diff --git a/docs/docs/images/version-creation-form.png b/docs/docs/images/version-creation-form.png
new file mode 100644
index 000000000..a7e83db62
Binary files /dev/null and b/docs/docs/images/version-creation-form.png differ
diff --git a/docs/docs/workflows/improve-your-data/_category_.json b/docs/docs/workflows/improve-your-data/_category_.json
index a23fc52b2..1a42bf7c9 100644
--- a/docs/docs/workflows/improve-your-data/_category_.json
+++ b/docs/docs/workflows/improve-your-data/_category_.json
@@ -1,6 +1,6 @@
 {
   "label": "Improve Your Data",
-  "position": 4,
+  "position": 6,
   "collapsible": true,
   "link": {
     "type": "generated-index",
diff --git a/docs/docs/workflows/using-2d-embeddings.md b/docs/docs/workflows/using-2d-embeddings.md
index 6c8558afb..fdfa93f54 100644
--- a/docs/docs/workflows/using-2d-embeddings.md
+++ b/docs/docs/workflows/using-2d-embeddings.md
@@ -1,13 +1,13 @@
 ---
-sidebar_position: 3
+sidebar_position: 5
 ---
 
 # Using 2D Embeddings
 
 **Use 2D embeddings to improve your active learning cycle**
 
-Encord Active presents you with a 2D embedding plot for all the images in your project. You can use it to find interesting 
-clusters, understand your data better, weakly label your images, remove detrimental images, etc. The 2D 
+Encord Active presents you with a 2D embedding plot for all the images in your project. You can use it to find interesting
+clusters, understand your data better, weakly label your images, remove detrimental images, etc. The 2D
 embedding plot can be found in the **Explorer** pages of the **Data Quality** and **Label Quality** sections. Here is an example
 embedding plot for the COCO 2017 validation set.
 
@@ -25,17 +25,17 @@ snowy images.
 ![img.png](../images/workflows/using-2d-embeddings/using-2d-embeddings-3.png)
 
 Now you can perform several actions with the selected group:
-- You can tag and export them in the **Actions** => **Filter & Export** page so that you can label these images 
+- You can tag and export them in the **Actions** => **Filter & Export** page so that you can label these images
 automatically via a script.
 - You can explore the performance of the selected samples in the model performance page.
 - You can create different subsets like these and compare them.
 
 Samples in the 2D embedding plot in the **Data Quality => Summary** page have no label information; therefore, they all share
-the same color. On the other hand, points colored according to labels on the **Label Quality => Summary** page. 
+the same color. On the other hand, points colored according to labels on the **Label Quality => Summary** page.
 
 ![using-2d-embeddings-4](../images/workflows/using-2d-embeddings/using-2d-embeddings-4.png)
 
 Now you can:
 - Check which classes are confused with each other
 - Spot wrongly labeled samples (e.g., a different class inside the large cluster of another class)
-- Detect outliers and remove them from the dataset
\ No newline at end of file
+- Detect outliers and remove them from the dataset
diff --git a/docs/docs/workflows/versioning.md b/docs/docs/workflows/versioning.md
new file mode 100644
index 000000000..20771e089
--- /dev/null
+++ b/docs/docs/workflows/versioning.md
@@ -0,0 +1,45 @@
+---
+sidebar_position: 4
+---
+
+# Versioning
+
+## Why?
+When you do experiments and test hypotheses, you typically want to jump back and forth between different versions of your data, labels, and models. For example, when you train a model on a specific subset of your data, you will typically find that there is an edge case for which your model performs poorly. Hence, you expand your dataset with more data to better cover the edge case and train a new model.
+
+In order to track your experiments and compare not only the model performance but also the underlying data shown to the model, you can use the project versioning feature of Encord Active.
+
+## What?
+
+The versioning is global for the project, so everything from what data and labels were available at a given point in time to what the model predictions were will be versioned and available.
+
+Currently, versioning works like checkpoints, you can create a checkpoint and jump between them to view previous checkpoints.
+
+
+## How?
+
+#### Creating a new version
+
+In order to create a new version, navigate to the Versioning page in the Actions menu item.
+
+Provide a version name and click `create`.
+
+![Version creation form](../images/version-creation-form.png)
+
+
+:::tip
+
+You also have the ability to discard any outstanding changes, i.e. everything after the last version.
+
+:::
+
+#### Viewing a previous version
+
+On the left sidebar, there is a dropdown which allows version selection. Selecting an old version will temporary save any outstanding changes until the latest version is selected again.
+
+:::warning
+
+While on a previous version the app will be in read-only mode. Any changes made will be dicarded.
+
+:::
+