Interview scorecard
Data Engineer
Score each skill from 1 to 5. A 1 is below the bar for this role, a 3 meets it, and a 5 is above it. The examples show what each sounds like. Use 2 and 4 for answers that fall in between.
Data pipelines
Builds and runs jobs that move data reliably from sources to where people use it.
Questions
- Tell me about a pipeline you built. Where did the data come from and go?
- Tell me about a time a pipeline failed in production. How did you handle it?
- How did you make a pipeline easy to rerun or fix after a failure?
1: Below the bar
Describes only a script or tool they used. Can't explain what happens when a run fails or repeats.
3: Meets the bar
Explains the flow from source to destination and fixes failures once alerted. Has some retry or rerun approach.
5: Above the bar
Designs for failure up front, with safe reruns, alerts and monitoring. Explains how they found root causes and prevented repeats.
Data modeling
Designs table structures that match how the business asks questions and keep data consistent over time.
Questions
- Tell me about a data model you designed. What business questions did it need to answer?
- Describe a time a model had to change after launch. What drove the change?
- Tell me about a trade-off you made between a clean model and fast queries.
1: Below the bar
Describes tables without tying them to business questions. Can't explain why the structure was chosen or how it changed.
3: Meets the bar
Links the model to main business questions and explains key choices. Handles change reactively and names one trade-off.
5: Above the bar
Starts from how people will query the data, plans for change, and explains trade-offs with clear examples of the results.
SQL
Writes correct, readable queries and fixes slow ones on large, messy datasets.
Questions
- Tell me about a complex query you wrote recently. How did you check its results?
- Tell me about a time a query ran too slowly. What did you do?
- Walk me through how you'd combine three tables where some rows have no match.
1: Below the bar
Writes only simple queries. Can't explain joins or why results look wrong, and has no way to check them.
3: Meets the bar
Writes correct joins and aggregations, and checks results against known numbers. Speeds up slow queries with some guidance.
5: Above the bar
Uses window functions and clean structure, checks results unprompted, and finds slow spots by reading the query plan.
Data quality
Detects, prevents, and fixes bad data before it reaches the people who rely on it.
Questions
- Tell me about a time bad data reached users. How did you find out?
- Describe checks you added to catch data problems early. Why those checks?
- Tell me about a time you disagreed with a data owner about what counted as correct.
1: Below the bar
Relies on users to report errors. Fixes the single bad record and adds no checks to prevent a repeat.
3: Meets the bar
Adds basic checks like nulls and duplicates after an issue. Traces the cause and fixes it, with some prompting.
5: Above the bar
Sets checks before problems appear, ranks them by business impact, and agrees correctness rules with owners. Tracks issues to the root cause.
System design
Plans how data systems fit together and weighs trade-offs between cost, speed, and reliability.
Questions
- Tell me about a data platform you designed. What did you ask before starting?
- What would break first if your data volume grew tenfold, and why?
- Tell me about a design choice you later changed. What made you change it?
1: Below the bar
Starts with tools before clarifying needs. Cannot explain why one design beats another or where it might fail.
3: Meets the bar
Describes a working platform and why its parts were chosen. Names cost and speed trade-offs but not failure handling.
5: Above the bar
Starts from who uses the data and how fresh it must be, compares options on cost, and plans for failed loads.
Overall recommendation
Decide on your own, before you compare notes with the rest of the panel.