diff --git a/07_Window_Functions/03_DENSE_RANK.md b/07_Window_Functions/03_DENSE_RANK.md
new file mode 100644
index 0000000..b02d14e
--- /dev/null
+++ b/07_Window_Functions/03_DENSE_RANK.md
@@ -0,0 +1,106 @@
+# DENSE_RANK()
+
+## Overview
+
+`DENSE_RANK()` behaves like `RANK()` except it **never leaves a gap** in
+the ranking sequence after a tie.
+
+```
+Ordered values: [10, 20, 20, 40]
+RANK():         [1,  2,  2,  4]
+DENSE_RANK():   [1,  2,  2,  3]
+```
+
+## Learning Objectives
+
+- Assign gap-free competition ranks.
+- Correctly choose between `RANK()` and `DENSE_RANK()` based on business
+  requirements.
+- Compare all three ranking functions side by side.
+
+## Prerequisites
+
+- `01_ROW_NUMBER`
+- `02_RANK`
+
+## Syntax
+
+```sql
+SELECT
+    column_a,
+    DENSE_RANK() OVER (ORDER BY sort_column) AS dense_rank_value
+FROM table_name;
+```
+
+## Dataset Used
+
+`employes`
+
+## Examples
+
+See [`03_dense_rank.sql`](./03_dense_rank.sql).
+
+## Real World Applications
+
+- Salary bands / pay-grade tiers where consecutive tier numbers matter.
+- Product pricing tiers where no tier number should ever be "skipped".
+
+## Business Use Cases
+
+| Domain | Scenario |
+|---|---|
+| Finance | Assign consecutive risk tiers to loan applicants |
+| Retail | Assign consecutive pricing tiers to products |
+| HR | Assign consecutive seniority bands per manager group |
+
+## Common Mistakes
+
+- Using `DENSE_RANK()` when the business actually wants gap-aware
+  standings (use `RANK()` instead).
+- Assuming `DENSE_RANK()` and `ROW_NUMBER()` are interchangeable when
+  there are no ties in the sample data -- they diverge as soon as a tie
+  appears.
+
+## Best Practices
+
+- Always test ranking functions against a dataset that **contains ties**
+  before shipping to production; behavior on unique data hides real bugs.
+- Comment the choice of ranking function inline so the next engineer
+  does not have to reverse-engineer the intent.
+
+## Engineering Notes
+
+`DENSE_RANK()` maintains an internal counter that increments only when
+the `ORDER BY` value changes from the previous row -- this is the single
+mechanical difference from `RANK()`, which instead increments by the
+number of rows seen so far.
+
+## Practice Questions
+
+1. Assign dense rank to employees ordered by `manager_id`.
+2. Show employee name and dense rank (ordered by `emp_id`).
+3. Show employees with dense rank ≤ 3.
+4. Compare `RANK()` and `DENSE_RANK()` side by side.
+5. Show the employee(s) with dense rank = 1.
+
+## Difficulty
+
+Beginner
+
+## Estimated Time
+
+20 minutes
+
+## Learning Outcomes
+
+- Predict `DENSE_RANK()` output on tied data without running the query.
+- Justify, in an interview, when to use `DENSE_RANK()` over `RANK()`.
+
+## Related Topics
+
+- `02_RANK`
+- `04_PARTITION_BY`
+
+## Next Topic
+
+[`04_PARTITION_BY`](../04_PARTITION_BY/README.md)