Mettl SQL Test Questions and Answers: Worked Examples and Prep Guide 2026

Mettl SQL test questions and answers cover SELECT queries, joins, GROUP BY with HAVING, subqueries, window functions, and data-cleaning tasks, delivered in a timed, proctored Mercer | Mettl session. The hiring company sets the question count and cutoff, so no single public answer key exists. Prepare by practicing query patterns, not memorizing answers.

Mettl SQL Test Questions and Answers: Key Facts (Updated October 2026)

Most people searching for mettl sql test questions and answers want a leaked answer sheet. That does not exist in any reliable form, because Mercer | Mettl builds each assessment per employer and draws items from a question library. What does exist is a predictable set of patterns. If you can write a join, aggregate with GROUP BY, filter groups with HAVING, and rank rows with a window function, you can solve nearly everything these tests ask.

Before studying queries, run a system check. The msb demo test lets you confirm your browser, webcam, and connection work with the Mettl lockdown environment, which removes a surprising source of failed attempts. A proctored assessment needs Chrome 63+, Firefox 52+, or Safari 12+ on a computer, so a phone or Edge will not work. Spending ten minutes on this check protects every hour of SQL practice you do afterward.

Candidates usually meet two question types. Multiple-choice items ask what a query returns, which keyword is missing, or which index speeds a lookup. Coding items give a schema, such as Employees and Departments tables, and ask you to write a query that passes hidden test cases. Both are scored automatically, so output must match exactly, including column order and sorting. A correct idea with the wrong column alias can still fail a case.

Because the sponsor controls the settings, difficulty varies widely. A screening for an entry-level role may include ten multiple-choice questions and two easy queries in 30 minutes. A data analyst or backend role may add window functions, self-joins, and query optimization over 60 to 90 minutes. Read your invitation email carefully, since it states the duration and sections. Treat anything beyond that as unknown rather than guessing from forum posts.

If your assessment also includes programming, read how mettl scores code, because SQL tasks often sit beside Java or Python problems in the same session. Partial credit generally depends on how many hidden test cases pass, so a query that handles NULLs and duplicates earns more than one that works only on the sample rows. Build the habit of testing edge cases before you click submit.

This guide walks through the question types, worked examples with answers, a realistic format, and a short study plan. Every example uses standard SQL that runs on MySQL, PostgreSQL, and SQL Server with small changes. Where dialects differ, such as TOP versus LIMIT, the text says so. Use the examples as templates: change the table, predict the output, and then verify it in a free sandbox such as SQLite or an online fiddle.

One more point on honesty: any site promising the exact questions from your test is guessing or selling something. Sharing live assessment content likely violates the rules you accept before starting, and it can get your result invalidated. The honest edge is pattern fluency. Practicing 40 to 60 varied queries over two weeks builds the speed that leaves you time to review, and review time is where borderline scores get rescued.

What Are Typical Mettl Test Sample Questions for SQL?

Typical Mettl test sample questions for SQL include finding the second-highest salary, counting employees per department, listing customers with no orders, and removing duplicate rows. Each maps to one core concept: subqueries, aggregation, outer joins, and window functions. Below are four patterns with answers you can adapt to any schema, plus the mistakes that cost points.

Question one: return the second-highest salary from an Employees table. Answer: SELECT MAX(salary) FROM Employees WHERE salary < (SELECT MAX(salary) FROM Employees); The inner query finds the top value and the outer query finds the maximum below it. A window alternative uses DENSE_RANK() OVER (ORDER BY salary DESC) and filters for rank 2, which also handles ties correctly.

Question two: list departments with more than five employees. Answer: SELECT dept_id, COUNT(*) AS headcount FROM Employees GROUP BY dept_id HAVING COUNT(*) > 5; The common trap is using WHERE for the aggregate condition. WHERE filters rows before grouping, while HAVING filters groups after it. Test authors love this distinction because the wrong answer throws a syntax error, which a hidden test case treats as a total failure.

Question three: find customers who never placed an order. Answer: SELECT c.name FROM Customers c LEFT JOIN Orders o ON c.id = o.customer_id WHERE o.id IS NULL; The LEFT JOIN keeps every customer, and unmatched rows carry NULL in the Orders columns. Writing = NULL instead of IS NULL is a classic error, because NULL never equals anything, so that query returns zero rows and fails every test.

Question four: delete duplicate emails while keeping the lowest id. Answer: DELETE FROM Users WHERE id NOT IN (SELECT MIN(id) FROM Users GROUP BY email); On MySQL you may need to wrap the subquery in a derived table, because MySQL blocks selecting from the table you are deleting. On PostgreSQL or SQL Server, a ROW_NUMBER() approach is cleaner. Knowing the quirk shows you have run real queries, not only read about them.

