AMAADOR ACADEMYCertificatesVerify a certificate

DEET

Final check — Data Engineering: ETL, Warehousing, dbt and Airflow

Answer the 42 questions. You need 80% or more to pass. You can retry as many times as you like.

Study the course →

Question 1A retailer must stop selling an item the moment the last unit is bought, across web and shops. Which design fits?
Question 2What does the abbreviation ELT stand for, in order?
Question 3Company policy says employees’ bank account numbers must not be stored in the analytics platform at all. The HR system holds them. What do you do?
Question 4Which reason best supports choosing ELT on a cloud warehouse?
Question 5The board pack is built once a month from sales KPIs. A colleague proposes a streaming platform “to be modern”. What is your answer?
Question 6Which difficulty is typical of streaming pipelines but much smaller in batch?
Question 7Your incremental load copies rows with updated_at after the last run. Over months, the warehouse holds more orders than the source. Why?
Question 8A task is idempotent when…
Question 9A Dag loads “yesterday’s” data with a filter on the current date minus one day. It failed for two days and was fixed on day three. What went missing?
Question 10Why is Parquet usually preferred to CSV for analytics storage?
Question 11Some orders are committed in the source a few minutes after their updated_at time, so an incremental load sometimes skips them. What is the standard fix?
Question 12Last night 1,000,000 rows were exported but the table received only 65,536, and the job still showed green. Which control would have caught it?
Question 13In dimensional modelling, what does the “grain” of a fact table define?
Question 14A fact table has one row per order line and a column “monthly_budget”. Totals per month are far too high. What is wrong?
Question 15Which slowly changing dimension type adds a new row and keeps full history?
Question 16A customer’s last name was misspelled in the source and has just been corrected. Reports should show the correct spelling for all years. Which handling fits?
Question 17Why use a surrogate key in a dimension instead of the source system’s id?
Question 18A report sums the end-of-month account balance over twelve months and gets a number twelve times too large. What is the right approach?
Question 19What is the main benefit of a columnar warehouse for analytical queries?
Question 20Most queries filter on event_date and then on country. The 3 TB events table is partitioned by event_date only, and queries on one country are still slow. What helps most?
Question 21Which capability do open table formats such as Iceberg and Delta Lake add on top of Parquet files?
Question 22Executives query a dashboard that reads the raw landing tables directly. Numbers change when a late file arrives and duplicates appear. What is the better design?
Question 23A team must store tables, JSON logs and images, query tables with SQL and train models on the same data, without copying it between systems. Which pattern fits best?
Question 24What does partitioning a large table by date let the engine do?
Question 25What is the main role of the ref() function in a dbt model?
Question 26During dbt build, the unique test on stg_orders.order_id fails and downstream marts are skipped. A colleague suggests removing the test to unblock the run. What should you do?
Question 27When does a dbt data test pass?
Question 28A 500-million-row events model is rebuilt fully every night and now takes three hours and a lot of compute. New events arrive only for the last day or two. What is the best change?
Question 29You rewrite a CASE expression that assigns customer tiers and want to be sure the edge cases work before the real data contains them. Which dbt feature helps?
Question 30What does a dbt snapshot do?
Question 31Why is a workflow in Airflow described as a Dag?
Question 32A report Dag is scheduled for 06:00. When the nightly load runs late, the report publishes with yesterday’s data. How do you remove the timing guess?
Question 33A load task fails now and then because the source API times out for a few seconds. The run succeeds if you simply start it again. What setting suits this?
Question 34In Airflow 3, which concept replaced execution_date in the task context?
Question 35A task passes a 20 GB extract to the next task through XCom, and the Airflow metadata database slows down. What is the right pattern?
Question 36During a review you find a database password written in clear text inside a Dag file that is stored in Git. What do you do?
Question 37Which data quality dimension asks “is the data fresh enough for its use?”
Question 38A source team renames a column without notice. The pipeline still runs green, but the revenue on the dashboard becomes empty. Which control would have caught it first?
Question 39What does data lineage let you see?
Question 40The cloud bill doubles. You find a large development warehouse that ran all weekend with nobody using it, and no cost tags on it. What should you set up?
Question 41A customer in the EU asks the company to delete their personal data. Your platform has a raw layer, marts and backups. What makes the request manageable?
Question 42Which GDPR principle says you should collect and keep only the personal data you need for the stated purpose?