Update the breaking change analysis doc

Signed-off-by: popcorny <[email protected]>

Author: popcornylu

docs/assets/images/features/breaking-change-disabled.png

@@ -3,21 +3,23 @@ title: Breaking Change Analysis
 icon: octicons/diff-modified-24
 ---
 
-Breaking change analysis examines models for modifications to the SQL semantic tree that could potentially alter the data generated by the model, resulting in downstream impact.
+**Breaking Change Analysis** examines modified models and categorizes changes into three types: **Breaking change**, **Partial breaking change**, and **Non breaking change**.
 
-Currently, all modified models are treated as **breaking changes** unless:
+Traditionally, any modification to a model’s SQL was assumed to affect all downstream models. However, not all changes have the same level of impact. For example, formatting adjustments or adding a new column should not break any downstream dependencies. **Breaking Change Analysis** helps you assess whether a change affects downstream models, and if so, to what extent.
 
-1. The semantics are unchanged.
-1. The modification only involves adding a new column.
 
-If no structural differences are detected or the changes involve simply adding new columns, the modification is classified as non-breaking changes.
+##  Categories
+Category | Downstream Impact | Examples
+---|---|---
+Non breaking change | No downstream models are affected |  Adding a new column, formatting SQL, adding comments
+Partial breaking change | Only downstream models referencing certain columns are affected | Removing, renaming, or modifying the definition of a column
+Breaking change | All downstream models are affected | Changing filter conditions (e.g. `WHERE`), sort order (`ORDER BY`), or other SQL logic
 
 ## Usage
 
-To enable the feature, click the **breaking Change Analysis** toggle on the  **Lineage** page. 
+To enable the feature, click the **Breaking Change Analysis** toggle on the  **Lineage** page. 
 
-- Models with **breaking changes** can be identified by a solid orange border.
-- Models with **non-breaking changes** can be identified by the dashed orange border.
+All modified models will display their change category directly on the node. Additionally, partial breaking changes are highlighted with a dashed orange border to indicate that they may not impact your downstream models.
 
 === "Disabled"
 
@@ -28,12 +30,16 @@ To enable the feature, click the **breaking Change Analysis** toggle on the  **L
     ![Breaking Change Analysis](../../assets/images/features/breaking-change.png){: .shadow}
 
 
+## Column-Level Lineage
+
+In models classified as **Non breaking** or **Partial breaking**, added, removed, or modified columns will be listed. You can click on a column to open its [Column-Level Lineage](./column-level-lineage.md)
+
+![Column-level lineage](../../assets/images/features/breaking-change-lineage.png){: .shadow}
 
 ## Limitations
 
-The current implementation of breaking change analysis is still very conservative. As a result, a modified model may be classified as a breaking change when it is actually non-breaking. Common cases include:
+The current implementation of breaking change analysis is still very conservative. As a result, a modified model may be classified as a breaking change when it is actually non breaking or partial breaking changes. Common cases include:
 
-1. Adding a column using a `CASE WHEN` statement.
 1. Logical equivalence in operations, such as changing `a + b` to `b + a`.
 1. Adding a `LEFT JOIN` to a table and selecting columns from it. This is often used to enrich the current model with additional dimension table data without affecting existing downstream tables.
 1. All modified python models or seeds are treated as breaking change.