These examples mirror the themes in mettl questions across technical roles, where the SQL portion is often one section of a broader assessment alongside aptitude and logic. Multiple-choice items test the same ideas in reading form: predict the output of a join, spot the missing GROUP BY column, or choose between UNION and UNION ALL. UNION removes duplicates, while UNION ALL keeps everything and runs faster, a frequent one-mark question.

More multiple-choice staples: the difference between DELETE, TRUNCATE, and DROP (selected rows with logging, all rows quickly, the whole table); what a PRIMARY KEY guarantees (unique and not null); and why an index speeds reads but slows writes. Also expect questions on normalization up to third normal form and on ACID properties. Short definitions with one concrete example each are enough to answer these quickly and bank time for the coding items.

Practice tip: for each example, write the query from memory, run it on a small table you build yourself, and then deliberately break it. Remove the GROUP BY column, swap LEFT for INNER, insert a NULL salary. Seeing the output change teaches more than rereading the answer. Ten minutes of breaking queries beats thirty minutes of rereading solutions, and it prepares you for hidden test cases that probe exactly these edges.

How Hard Is the Mettl Test and How Is It Scored?

The Mettl SQL test is moderately hard for beginners and routine for anyone who writes queries weekly, and scoring is automatic. Coding answers run against hidden test cases, and multiple-choice items are marked right or wrong. Difficulty comes mostly from time pressure and exact output matching, not obscure syntax. A candidate comfortable with joins and aggregation usually finds the first half of the test straightforward and the last two items demanding.

Scoring rules are set by the hiring company. Some employers use an overall percentage cutoff, such as 60 percent, while others require a minimum in each section so strong SQL cannot hide weak logic. Mettl does not publish a universal passing score, so ask the recruiter if the invitation is vague. Knowing whether sections are scored separately changes your strategy, because it tells you whether to spread time evenly or protect your weakest area.

Time allocation matters more than most candidates expect. In a 60-minute test with 15 multiple-choice items and three queries, spend about 15 minutes on the multiple-choice block and 15 minutes on each query. Skip a stuck item after five minutes and return later. Many platforms let you move between questions, but confirm that on the instruction screen, since some sections lock once submitted and you cannot revisit them.

Hidden test cases usually probe the same traps: NULL values, duplicate rows, ties in ranking, empty tables, and case sensitivity in text comparisons. After your query passes the sample output, ask what happens if two employees share the top salary or if a department has no staff. Adding DISTINCT, COALESCE, or DENSE_RANK in the right place often converts a partial score into a full one without changing your core approach.

Anyone curious how the wider platform works can read the guide to test your mettle on Mettl online assessments, which covers invitations, identity checks, and the session flow. The phrase itself comes from the English word mettle, meaning courage and resilience under pressure. People who ask what does mettle mean, or what is mettle, are often surprised that the company name Mettl echoes that word.

Technical failures are covered by a published resume rule. If your connection drops, you can resume within 30 minutes using the same email address, and your attempts and remaining time are kept as of the last sync with the server. Do not clear cookies or switch devices in between. Take a screenshot of any error message, then contact the organization that invited you, since only the test owner can grant extra attempts.

Proctoring adds another layer of difficulty that has nothing to do with SQL. Mercer | Mettl describes AI flagging, live human proctoring, and a secure lockdown browser, depending on what the employer enables. Sit in a well-lit room, keep your face in frame, and close every other application and monitor before starting. Looking away to think is natural, but frequent glances off-screen can generate flags that a reviewer must then assess.

Finally, accept that you will not know everything. Scores are measured against the employer's bar, and a candidate who solves five of six items cleanly often outperforms one who attempts all six with messy, failing queries. Prioritize correct, tested answers on the items you understand best. A calm second pass through flagged questions in the final ten minutes usually recovers more points than rushing the first pass.

What Are Mettl Technical Test Questions and Answers Beyond Queries?

Mettl technical test questions and answers beyond queries cover database theory, such as normalization, transactions, indexes, and keys. These appear as short multiple-choice items and are the fastest points on the test if you hold clean definitions. Below are the concepts that recur most often, with one-line answers you can memorize and then restate in your own words when an item phrases them differently.

Normalization reduces redundancy. First normal form means atomic column values and no repeating groups. Second normal form removes partial dependencies on part of a composite key. Third normal form removes transitive dependencies, so non-key columns depend only on the key. A typical question shows a table with a customer city stored beside every order and asks which normal form is violated. The answer is usually third normal form, because city depends on the customer, not the order.

ACID stands for atomicity, consistency, isolation, and durability. Atomicity means a transaction fully completes or fully rolls back. Isolation levels, from READ UNCOMMITTED to SERIALIZABLE, control which anomalies appear: dirty reads, non-repeatable reads, and phantom reads. Expect a scenario question, such as two users updating one account balance at once, asking which property prevents the lost update. SERIALIZABLE prevents all three anomalies but costs the most concurrency.

