SQL GROUP BY Multiple Columns Guide

When you GROUP BY two or more columns, SQL creates one group for each unique combination of values across those columns. A table with 3 departments and 4 job titles could produce up to 12 groups (3 x 4), though only combinations that exist in the data create actual groups. Each group gets one row in the result.

-- Revenue by department and quarter
SELECT
  department,
  quarter,
  SUM(revenue) AS total_revenue,
  COUNT(*) AS deal_count
FROM sales
GROUP BY department, quarter
ORDER BY department, quarter;

-- If department has values (Sales, Marketing)
-- and quarter has values (Q1, Q2, Q3, Q4)
-- up to 8 groups are possible:
-- Sales/Q1, Sales/Q2, Sales/Q3, Sales/Q4
-- Marketing/Q1, Marketing/Q2, Marketing/Q3, Marketing/Q4

Hierarchical grouping produces summaries at the intersection of organizational levels. Grouping by country and city shows totals for each city within each country. Grouping by year, month, and day creates daily summaries nested within months and years. This pattern powers most business reporting dashboards.

-- Headcount by country and city
SELECT
  country,
  city,
  COUNT(*) AS employee_count,
  AVG(salary) AS avg_salary
FROM employees
GROUP BY country, city
ORDER BY country, city;

-- Result:
-- USA  | New York      | 45 | 95000
-- USA  | San Francisco | 32 | 110000
-- UK   | London        | 28 | 85000
-- UK   | Manchester    | 12 | 72000

-- Each row is a unique (country, city) pair
-- "USA" alone is not a group; "USA, New York" is

Combining multi-column GROUP BY with conditional aggregation produces a cross-tabulation: one dimension on the rows, another on the columns. This is the SQL equivalent of a pivot table. The GROUP BY defines the row dimension. CASE WHEN inside aggregate functions creates the column dimension.

-- Sales by product (rows) and quarter (columns)
SELECT
  product_name,
  SUM(CASE WHEN quarter = 'Q1' THEN revenue ELSE 0 END) AS q1,
  SUM(CASE WHEN quarter = 'Q2' THEN revenue ELSE 0 END) AS q2,
  SUM(CASE WHEN quarter = 'Q3' THEN revenue ELSE 0 END) AS q3,
  SUM(CASE WHEN quarter = 'Q4' THEN revenue ELSE 0 END) AS q4,
  SUM(revenue) AS full_year
FROM sales
GROUP BY product_name
ORDER BY full_year DESC;

-- Two-dimension grouping with conditional aggregation
SELECT
  department,
  job_level,
  COUNT(*) AS headcount,
  COUNT(*) FILTER (WHERE gender = 'F') AS female_count,
  COUNT(*) FILTER (WHERE gender = 'M') AS male_count
FROM employees
GROUP BY department, job_level;

HAVING filters groups after aggregation. With multi-column GROUP BY, HAVING filters based on the aggregate values of composite groups. This lets you ask questions like 'which department-quarter combinations had more than $1M in revenue' or 'which city-product pairs sold fewer than 10 units.'

-- Department-quarter pairs with high revenue
SELECT
  department,
  quarter,
  SUM(revenue) AS total_revenue
FROM sales
GROUP BY department, quarter
HAVING SUM(revenue) > 1000000
ORDER BY total_revenue DESC;

-- City-category pairs with low sales volume
SELECT
  city,
  product_category,
  COUNT(*) AS order_count,
  SUM(amount) AS total_sales
FROM orders
JOIN products ON orders.product_id = products.product_id
GROUP BY city, product_category
HAVING COUNT(*) < 10;