Interview rubric
How SQL Interviews Are Scored: Rubric by Level (2026)
A sql interview is scored on 6 dimensions: query correctness (30%), problem decomposition (15%), edge cases and data quality (15%), query efficiency (10%), business interpretation (15%), communication (15%). Each is rated 1 to 4 against anchors that rise with the level. At mid level, an outstanding score on query correctness means: "Writes correct, compact queries and explains why each join type and window frame is right."
The dimensions and their weights
| Dimension | Weight | Solid (3) at mid level |
|---|---|---|
| Query correctness | 30% | Writes correct joins, grouping and window functions and handles NULLs right. |
| Problem decomposition | 15% | Plans the steps unprompted and builds the query in named, readable CTEs. |
| Edge cases and data quality | 15% | Raises NULLs, duplicates and time zones unprompted and handles them in the query. |
| Business interpretation | 15% | Explains what the numbers mean, the caveats and what to check next, unprompted. |
| Communication | 15% | Talks through the plan before writing and explains why each choice was made. |
| Query efficiency | 10% | Avoids needless scans and explains which indexes or partitions the query relies on. |
What changes between levels
On query correctness, the heaviest dimension, an outstanding mid-level answer "writes correct, compact queries and explains why each join type and window frame is right", while an outstanding senior answer "writes correct queries for tricky cases and offers an alternative formulation with trade-offs". The same answer can score 4 at mid level and 2 at senior: senior candidates are expected to raise things unprompted that mid-level candidates may be asked about.
Every score, at every level
Query correctness (30%)
| Level | 1 · Poor | 2 · Borderline | 3 · Solid | 4 · Outstanding |
|---|---|---|---|---|
| Intern / new grad | Writes a query that does not run or answers a different question. | Gets a basic SELECT with filters right only after several corrections. | Writes correct SELECT, JOIN and GROUP BY queries with a few fixes after hints. | Writes correct joins and aggregations unprompted, including a HAVING or subquery. |
| Junior | Writes a query that does not run or answers a different question. | Joins and groups correctly only with major interviewer help. | Writes correct joins and aggregations; one window function needs a hint. | Writes correct joins, aggregations and window functions with no help. |
| Mid-level | Writes a query that does not run or answers a different question. | Query runs but returns wrong results, such as a fan-out join, until corrected. | Writes correct joins, grouping and window functions and handles NULLs right. | Writes correct, compact queries and explains why each join type and window frame is right. |
| Senior | Writes a query that does not run or answers a different question. | Mostly correct but misses a subtle error such as join fan-out or a wrong window frame. | Writes correct queries for tricky cases such as gaps and islands or retention, unaided. | Writes correct queries for tricky cases and offers an alternative formulation with trade-offs. |
| Staff | Writes a query that does not run or answers a different question. | Mostly correct but misses a subtle error such as join fan-out or a wrong window frame. | Writes correct queries for tricky cases unaided and checks results against expectations. | Writes correct queries and designs reusable models or views so others get it right too. |
Problem decomposition (15%)
| Level | 1 · Poor | 2 · Borderline | 3 · Solid | 4 · Outstanding |
|---|---|---|---|---|
| Intern / new grad | Tries to write one large query with no plan and gets lost. | Breaks the problem into steps only when the interviewer suggests them. | Describes the steps in words when nudged, then writes the query step by step. | Plans the steps unprompted and builds the query one step at a time. |
| Junior | Tries to write one large query with no plan and gets lost. | Plans steps only when asked and writes nested queries that are hard to follow. | Plans the steps and builds the query in readable parts such as CTEs. | Plans the steps unprompted and checks each intermediate result before moving on. |
| Mid-level | Tries to write one large query with no plan and gets lost. | Builds the query in parts but the parts are tangled or repeated. | Plans the steps unprompted and builds the query in named, readable CTEs. | Decomposes cleanly and checks each intermediate result before combining. |
