Coding/Learn/sql/sql sum avg
SQL & Relational Database Engineering Complete Course
01.Introduction & Relational Database Architecture
02.Syntax, Clauses & Execution Order
03.SELECT: Column Projection & Expressions
04.SELECT DISTINCT & Row Deduplication
05.WHERE Clause: Filtering Predicates & Comparison
06.ORDER BY: Sorting, Multi-Column & NULLS Ordering
07.Logical Operators: AND, OR, NOT & Operator Precedence
08.Comments: Single-Line (--), Multi-Line (/* */) & Query Hints
09.Operators: Arithmetic, Bitwise & String Concatenation
10.LIKE Operator & Wildcard Characters (%, _)
11.IN & NOT IN: Membership Operators & Subqueries
12.BETWEEN: Range Filtering for Numbers, Dates & Text
13.NULL Values: Three-Valued Logic, IS NULL & IS NOT NULL
14.Aliases: AS Keyword for Columns & Tables
15.SELECT TOP, LIMIT & OFFSET Pagination
16.INSERT INTO: Single & Multi-Row Insertions
17.UPDATE: Modifying Existing Records & Safe WHERE Rules
18.DELETE: Removing Rows vs TRUNCATE TABLE
19.SELECT INTO & Table Backups (CREATE TABLE AS SELECT)
20.INSERT INTO SELECT: Bulk Data Migration
21.Aggregate Functions: Statistical Summaries
22.MIN() & MAX(): Finding Extremes Across Types
23.COUNT: COUNT(*), COUNT(column) & COUNT(DISTINCT)
24.SUM() & AVG(): Numerical Aggregation & Averages
25.GROUP BY: Aggregation Buckets & Categorical Metrics
26.HAVING: Filtering Aggregated Groups vs WHERE
27.Joins Architecture: Relational Modeling & Venn Diagrams
28.INNER JOIN: Matching Intersections Across Tables
29.LEFT JOIN: Preserving Unmatched Left Rows & Finding Orphans
30.RIGHT JOIN: Preserving Unmatched Right Rows
31.FULL JOIN: Full Outer Join & Complete Reconciliation
32.SELF JOIN: Hierarchies, Trees & Parent-Child Relationships
33.UNION: Combining Result Sets with Automatic Deduplication
34.UNION ALL vs UNION: Performance & Duplicates
35.EXISTS & NOT EXISTS: Correlated Subqueries
36.ANY & ALL: Quantified Comparison Subqueries
37.CASE: Conditional Expressions & Data Transposition
38.Null Functions: COALESCE, IFNULL, ISNULL & NULLIF
39.Stored Procedures, Functions & Query Parameters
40.CREATE & DROP DATABASE: Environment Lifecycle
41.CREATE TABLE: Defining Schemas & Types
42.DROP & TRUNCATE TABLE: Structure vs Data Purging
43.ALTER TABLE: Modifying Columns & Schema Evolution
44.AUTO INCREMENT & Sequences: Identity Generation
45.Date & Time Formats: Timestamp Manipulation
46.Views: Virtual Tables & Security Layers
47.Constraints Overview: Relational Integrity
48.NOT NULL & UNIQUE: Data Invariants
49.PRIMARY KEY: Entity Uniqueness & Clustering
50.FOREIGN KEY: Referential Integrity & Cascades
51.CHECK & DEFAULT: Business Rules at DB Layer
52.CREATE INDEX: B-Trees & Query Optimization
53.Injection Defense: Threat Models & Prevention
54.Parameters & Prepared Statements: Safe Execution
55.Database Engines & Hosting Architecture
56.Dialect References & Data Type Standards
57.Production Best Practices & Query Tuning Checklist
58.Interactive Exercises: Multi-Table Analytics Challenges
59.Comprehensive Assessment & Knowledge Check
60.Capstone Project: Enterprise Analytics Engine
61.Mastery Certification & Graduate Verification
Course Certificate0%

Complete all lessons to earn

LearnSQL SUM() & AVG(): Numerical Aggregation & Averages
24 / 61 0%
Lesson 24 of 61
5 mins read

SQL SUM() & AVG(): Numerical Aggregation & Averages

PreviousNext Lesson
In Plain English

`SUM()` computes the mathematical total sum of numerical values in a column, while `AVG()` computes the arithmetic mean (`SUM(col) / COUNT(col)`). Both functions ignore NULL values during calculation.

The Mental Model

"`SUM()` adds up every deposit on a bank statement to find total incoming cash; `AVG()` divides by the number of transactions to find the average deposit size."

Deep Dive: How It Works

Null Handling in AVG(): If a table has 4 rows with scores (100, 80, NULL, 60), `AVG(score)` evaluates as `(100+80+60) / 3 = 80.0`, NOT dividing by 4.

Handling Missing Values as Zero: If you want NULLs treated as 0 in an average: `AVG(COALESCE(score, 0))`.

Conditional SUM with CASE: `SUM(CASE WHEN status = 'refunded' THEN amount ELSE 0 END)` calculates refunded revenue inline.

Syntax Blueprint

SELECT
  SUM(sales_amount) AS total_revenue,
  AVG(sales_amount) AS average_sale
FROM transactions;

Use SUM for grand totals and AVG for arithmetic means.

PointCore Rules to Remember

Point star
Numeric Only: SUM and AVG operate on integers, floats, and decimals (fails on strings).
Point star
Float Return: AVG returns floating point numbers even when operating on integer columns.
Point star
Conditional Sums: Combine SUM with CASE to create pivot tables and analytical dashboards.

Common Beginner Traps & How to Fix Them

Mistake:Forgetting that AVG() ignores NULL rows and assuming it divides by the total row count.

Why it happens: AVG() divides by COUNT(col), not COUNT(*).

How to fix: Use `AVG(COALESCE(col, 0))` if unrecorded rows should be treated as zero.

Pro Tip: Use `SUM(col) * 1.0 / COUNT(*)` to calculate exact averages including NULL rows as zero.

Live Interactive Example

Hit Run Code to see it live
Calculating Total Revenue and Average Ticket Size
Python 3.12
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
Output Console
Click "Run Code" to view the rendered output.
How it works: SUM calculated $800 total, AVG computed $200 mean, and conditional SUM calculated $650 wholesale.

Your Turn: Micro Challenge

No pressure! Edit the starter code below and test your solution with instant feedback.

Micro Exercise

Calculate Sum and Average

  • PointSelect `SUM(cost) AS total_cost`, `AVG(cost) AS avg_cost` from `expenses`.
1
2
3
4
5
Sandbox Output
Click "Run & Check" to test your solution.

Finished reading and practicing?

Mark this lesson as completed to update your course progress.

PreviousNext Lesson
Overview100 SheetLearnCertificates