Interview scorecard for data analysts, with anchors for each competency
On this page
- Which competencies should a data analyst scorecard cover?
- What does the scorecard look like?
- What do the anchors sound like for each competency?
- Which work samples produce the evidence?
- What does a filled scorecard look like?
- How do you adjust the scorecard by level and type of analyst?
- What mistakes are specific to scoring data analysts?
- Where does the scorecard fit in the hiring process?
- Questions people ask
An interview scorecard for data analysts should score six things: turning a vague request into an answerable question, SQL and data handling, analytical reasoning, skepticism about data quality, communicating findings, and rigor in checking and documenting work. Below is a copy-ready scorecard with example weights, what a 1, a 3 and a 5 look like for each competency, two short work samples, and a filled example. The anchors reward a verified answer to the right question over a fast query or a long list of tools.
The levels follow the 1-5 interview rating scale, where 3 is a real pass at the level being hired and 5 is rare. For how weights and must-haves work on any scorecard, see the interview scorecard template. The example weights below match the ones in the data analyst interview guide, so the guide tells you which stage owns each row and this page tells you how to score it.
Which competencies should a data analyst scorecard cover?
- Problem framing and requirements. Finds the decision behind a request, defines the terms ("active", "top", "last month") and agrees the scope before pulling data.
- SQL and data handling. Joins, aggregates and filters correctly, and knows what each join does to the row count.
- Analytical reasoning. Picks a fair comparison, separates correlation from cause, and knows when a difference is too small to act on.
- Data quality skepticism. Checks totals against a known source, notices duplicates, gaps and definition changes, and says so.
- Communicating findings. States the finding in one sentence, shows it in a chart a non-analyst can read, and gives a recommendation with its caveats.
- Rigor and reproducibility. Leaves work someone else can rerun, gets queries reviewed, and owns an error when one surfaces.
What does the scorecard look like?
Example weights for a mid-level analyst supporting business teams. Set your own at intake and lock them before the first interview.
| Competency | Example weight | 1 looks like | 3 looks like | 5 looks like |
|---|---|---|---|---|
| Problem framing and requirements | 15% | Starts querying without asking what the result is for. | Asks what decision the request supports and pins down the key definitions. | Reframed a request so the answer changed a decision the requester had not planned to make, and can name it. |
| SQL and data handling | 20% | Query runs but answers a different question, and the candidate does not notice. | Correct joins and aggregation, with a row-count or total check after each step. | Spots a subtle join or grain problem in someone else's query and fixes it cleanly. |
| Analytical reasoning | 20% | Reads any difference as a cause. | Uses a fair comparison and names at least one other explanation. | Designs the comparison that would settle the question, and says what result would prove them wrong. |
| Data quality skepticism | 15% | Trusts the first number the data returns. | Checks the result against a known figure and flags a problem they found. | Traced a long-standing reporting error to its source and got it fixed upstream. |
| Communicating findings | 20% | Walks through every step before saying what they found. | Finding first, one clear chart, a recommendation with caveats. | A non-analyst can repeat the finding and its limits correctly after one pass. |
| Rigor and reproducibility | 10% | Work lives in an unsaved notebook; no review. | Saves and documents queries and has had work reviewed. | Found and disclosed their own published error, and changed how the team checks work. |
A copy-ready version for each interviewer:
Candidate: [name] Role: [data analyst, team or domain, level]
Interviewer: [name] Stage: [screen / SQL / analysis case / stakeholder / hiring manager]
Problem framing and requirements (15%) Score 1-5: [ ]
Evidence: [questions asked, definitions agreed]
SQL and data handling (20%) Score 1-5: [ ]
Evidence: [correctness, checks after joins, grain]
Analytical reasoning (20%) Score 1-5: [ ]
Evidence: [comparison used, other explanations named]
Data quality skepticism (15%) Score 1-5: [ ]
Evidence: [what they checked, what they flagged]
Communicating findings (20%) Score 1-5: [ ]
Evidence: [first sentence, chart, recommendation, caveats]
Rigor and reproducibility (10%) Score 1-5: [ ]
Evidence: [documentation, review, an error they owned]
Weighted score: [ ] Recommendation: [Strong yes / Yes / No / Strong no]
Open question for the next interviewer: [ ]
What do the anchors sound like for each competency?
The table gives 1, 3 and 5. These are the questions to ask and the in-between levels that are hardest to call.
Problem framing and requirements
Ask: "Tell me about a request where what the person asked for was not what they needed."
A 2: clarified definitions but never asked what the result was for. A 4: narrowed a large request to the one number that mattered and delivered it sooner.
SQL and data handling
Ask: use the query review below, plus the live session in the interview guide if you run one.
A 2: right idea, but a join changes the row count and they only find it when prompted. A 4: correct and checked, and they explain the grain of each table before joining.
Analytical reasoning
Ask: "Tell me about an analysis where the obvious explanation turned out to be wrong."
A 2: compared before and after with no thought for seasonality or a mix change. A 4: ruled out a named alternative with a second cut of the data.
Data quality skepticism
Ask: "When did the data turn out to be wrong? How did you notice, and who did you tell?"
A 2: found the problem after a stakeholder did. A 4: caught it with a routine check before anything was published.
Communicating findings
Ask: "Tell me the finding from your last analysis in one sentence, then what you recommended."
A 2: accurate, but the recommendation is missing or hedged into nothing. A 4: the recommendation is clear and the caveat is stated in plain words.
Rigor and reproducibility
Ask: "If you left tomorrow, could someone rerun your last analysis? How?"
A 2: the code is saved but undocumented. A 4: queries are versioned or reviewed and the candidate describes a mistake review caught.
Which work samples produce the evidence?
