ETL & ELT interview questions

1. A.A pattern that loads raw data into the warehouse first and transforms it in-place using warehouse compute, leveraging cheap scalable storage and SQL.
2. B.The stored maximum value of a watermark column from the last successful run, used to bound the next incremental extract.
3. C.A full load replaces the whole target each run; an incremental load processes only new or changed rows since the last run for efficiency.
4. D.An intermediate landing zone where raw extracted data is held before transformation, isolating sources from the warehouse and enabling reprocessing.

Answer + AI explanation with Pro

1. A.staging area
2. B.change data feed
3. C.idempotent load
4. D.reverse ETL

Answer + AI explanation with Pro

1. A.staging area — A pattern that loads raw data into the warehouse first and transforms it in-place using warehouse compute, leveraging cheap scalable storage and SQL.
2. B.staging area — The stored maximum value of a watermark column from the last successful run, used to bound the next incremental extract.
3. C.staging area — An intermediate landing zone where raw extracted data is held before transformation, isolating sources from the warehouse and enabling reprocessing.
4. D.staging area — Moving modeled data from the warehouse back out into operational SaaS tools (CRM, ad platforms) to activate analytics in business workflows.

Answer + AI explanation with Pro

1. A.Re-running a pipeline over historical date ranges to populate or correct past partitions, ideally idempotently so reruns are safe.
2. B.A layered design with bronze (raw), silver (cleaned and conformed), and gold (business-ready aggregates) tables popularized by the lakehouse pattern.
3. C.A pattern that loads raw data into the warehouse first and transforms it in-place using warehouse compute, leveraging cheap scalable storage and SQL.
4. D.A stream of inserts, updates, and deletes emitted by a source so downstream pipelines apply only deltas instead of full reloads.

Answer + AI explanation with Pro

1. A.watermark column
2. B.exactly-once ingestion
3. C.medallion architecture
4. D.slowly changing fact

Answer + AI explanation with Pro

1. A.medallion architecture — A layered design with bronze (raw), silver (cleaned and conformed), and gold (business-ready aggregates) tables popularized by the lakehouse pattern.
2. B.medallion architecture — Re-running a pipeline over historical date ranges to populate or correct past partitions, ideally idempotently so reruns are safe.
3. C.medallion architecture — A correction or restatement of previously loaded fact rows, handled by reversing entries, snapshotting, or restating affected partitions.
4. D.medallion architecture — Ensuring each source record lands once despite retries, achieved with idempotent writes, dedup keys, or transactional offset commits.

Answer + AI explanation with Pro

1. A.A stream of inserts, updates, and deletes emitted by a source so downstream pipelines apply only deltas instead of full reloads.
2. B.A pattern that loads raw data into the warehouse first and transforms it in-place using warehouse compute, leveraging cheap scalable storage and SQL.
3. C.A monotonically increasing column (timestamp or id) tracked between runs so incremental extracts pull only rows newer than the last loaded value.
4. D.Moving modeled data from the warehouse back out into operational SaaS tools (CRM, ad platforms) to activate analytics in business workflows.

Answer + AI explanation with Pro

1. A.reverse ETL
2. B.medallion architecture
3. C.exactly-once ingestion
4. D.backfill

Answer + AI explanation with Pro

1. A.reverse ETL — A monotonically increasing column (timestamp or id) tracked between runs so incremental extracts pull only rows newer than the last loaded value.
2. B.reverse ETL — Moving modeled data from the warehouse back out into operational SaaS tools (CRM, ad platforms) to activate analytics in business workflows.
3. C.reverse ETL — A full load replaces the whole target each run; an incremental load processes only new or changed rows since the last run for efficiency.
4. D.reverse ETL — A layered design with bronze (raw), silver (cleaned and conformed), and gold (business-ready aggregates) tables popularized by the lakehouse pattern.

Answer + AI explanation with Pro

1. A.A load designed so rerunning it for the same window yields the same final state, typically via delete-insert or merge keyed on a partition.
2. B.Re-running a pipeline over historical date ranges to populate or correct past partitions, ideally idempotently so reruns are safe.
3. C.A pattern that loads raw data into the warehouse first and transforms it in-place using warehouse compute, leveraging cheap scalable storage and SQL.
4. D.The stored maximum value of a watermark column from the last successful run, used to bound the next incremental extract.

