How to assess · For hiring teams
How to Assess ETL Skills When Hiring
The test formats that actually work for ETL, what a strong answer looks like, sample questions and a scoring rubric you can use as-is.
The short answer
Assess ETL with a task, not a conversation: design a pipeline against an awkward source, fix a lossy job, ai-scored assessment (e.g. cohesyve) or schema-change conversation. Score it against written criteria you fix before you see any submissions, and weight the criteria that the role actually depends on.
- Extracts incrementally with watermarks and handles pagination, rate limits and timeouts without losing records
- Writes transformations that are deterministic, testable and readable
- Loads idempotently, so retries and backfills never duplicate
- Reconciles counts and totals against the source and alerts on discrepancy
Paste a job description; Cohesyve generates a role-specific assessment and rubric. Ten candidates free, no card.
ETL is unglamorous and consequential: it is the work that decides whether the numbers in every downstream system are complete, current and correct. The skill is not knowing a tool. It is designing extraction that does not miss records, transformation that is testable and explainable, and loading that can be retried, backfilled and audited. This page covers how to assess ETL for data engineering and integration roles: extraction reliability, transformation correctness, load safety, and the operational habits that keep data trustworthy.
Why ETL is worth testing
ETL failures are silent. A job that misses records under an API timeout, a transformation that rounds where it should not, a load that duplicates on retry — none raise an error, and all corrupt the data people rely on. Testing shows whether a candidate designs for the failures that actually occur, and it is far cheaper than discovering them through a wrong quarterly number.
What strong ETL looks like
- Extracts incrementally with watermarks and handles pagination, rate limits and timeouts without losing records
- Writes transformations that are deterministic, testable and readable
- Loads idempotently, so retries and backfills never duplicate
- Reconciles counts and totals against the source and alerts on discrepancy
- Handles schema change in the source without silent breakage
- Logs enough to answer "what ran, on what, and what happened" for any run
- Knows when ELT into a warehouse beats transforming in flight
Ways to assess ETL
Design a pipeline against an awkward source
Describe a paginated API with rate limits and occasional timeouts that must be loaded daily into a warehouse with exact record counts. Ask for the design, failure handling and reconciliation.
Pros
Cons
Best for Mid and senior roles.
Fix a lossy job
Provide a small pipeline that drops records on timeout and duplicates on retry. Ask the candidate to find and fix both.
Pros
Cons
Best for Any level.
AI-scored assessment (e.g. Cohesyve)
Generate a ETL task from the job description — a pipeline design, a data-loss diagnosis, a reconciliation plan — with a rubric. Each candidate receives a different variant; the reasoning is scored alongside the work.
Pros
Cons
Best for Screening an applicant pool fairly before interview time is spent.
Schema-change conversation
Ask what happens when the source adds, renames or drops a column, and how their pipeline responds.
Pros
Cons
Best for Senior engineers.
Cohesyve
Run a ETL assessment on your next opening
Cohesyve generates a unique ETL task per candidate from your job description, with the scoring rubric attached. Questions are different for every applicant, so they cannot be shared or looked up.
What to test
Extraction
Whether every record arrives.
Transformation
Whether logic is correct and testable.
Loading
Whether loads are safe.
Operations
Whether the pipeline is trustworthy.
Sample ETL questions
What is a watermark in incremental extraction, and what goes wrong without one?
EntryLook for A marker of the last extracted point; without it, full reloads or missed records.
A job retried after a failure and now the table has duplicates. What was missing?
EntryLook for Idempotent load — merge on key or delete-and-insert by partition.
An API times out on page 40 of 100. How does your pipeline behave?
MidLook for Retries with backoff, resumes from the last successful page, does not mark the run complete, reconciles counts.
How do you know today's load is complete and correct?
MidLook for Row count and total reconciliation against source, freshness check, anomaly detection on volume.
The source renames a column overnight. What happens and what should happen?
SeniorLook for Fail loudly rather than load nulls; schema checks at extraction; alert and a documented process.
Red flags
- Full reloads every run because incremental is "hard"
- No idea whether yesterday's load was complete
- Retries that duplicate
- Transformations with no tests
- Silent null-filling on schema change
Scoring rubric
| Criterion | Weight | What strong looks like |
|---|---|---|
| Extraction reliability | 30% | No records lost under real conditions. |
| Load safety | 25% | Idempotent, backfillable. |
| Transformation quality | 20% | Deterministic, tested, readable. |
| Operations | 25% | Reconciled, monitored, resilient to change. |
Mistakes hiring teams make
- Testing tool syntax instead of failure handling
- Not asking how completeness is verified
- Accepting a pipeline with no retry story
- Skipping schema change
- Assuming ETL tool certification equals engineering judgement
Roles that need ETL
Common questions
ETL or ELT — does it change the assessment?
The failure modes are the same: lost records, duplicated loads, silent breakage. Test those regardless of where the transformation runs.
What is the best single ETL question?
Ask how they know a load was complete and correct. Reconciliation habit separates engineers from job-runners.
Should I test a specific ETL tool?
Only if the role is locked to it. Reliability design transfers; tool syntax does not matter much.
How long should an ETL assessment take?
Sixty minutes for a design exercise; two hours capped for a small hands-on pipeline.
Cohesyve · Skill assessments for hiring
Test ETL before the first interview
Generate a role-specific ETL assessment from your job description and see who can do the work before you spend interview time on them.
1,500+
assessments completed
50%
faster time-to-hire
90%
completion rate
5 min
from JD to assessment
No credit card · 10 free candidates · Plans sized to your hiring volume
From the blog