The interview guide covers a live SQL session and a capped analysis case. These two shorter samples test reading SQL and handling a stakeholder, and fit a screen or a hiring manager round. Use invented data and the same version for every candidate.
Query review (15 minutes). Show this query, written by an invented colleague to answer "How many customers do we have in each region, and how many ordered this year?"
SELECT c.region, COUNT(*) AS customers
FROM customers c
LEFT JOIN orders o ON o.customer_id = c.customer_id
WHERE o.order_date >= '2026-01-01'
GROUP BY c.region;
What to score: SQL and data handling and data quality skepticism. A 3 finds both planted problems: the date filter in the WHERE clause drops customers with no orders this year, so the left join behaves like an inner join; and COUNT(*) counts orders rather than customers. A 5 also writes the fix, for example counting distinct customer IDs and moving the date condition into the join or into a conditional count, and asks what "customer" means (all accounts, or active ones).
Stakeholder intake (10 minutes). The interviewer plays a sales director: "Can you build me a dashboard with everything about our top customers by Friday?" The candidate has ten minutes to agree what they will deliver.
What to score: problem framing (do they ask what decision it supports and how "top" is defined?) and communicating findings (do they end with a clear statement of what will be delivered, when, and what will not be in the first version?).
What does a filled scorecard look like?
A made-up candidate for a mid-level analyst on a marketing team, after a screen, the query review and the stakeholder intake:
Problem framing and requirements: 3. In the intake, asked what the dashboard would be used for and agreed that "top" meant revenue in the last twelve months.
SQL and data handling: 4. Found both problems in the query review within five minutes and rewrote it with a conditional distinct count.
Analytical reasoning: 3. Described checking whether a lift in sign-ups came from one paid channel before crediting a site change.
Data quality skepticism: 4. Described catching duplicated web events by comparing daily totals with the billing system, before a monthly report went out.
Communicating findings: 2. Explained an old analysis step by step and reached the recommendation only when asked.
Rigor and reproducibility: 3. Keeps queries in a shared repository and has had them reviewed by a senior analyst.
Weighted score (example): 3 x 0.15 + 4 x 0.20 + 3 x 0.20 + 4 x 0.15 + 2 x 0.20 + 3 x 0.10 = 3.15. Recommendation: Yes, if the analysis case confirms they can lead with the finding; that is the question for the next interviewer.
How do you adjust the scorecard by level and type of analyst?
| Seat | Adjustment |
|---|---|
| Junior or first analyst job | Smaller schema and narrower questions; accept evidence from coursework or projects. Weight SQL and quality checks over stakeholder work. |
| Senior analyst | A 3 on framing needs a request they reshaped with a senior stakeholder; add mentoring or reviewing others' work. |
| Reporting and BI | Raise SQL and rigor; add a row for dashboard design scored on whether users can answer their question from it. |
| Product or experimentation | Raise analytical reasoning; ask about a test they analyzed, including one with an unclear result. |
| Finance or operations analyst | Add reconciliation to the quality row; the known figure to check against is usually the ledger or the system of record. |
What mistakes are specific to scoring data analysts?
| Mistake | Why it fails | Fix |
|---|---|---|
| Scoring tool names | A list of tools says nothing about whether the numbers were right. | Score the checks they ran and the problems they caught. |
| Rewarding speed over verification | A fast query that double-counts is worse than a slow one that is right. | Make the row-count or total check part of the 3 anchor. |
| Penalizing syntax lookups | It tests memory, not the job. | Allow references; score correctness and reasoning. |
| Using real company data in a take-home | It is unpaid work and may expose data you should not share. | Invent the dataset; cap the time in writing. |
| Scoring statistical vocabulary | Terms can be recited; judgment about a fair comparison cannot. | Ask what else could explain the result. |
Where does the scorecard fit in the hiring process?
The data analyst phone screen sorts out which kind of analyst the candidate is and how deep their SQL goes before the technical rounds, and the rows here should match the must-haves in your data analyst job description. After the hire, the same competencies become the check-ins in the 30-60-90 day plan for data analysts. For more questions, see interview questions for analytical skills and interview questions for attention to detail; for take-homes, see how to evaluate a take-home assignment.
Data analysis has no license to verify, and vendor certificates are not scored. Score only the essential functions the job description lists, and allow reasonable accommodation in timed SQL tasks, which the ADA regulations define to include changes to the application process (29 CFR 1630.2(o)). Each interviewer submits scores before the debrief and before reading anyone else's.
If you run the interview in Interview Signal (we make it), you can enter these six competencies as the scorecard areas of the role's kit; the app scores each area from 1 to 10 during the call and keeps the candidate's own quotes as evidence, deleting any quote that is not in the transcript word for word.
Questions people ask
What should a data analyst interview scorecard include?
Score turning a vague request into an answerable question, SQL and data handling, analytical reasoning, skepticism about data quality, communicating findings, and rigor in checking and documenting work. Give each a weight and anchors for a 1, a 3 and a 5, and score from what the candidate did in a SQL task or analysis case as well as what they described.
Should a data analyst be allowed to look up SQL syntax during the interview?
Yes. Analysts look up syntax every day, so penalizing it measures memory rather than skill. Score whether the query answers the question that was asked, whether the candidate checks row counts and totals, and whether they notice problems in the data.
How do you score a data analyst take-home assignment?
Use invented data, cap the time in writing, give every candidate the same brief, and score it on the same anchors as the live interview. Then discuss it live, because questions such as what else could explain the result show the reasoning that a finished slide deck can hide.
How do you score a junior data analyst with no work experience?
Keep the anchors and shrink the scope: a smaller schema, a narrower question and evidence from coursework, projects or a previous non-analyst job where they used data. A junior candidate who checks their own row counts and says what they are unsure of is showing the habits the scorecard rewards.