Answer + AI explanation with Pro

1. A.staging area
2. B.watermark column
3. C.ELT
4. D.backfill

Answer + AI explanation with Pro

1. A.ELT — An intermediate landing zone where raw extracted data is held before transformation, isolating sources from the warehouse and enabling reprocessing.
2. B.ELT — Moving modeled data from the warehouse back out into operational SaaS tools (CRM, ad platforms) to activate analytics in business workflows.
3. C.ELT — A pattern that loads raw data into the warehouse first and transforms it in-place using warehouse compute, leveraging cheap scalable storage and SQL.
4. D.ELT — A load designed so rerunning it for the same window yields the same final state, typically via delete-insert or merge keyed on a partition.

Answer + AI explanation with Pro

1. A.A full load replaces the whole target each run; an incremental load processes only new or changed rows since the last run for efficiency.
2. B.A stream of inserts, updates, and deletes emitted by a source so downstream pipelines apply only deltas instead of full reloads.
3. C.Re-running a pipeline over historical date ranges to populate or correct past partitions, ideally idempotently so reruns are safe.
4. D.A correction or restatement of previously loaded fact rows, handled by reversing entries, snapshotting, or restating affected partitions.

Answer + AI explanation with Pro

1. A.slowly changing fact
2. B.ELT
3. C.change data feed
4. D.staging area

Answer + AI explanation with Pro

1. A.change data feed — A full load replaces the whole target each run; an incremental load processes only new or changed rows since the last run for efficiency.
2. B.change data feed — A layered design with bronze (raw), silver (cleaned and conformed), and gold (business-ready aggregates) tables popularized by the lakehouse pattern.
3. C.change data feed — A stream of inserts, updates, and deletes emitted by a source so downstream pipelines apply only deltas instead of full reloads.
4. D.change data feed — A load designed so rerunning it for the same window yields the same final state, typically via delete-insert or merge keyed on a partition.

Answer + AI explanation with Pro

1. A.Re-running a pipeline over historical date ranges to populate or correct past partitions, ideally idempotently so reruns are safe.
2. B.Moving modeled data from the warehouse back out into operational SaaS tools (CRM, ad platforms) to activate analytics in business workflows.
3. C.An intermediate landing zone where raw extracted data is held before transformation, isolating sources from the warehouse and enabling reprocessing.
4. D.A pattern that loads raw data into the warehouse first and transforms it in-place using warehouse compute, leveraging cheap scalable storage and SQL.

Answer + AI explanation with Pro

1. A.exactly-once ingestion
2. B.high-water mark
3. C.backfill
4. D.idempotent load

Answer + AI explanation with Pro

1. A.backfill — A pattern that loads raw data into the warehouse first and transforms it in-place using warehouse compute, leveraging cheap scalable storage and SQL.
2. B.backfill — Ensuring each source record lands once despite retries, achieved with idempotent writes, dedup keys, or transactional offset commits.
3. C.backfill — A load designed so rerunning it for the same window yields the same final state, typically via delete-insert or merge keyed on a partition.
4. D.backfill — Re-running a pipeline over historical date ranges to populate or correct past partitions, ideally idempotently so reruns are safe.

Answer + AI explanation with Pro

1. A.A monotonically increasing column (timestamp or id) tracked between runs so incremental extracts pull only rows newer than the last loaded value.
2. B.A layered design with bronze (raw), silver (cleaned and conformed), and gold (business-ready aggregates) tables popularized by the lakehouse pattern.
3. C.Re-running a pipeline over historical date ranges to populate or correct past partitions, ideally idempotently so reruns are safe.
4. D.Moving modeled data from the warehouse back out into operational SaaS tools (CRM, ad platforms) to activate analytics in business workflows.