| Senior | Tries to write one large query with no plan and gets lost. | Readable steps, but the plan needs reworking when the question changes. | Decomposes into clear CTEs that adapt easily when a follow-up changes the question. | Decomposes so each step is testable and reusable, and explains why. |
| Staff | Tries to write one large query with no plan and gets lost. | Readable steps, but the plan needs reworking when the question changes. | Decomposes into testable steps and names which ones belong in a shared data model. | Decomposes the problem into a reusable pipeline design and explains how teams would share it. |
Edge cases and data quality (15%)
| Level | 1 · Poor | 2 · Borderline | 3 · Solid | 4 · Outstanding |
|---|---|---|---|---|
| Intern / new grad | Ignores missing or duplicate data even when it breaks the answer. | Handles NULLs or duplicates only after the interviewer points them out. | Asks about NULLs or duplicates when prompted and handles them in the query. | Raises NULLs and duplicates unprompted and handles them in the query. |
| Junior | Ignores missing or duplicate data even when it breaks the answer. | Handles data issues only after the interviewer points them out. | Handles NULLs and duplicates when prompted and checks the result for sense. | Raises NULLs, duplicates and empty groups unprompted and handles each one. |
| Mid-level | Ignores missing or duplicate data even when it breaks the answer. | Handles NULLs but misses duplicates or time zones until prompted. | Raises NULLs, duplicates and time zones unprompted and handles them in the query. | Also raises late-arriving data or bad source rows and proposes checks for them. |
| Senior | Ignores missing or duplicate data even when it breaks the answer. | Handles common issues but misses late data or definition ambiguity until prompted. | Handles common issues unprompted and clarifies ambiguous definitions such as active user. | Proposes data-quality checks that would catch these issues before analysis. |
| Staff | Ignores missing or duplicate data even when it breaks the answer. | Handles issues in the query but never asks where they come from. | Handles issues unprompted and traces them to source systems to fix upstream. | Proposes data contracts or monitoring that prevent these issues across teams. |
Query efficiency (10%)
| Level | 1 · Poor | 2 · Borderline | 3 · Solid | 4 · Outstanding |
|---|---|---|---|---|
| Intern / new grad | Cannot say anything about how the query would perform. | Mentions performance only when asked, with a wrong claim. | Says when asked that filtering early or an index would make the query faster. | Avoids an obviously wasteful step unprompted and mentions indexes. |
| Junior | Writes a needlessly expensive query and cannot say why when asked. | Mentions indexes vaguely when asked but cannot apply the idea. | Avoids SELECT * and needless subqueries and explains a useful index when asked. | Explains unprompted which index or filter keeps the query cheap. |
| Mid-level | Writes a needlessly expensive query and cannot say why when asked. | Notices cost issues only when asked, such as a correlated subquery. | Avoids needless scans and explains which indexes or partitions the query relies on. | Reasons about the query plan and rewrites a slow step with a stated reason. |
| Senior | Writes a needlessly expensive query and cannot say why when asked. | Knows general rules but cannot reason about this query at scale. | Reasons about the plan and cost at the stated data size and chooses accordingly. | Proposes pre-aggregation or partitioning that changes the cost profile, with trade-offs. |
| Staff | Writes a needlessly expensive query and cannot say why when asked. | Knows general rules but cannot reason about this query at scale. | Reasons about cost at scale and names storage or modelling changes that would help. | Weighs compute cost, freshness and maintenance across the warehouse and recommends a design. |
Business interpretation (15%)
| Level | 1 · Poor | 2 · Borderline | 3 · Solid | 4 · Outstanding |
|---|---|---|---|---|
| Intern / new grad | Cannot say what the result means. | Reads the numbers back without saying what they mean. | Explains what the result means in one or two plain sentences when asked. | Explains what the result means unprompted and names one thing to check next. |
| Junior | Cannot say what the result means. | Reads the numbers back without saying what they mean. | Explains what the result means and one caveat about it. | Explains the meaning unprompted, a caveat, and a sensible next question to ask. |
| Mid-level | Cannot say what the result means, or draws a wrong conclusion. | Interprets the result only when asked and ignores caveats. | Explains what the numbers mean, the caveats and what to check next, unprompted. | Turns the result into a recommendation and names what data would change it. |