On indexes, know that a clustered index defines the physical row order and a table can have only one, while non-clustered indexes are separate structures and a table can have many. A PRIMARY KEY is unique and not null; a FOREIGN KEY enforces a link to another table; a UNIQUE constraint on SQL Server allows one NULL. Questions often ask which index helps a WHERE clause on a frequently filtered column.

Delivery details matter too. Candidates who read about mettl test questions inside the Mettl Secure Browser learn that copy-paste, other tabs, and screen sharing can be restricted, so you cannot look up syntax mid-test. Practice without autocomplete, documentation, or an AI assistant. If your normal editor suggests column names, switch it off for a week so your fingers remember the syntax when the test gives you only a basic editor.

Other recurring theory items: a view is a saved query, not stored data unless it is materialized; a stored procedure can accept parameters and run multiple statements; a trigger fires automatically on INSERT, UPDATE, or DELETE. WHERE versus HAVING, CHAR versus VARCHAR, and INNER versus OUTER joins appear constantly. Write a one-page sheet with these pairs and quiz yourself daily, because comparison questions reward crisp contrasts more than long explanations.

Dialect awareness helps if the test names a database. SQL Server uses TOP and OFFSET FETCH, ISNULL or COALESCE, GETDATE(), and square brackets for unusual identifiers. MySQL uses LIMIT, IFNULL, NOW(), and backticks. PostgreSQL uses LIMIT, COALESCE, NOW(), and double-quoted identifiers. If the problem statement does not name a dialect, stick to standard syntax such as COALESCE and CASE, which minimizes the risk of a syntax error.

A final theory trap is the order of execution. SQL is written starting with SELECT but evaluated FROM first, then WHERE, GROUP BY, HAVING, SELECT, and ORDER BY. That sequence explains alias rules, why window functions cannot appear in WHERE, and why filtering early with WHERE is cheaper than filtering late with HAVING. Draw this sequence from memory until it feels automatic; it resolves a surprising number of debugging puzzles.

How Do You Prepare for the Mettl Test in Two Weeks?

A two-week plan is enough if you already know basic SQL: spend days one to four on SELECT, joins, and aggregation, days five to nine on subqueries, CASE, and window functions, and days ten to fourteen on timed mixed sets. Study 60 to 90 minutes daily and write every query by hand in a sandbox. Reading solutions feels productive but produces little recall, while typing and debugging builds the habit you need.

Build a practice database. Create Employees, Departments, Customers, Orders, and Products tables with 20 to 50 rows each, then deliberately include NULLs, duplicate names, and tied salaries. Every query you write should be checked against these traps. SQLite, a local MySQL install, or a browser-based fiddle all work and cost nothing. Having your own data lets you predict output before running, which is exactly the skill multiple-choice items measure.

For general format guidance, the mettl assessment test tips article covers the registration email, identity verification, and common mistakes across sections. Read it alongside this guide, then rehearse the full flow once: open the invitation link early, test your webcam and microphone, confirm your upload speed meets the 512 kbps minimum for proctored tests, and close background apps that consume bandwidth, such as cloud backups and video calls.

Run two timed mock sessions in the final week. Set a 45-minute timer, solve eight multiple-choice items and three queries, then review every miss. Categorize errors: syntax, logic, NULL handling, or time. Most candidates discover that one category causes most losses. If syntax slips dominate, drill clause order; if logic dominates, practice breaking problems into steps; if time dominates, practice skipping and returning to hard items.

On test day, read each question twice before typing. Write the SELECT list last, after you know which tables and joins are needed. Start with a simple version that returns something, then refine it. Use table aliases, qualify column names, and format the query on several lines so you can spot a missing comma. Run the query against the sample data every time, and confirm column names and ordering match the prompt exactly.

Treat integrity as part of preparation. Do not use brain dumps, pay for leaked questions, or let someone else sit the test. Proctoring tools and identity checks can detect these behaviors, and an invalidated result can cost you the opportunity and, in some cases, future applications to the same employer. Honest preparation also pays off after the test, since interviews often repeat the same concepts live, with an engineer watching you reason.

Finish with recovery. Sleep properly the night before, eat, and log in 15 minutes early. If something breaks, stay calm and document it. Afterward, note which topics appeared while they are fresh, without recording actual question text, and use that to prepare for the technical interview. Whether you pass or not, a short review of what felt slow gives you a concrete target for the next assessment you take.

Last, keep a one-page cheat sheet of patterns, not answers: top-N per group with ROW_NUMBER, anti-join with LEFT JOIN and IS NULL, running totals with SUM OVER, duplicate detection with GROUP BY and HAVING COUNT greater than one, and conditional counts with CASE inside SUM. Nearly every intermediate task is a variation on one of these five. Review the sheet the evening before, then stop studying and rest.