domain 1 sub-skills (ingestion patterns, format conversion, schema handling) from weakest to strongest

Distinguish batch versus streaming ingestion options in Google Cloud (Storage Transfer Service, Datastream, Pub/Sub, BigQuery load jobs) and record one sentence on when each is the right tool

Configure a Cloud Storage bucket with lifecycle rules, then load CSV and JSON files into BigQuery using both the console and the bq command line to see how schema autodetect behaves

Diagnose three deliberate load failures (wrong delimiter, mismatched column count, unsupported encoding) and document the exact error message and fix for each

Build a Dataflow pipeline from a provided template that reads from Pub/Sub and writes to BigQuery, then modify at least one parameter and redeploy

Compare Pub/Sub topic and subscription design choices (retention, acknowledgment deadlines, dead-letter topics) by sketching a design for a high-volume event feed

Apply schema evolution on a BigQuery table (add a nullable column, use a new schema on load) and confirm how existing rows are handled

Practice converting nested JSON into BigQuery repeated and record fields, then query the nested data with UNNEST

Write ten BigQuery SQL queries covering aggregation, window functions, joins across partitioned tables, and UNNEST of arrays, and save them as a reusable query set

Explain how BigQuery storage separates compute from storage and predict cost differences between a full table scan and a partition-pruned query on the same data

Create a materialized view and a scheduled query, then verify refresh behavior and query history in INFORMATION_SCHEMA

Identify when to use BigQuery BI Engine versus Looker Studio extract versus live connection for a dashboard scenario

Assemble an end-to-end mini pipeline: land raw files in Cloud Storage, transform with a Dataflow or scheduled query step, and serve a summary table in BigQuery

Build a Looker Studio dashboard on that summary table with at least two charts, one calculated field, and one filter control

Reconcile row counts and totals between the raw source and the dashboard to confirm no records were dropped or duplicated

Time-box a 25-question mixed Domain 1 and 2 practice set and tag every miss by sub-skill

Model a three-task dependency chain in Cloud Composer (or Workflows if Composer is unavailable) and trigger it on a schedule

Add retry, timeout, and alerting behavior to the pipeline and force a failure to confirm the alert fires

Compare Cloud Composer, Cloud Scheduler plus scheduled queries, and Dataflow templates as orchestration choices, and justify one for a nightly batch scenario

Use Cloud Logging and Cloud Monitoring to trace one pipeline run from trigger to completion

Implement idempotent loads so rerunning the same pipeline run does not duplicate rows, and prove it by running the job twice

Define IAM roles for a pipeline service account using least privilege and list the specific permissions each stage needs

Apply column-level access controls or policy tags on a BigQuery table and verify a lower-privileged account cannot read the restricted column

Configure dataset expiration and table partitioning with partition expiration, then confirm old data is dropped as expected

Design a partitioning and clustering strategy for a large fact table and measure query bytes scanned with and without pruning

Set up data quality checks (row count thresholds, null checks, freshness checks) and wire one to fail the pipeline when violated

Trace data lineage across your pipeline stages and document where PII enters, moves, and is masked

Review Cloud Audit Logs and INFORMATION_SCHEMA job history to answer who accessed what and when

Take two full-length timed practice exams and record per-domain accuracy after each attempt

Rework every missed question from Domain 3 and 4 into a one-line rule you can recall under time pressure

Rebuild from scratch, in under 60 minutes, the ingestion-to-dashboard pipeline without consulting notes

Write a final one-page cheat sheet of decision points (which ingestion tool, which orchestration service, which access control) and review it twice