Answer + AI explanation with Pro

1. A.medallion architecture
2. B.full load vs incremental
3. C.watermark column
4. D.exactly-once ingestion

Answer + AI explanation with Pro

1. A.watermark column — An incremental pattern that deletes the target rows for a key range then inserts the recomputed rows, giving idempotent partition replacement.
2. B.watermark column — A load designed so rerunning it for the same window yields the same final state, typically via delete-insert or merge keyed on a partition.
3. C.watermark column — A monotonically increasing column (timestamp or id) tracked between runs so incremental extracts pull only rows newer than the last loaded value.
4. D.watermark column — A stream of inserts, updates, and deletes emitted by a source so downstream pipelines apply only deltas instead of full reloads.

Answer + AI explanation with Pro

1. A.A correction or restatement of previously loaded fact rows, handled by reversing entries, snapshotting, or restating affected partitions.
2. B.A load designed so rerunning it for the same window yields the same final state, typically via delete-insert or merge keyed on a partition.
3. C.Moving modeled data from the warehouse back out into operational SaaS tools (CRM, ad platforms) to activate analytics in business workflows.
4. D.Re-running a pipeline over historical date ranges to populate or correct past partitions, ideally idempotently so reruns are safe.

Answer + AI explanation with Pro

1. A.slowly changing fact
2. B.watermark column
3. C.change data feed
4. D.backfill

Answer + AI explanation with Pro

1. A.slowly changing fact — Moving modeled data from the warehouse back out into operational SaaS tools (CRM, ad platforms) to activate analytics in business workflows.
2. B.slowly changing fact — A correction or restatement of previously loaded fact rows, handled by reversing entries, snapshotting, or restating affected partitions.
3. C.slowly changing fact — Re-running a pipeline over historical date ranges to populate or correct past partitions, ideally idempotently so reruns are safe.
4. D.slowly changing fact — A full load replaces the whole target each run; an incremental load processes only new or changed rows since the last run for efficiency.

Answer + AI explanation with Pro

1. A.A pattern that loads raw data into the warehouse first and transforms it in-place using warehouse compute, leveraging cheap scalable storage and SQL.
2. B.A full load replaces the whole target each run; an incremental load processes only new or changed rows since the last run for efficiency.
3. C.A load designed so rerunning it for the same window yields the same final state, typically via delete-insert or merge keyed on a partition.
4. D.Moving modeled data from the warehouse back out into operational SaaS tools (CRM, ad platforms) to activate analytics in business workflows.

Answer + AI explanation with Pro

1. A.idempotent load
2. B.backfill
3. C.full load vs incremental
4. D.ELT

Answer + AI explanation with Pro

1. A.idempotent load — A load designed so rerunning it for the same window yields the same final state, typically via delete-insert or merge keyed on a partition.
2. B.idempotent load — Moving modeled data from the warehouse back out into operational SaaS tools (CRM, ad platforms) to activate analytics in business workflows.
3. C.idempotent load — Re-running a pipeline over historical date ranges to populate or correct past partitions, ideally idempotently so reruns are safe.
4. D.idempotent load — A pattern that loads raw data into the warehouse first and transforms it in-place using warehouse compute, leveraging cheap scalable storage and SQL.

Answer + AI explanation with Pro

1. A.A load designed so rerunning it for the same window yields the same final state, typically via delete-insert or merge keyed on a partition.
2. B.A stream of inserts, updates, and deletes emitted by a source so downstream pipelines apply only deltas instead of full reloads.
3. C.An incremental pattern that deletes the target rows for a key range then inserts the recomputed rows, giving idempotent partition replacement.
4. D.A monotonically increasing column (timestamp or id) tracked between runs so incremental extracts pull only rows newer than the last loaded value.

Answer + AI explanation with Pro

1. A.high-water mark
2. B.exactly-once ingestion
3. C.delete-insert strategy
4. D.full load vs incremental

Answer + AI explanation with Pro

1. A.delete-insert strategy — An incremental pattern that deletes the target rows for a key range then inserts the recomputed rows, giving idempotent partition replacement.
2. B.delete-insert strategy — Re-running a pipeline over historical date ranges to populate or correct past partitions, ideally idempotently so reruns are safe.
3. C.delete-insert strategy — A layered design with bronze (raw), silver (cleaned and conformed), and gold (business-ready aggregates) tables popularized by the lakehouse pattern.
4. D.delete-insert strategy — A load designed so rerunning it for the same window yields the same final state, typically via delete-insert or merge keyed on a partition.

Answer + AI explanation with Pro