diff --git a/docs/02-getting-started.md b/docs/02-getting-started.md
index 95b9b00..1894fb1 100644
--- a/docs/02-getting-started.md
+++ b/docs/02-getting-started.md
@@ -54,6 +54,7 @@ from basalt.observability import start_observe, observe
 
 
 @start_observe(
+    feature_slug="workflow",
     name="my_workflow",
     identity={
         "organization": {"id": "123", "name": "ACME"},
@@ -89,6 +90,7 @@ from basalt.observability import start_observe, observe
 def process_request(user_id, data):
     # Create root span with identity
     with start_observe(
+            feature_slug="request-processing",
             name="process_request",
             identity={"user": user_id},
             metadata={"source": "api"}
@@ -119,6 +121,7 @@ from basalt.observability import start_observe, observe
 
 # Method 1: Set on root span (recommended)
 @start_observe(
+    feature_slug="chat-handler",
     name="chat_handler",
     identity={
         "organization": {"id": "123", "name": "ACME"},
@@ -131,7 +134,7 @@ def handle_chat(message):
 
 
 # Method 2: Set dynamically
-@start_observe(name="api_handler")
+@start_observe(feature_slug="api-handler", name="api_handler")
 def handle_request(auth_token):
     user_data = verify_token(auth_token)  # Returns {"id": "user-123", "name": "John Doe"}
     observe.set_identity({"user": user_data})
@@ -144,7 +147,7 @@ Add custom key-value pairs to your spans:
 
 ```python
 # On root span
-@start_observe(name="workflow", metadata={"model": "gpt-4", "temperature": 0.7})
+@start_observe(feature_slug="workflow", name="workflow", metadata={"model": "gpt-4", "temperature": 0.7})
 def workflow():
     pass
 
diff --git a/docs/03-prompts.md b/docs/03-prompts.md
index 70ca78a..63b2493 100644
--- a/docs/03-prompts.md
+++ b/docs/03-prompts.md
@@ -514,7 +514,7 @@ Combine prompt context managers with manual spans:
 from basalt import Basalt
 from basalt.observability import start_observe, observe
 
-@start_observe(name="QA System", identity={"user": {"id": "user_123"}})
+@start_observe(feature_slug="qa-system", name="QA System", identity={"user": {"id": "user_123"}})
 def answer_question(question: str):
     with basalt.prompts.get_sync("qa-prompt", variables={"question": question}) as prompt:
         # Prompt span is a child of "QA System"
diff --git a/docs/05-observability.md b/docs/05-observability.md
index 68bec10..cb20728 100644
--- a/docs/05-observability.md
+++ b/docs/05-observability.md
@@ -15,6 +15,7 @@ from basalt.observability import start_observe, observe, ObserveKind
 
 # Root span with identity tracking
 @start_observe(
+    feature_slug="request-processing",
     name="process_request",
     identity={"user": {"id": "user_123", "name": "Alice"}, "organization": {"id": "org_abc"}},
     metadata={"environment": "production", "version": "2.0"}
@@ -25,6 +26,7 @@ def process():
 
 # Root span with experiment tracking
 @start_observe(
+    feature_slug="ml-workflow",
     name="ml_workflow",
     experiment={"id": "exp_456", "name": "Model Comparison"},
     identity={"user": "analyst_001"}
@@ -34,6 +36,7 @@ def run_experiment():
 
 # Context manager form
 with start_observe(
+    feature_slug="batch-processing",
     name="batch_job",
     identity={"organization": {"id": "org_xyz", "name": "Acme Corp"}},
     metadata={"job_id": "batch_123"}
@@ -43,6 +46,8 @@ with start_observe(
 ```
 
 **Key features of `start_observe`:**
+- **`feature_slug`** (required): A unique identifier for the feature or workflow being traced (e.g., "user-auth", "payment-flow").
+- **`name`** (required): A descriptive name for the root span.
 - **`identity`**: Dict with `user` and/or `organization` keys for tracking. Can also use simple string format or callables for dynamic resolution.
 - **`experiment`**: Dict with `id`, `name`, and `variant` keys for A/B testing and feature tracking.
 - **`evaluate_config`**: Configuration for evaluators attached to the root span.
@@ -57,6 +62,7 @@ from basalt.observability import observe, start_observe, ObserveKind
 
 # Root span (required as entry point)
 @start_observe(
+    feature_slug="request-handler",
     name="process_request",
     identity={
         "organization": {"id": "123", "name": "ACME"},
@@ -182,7 +188,7 @@ Sometimes you need to set metadata or identity on the root span from deeply nest
 from basalt.observability import observe, start_observe
 
 
-@start_observe(name="API Handler")
+@start_observe(feature_slug="api-handler", name="API Handler")
 def handle_request(user_id):
     # Root span starts here with identity tracking
     observe.set_input({"user_id": user_id})
diff --git a/docs/06-manual-tracing.md b/docs/06-manual-tracing.md
index d743621..1c4c665 100644
--- a/docs/06-manual-tracing.md
+++ b/docs/06-manual-tracing.md
@@ -24,6 +24,7 @@ Every trace must begin with a root span created using `start_observe`:
 from basalt.observability import start_observe, observe
 