| Senior | Cannot say what the result means, or draws a wrong conclusion. | Interprets correctly but stops short of a recommendation. | Turns the result into a recommendation, with caveats such as correlation versus causation. | Frames the result as a decision for stakeholders and proposes how to test the causal claim. |
| Staff | Cannot say what the result means, or draws a wrong conclusion. | Recommends locally without the wider business context. | Ties the result to business strategy and recommends a decision with its risks. | Uses the result to reshape the question the business should be asking, with evidence. |
Communication (15%)
| Level | 1 · Poor | 2 · Borderline | 3 · Solid | 4 · Outstanding |
|---|---|---|---|---|
| Intern / new grad | Writes in silence and cannot explain the query when asked. | Explains the query only when asked, and unclearly. | Talks through the plan when nudged and explains each part of the query. | Talks through the plan before writing and explains each part unprompted. |
| Junior | Writes in silence and cannot explain the query when asked. | Explains what the query does only when the interviewer asks. | Talks through the plan before writing and explains the query clearly. | Talks through the plan, checks assumptions with the interviewer and explains clearly. |
| Mid-level | Writes in silence and cannot explain the query when asked. | Explains the query but not the reasons behind its choices. | Talks through the plan before writing and explains why each choice was made. | Drives the session, confirms definitions early and adapts quickly to follow-ups. |
| Senior | Writes in silence and cannot explain the query when asked. | Explains choices but needs the interviewer to steer the pacing. | Drives the session, explains choices and manages time across follow-ups. | Explains results to a non-technical listener as clearly as to a technical one. |
| Staff | Writes in silence and cannot explain the query when asked. | Explains when asked but does not lead the discussion. | Leads the discussion, frames trade-offs and explains results for stakeholders. | Leads crisply, invites challenge and tailors depth to the listener. |
Get scored against this rubric
Thita's AI interviewer runs a live sql round and scores you against exactly this rubric, dimension by dimension, for the level you choose.
Practise a SQL roundWhere this rubric comes from
These rubrics are researched: they follow the scoring methods big-tech companies describe publicly (behaviourally anchored ratings, bar raisers, leadership principles, named interview signals) and how their rounds are commonly run. No company publishes its actual scoring sheet, so the level-by-level wording is our own synthesis. It is the same rubric our AI interviewer scores every mock against.
Questions, answered
- How is a sql interview scored?
- A sql interview is scored on 6 dimensions: query correctness (30%), problem decomposition (15%), edge cases and data quality (15%), query efficiency (10%), business interpretation (15%), communication (15%). Each is rated 1 to 4 against anchors that rise with the level. At mid level, an outstanding score on query correctness means: "Writes correct, compact queries and explains why each join type and window frame is right."
- What is the difference between a mid-level and a senior sql answer?
- On query correctness, the heaviest dimension, an outstanding mid-level answer "writes correct, compact queries and explains why each join type and window frame is right", while an outstanding senior answer "writes correct queries for tricky cases and offers an alternative formulation with trade-offs". The same answer can score 4 at mid level and 2 at senior: senior candidates are expected to raise things unprompted that mid-level candidates may be asked about.
- What does a score of 3 mean?
- Scores are relative to the level you are interviewing for. A 3 means solid for that level: the bar is met without help on the key points. A 4 means the candidate also raised the important points unprompted. A 2 usually means the interviewer had to lead the candidate to them.
- Is this the official rubric used by Google, Meta or Amazon?
- These rubrics are researched: they follow the scoring methods big-tech companies describe publicly (behaviourally anchored ratings, bar raisers, leadership principles, named interview signals) and how their rounds are commonly run. No company publishes its actual scoring sheet, so the level-by-level wording is our own synthesis. It is the same rubric our AI interviewer scores every mock against.