What the interviewer checks in your SQL answer

Checks 1-4: NULLs and joins that lie

1. 01

   COUNT(*) vs COUNT(col)

   COUNT(phone) skips NULLs. Know which one the question wants.

   SELECT COUNT(*)     AS all_rows,
          COUNT(phone) AS has_phone
   FROM users;

2. 02

   = NULL never matches

   Comparing with NULL gives unknown, so the row is dropped. Use IS NULL.

   SELECT name FROM employees
   WHERE manager_id IS NULL;

3. 03

   LEFT JOIN + WHERE

   A filter on the right table in WHERE turns it into an inner join. Put it in ON.

   FROM users u
   LEFT JOIN orders o
     ON o.user_id = u.id
    AND o.status = 'Paid'

4. 04

   Rows that multiply

   One user, 5 orders, 3 addresses = 15 rows. Count rows before and after every join.

   SELECT COUNT(*) FROM orders;
   -- run again after the JOIN

Checks 5-8: ties, maths, dates, groups

1. 05

   Ties in ranking

   2nd highest salary with two people on top? DENSE_RANK keeps ties together.

   DENSE_RANK() OVER (
     ORDER BY salary DESC) AS rnk

2. 06

   Integer division

   In SQL Server and PostgreSQL, 7/2 gives 3 on whole numbers. Multiply by 100.0 first.

   SELECT paid * 100.0 / total AS paid_pct
   FROM summary;

3. 07

   Date ranges

   BETWEEN on a timestamp misses the last day after midnight. Use >= and <.

   WHERE order_ts >= '2026-09-01'
     AND order_ts <  '2026-10-01'

4. 08

   WHERE vs HAVING

   WHERE filters rows before grouping. HAVING filters the groups.

   GROUP BY city
   HAVING COUNT(*) > 10;

How you talk matters too

Test yourself live

Try the traps on real tables: customers who never ordered, highest paid with ties, cancellation %. Free, checked instantly.