 @start_observe(
+    feature_slug="data-processing",
     name="Data Processing Workflow",
     identity={
         "organization": {"id": "123", "name": "ACME"},
@@ -62,6 +63,7 @@ from basalt.observability import start_observe
 
 def process_request():
     with start_observe(
+            feature_slug="request-processing",
             name="Request Processing",
             identity={
                 "organization": {"id": "123", "name": "ACME"},
@@ -85,6 +87,7 @@ from basalt.observability import start_observe
 
 # With identity tracking
 @start_observe(
+    feature_slug="main-workflow",
     name="Main Workflow",
     identity={
         "organization": {"id": "123", "name": "ACME"},
@@ -97,6 +100,7 @@ def main_workflow(data):
 
 # With experiment tracking
 @start_observe(
+    feature_slug="ab-test",
     name="A/B Test",
     experiment={"id": "exp_001", "name": "Model Comparison", "variant": "model_a"},
     identity={
@@ -109,6 +113,7 @@ def run_experiment():
 
 # Context manager form
 with start_observe(
+    feature_slug="batch-job",
     name="Batch Job",
     identity={
         "organization": {"id": "123", "name": "ACME"},
@@ -121,7 +126,8 @@ with start_observe(
 
 **Key `start_observe` parameters:**
 
-- `name`: Span name (defaults to function name if used as decorator)
+- `feature_slug` (required): A unique identifier for the feature or workflow being traced
+- `name` (required): Span name (defaults to function name if used as decorator)
 - `identity`: Dict with `user` and/or `organization` keys for tracking
 - `experiment`: Dict with `id`, `name`, and `variant` for A/B testing
 - `evaluate_config`: Evaluator configuration for the root span
@@ -178,6 +184,7 @@ Decorators automatically create parent-child relationships. Always start with `s
 from basalt.observability import start_observe, observe
 
 @start_observe(
+    feature_slug="main-workflow",
     name="Main Workflow",
     identity={
         "organization": {"id": "123", "name": "ACME"},
@@ -347,7 +354,7 @@ You can mix decorators and context managers:
 ```python
 from basalt.observability import start_observe, observe
 
-@start_observe(name="Main Workflow")
+@start_observe(feature_slug="main-workflow", name="Main Workflow")
 def main_workflow(data):
     # Root span via decorator
     prepare_data(data)
@@ -426,7 +433,7 @@ basalt = Basalt(api_key="your-api-key")
 openai_client = OpenAI(api_key="your-openai-key")
 
 # Root trace
-with start_observe(name="process_request", feature_slug="support-ticket"):
+with start_observe(feature_slug="support-ticket", name="process_request"):
     # Fetch prompt with context manager
     with basalt.prompts.get_sync("joke-analyzer", variables={"jokeText": "..."}) as prompt:
         # Auto-instrumented OpenAI call automatically receives prompt attributes
diff --git a/docs/08-async-observability.md b/docs/08-async-observability.md
index 2c856d6..df33694 100644
--- a/docs/08-async-observability.md
+++ b/docs/08-async-observability.md
@@ -18,7 +18,7 @@ import asyncio
 from basalt.observability import async_start_observe, async_observe
 
 async def main():
-    async with async_start_observe(name="async_workflow") as span:
+    async with async_start_observe(feature_slug="async-workflow", name="async_workflow") as span:
         span.set_input({"task": "process_data"})
         result = await process_data()
         span.set_output(result)
@@ -42,6 +42,7 @@ from basalt.observability import async_start_observe
 # Context manager form
 async def workflow():
     async with async_start_observe(
+        feature_slug="user-request",
         name="User Request",
         identity={"user": {"id": "user_123"}},
         metadata={"environment": "production"}
@@ -52,6 +53,7 @@ async def workflow():
 
 # Decorator form
 @async_start_observe(
+    feature_slug="api-handler",
     name="API Handler",
     identity={"user": {"id": "user_456"}}
 )
@@ -95,7 +97,7 @@ The `@observe` and `@start_observe` decorators automatically detect async functi
 from basalt.observability import observe, start_observe
 
 # These automatically work with async functions
-@start_observe(name="Async Root")
+@start_observe(feature_slug="async-root", name="Async Root")
 async def async_root():
     await nested_operation()
 
@@ -133,6 +135,7 @@ async def generate_greeting(user: dict) -> str:
 async def handle_user_request(user_id: str) -> str:
     """Main handler with root span"""
     async with async_start_observe(
+        feature_slug="user-request",
         name="user_request",
         identity={"user": {"id": user_id}},
         metadata={"version": "2.0"}
diff --git a/docs/11-experiments.md b/docs/11-experiments.md
index e803131..7403fc6 100644
--- a/docs/11-experiments.md
+++ b/docs/11-experiments.md
@@ -54,6 +54,7 @@ from basalt.observability import start_observe, observe
 
 
 @start_observe(
+    feature_slug="ab-test",
     name="experiment.variant_a",
     experiment="exp-456",
     identity={
@@ -73,6 +74,7 @@ def run_variant_a():
 
 # Or using context manager
 with start_observe(
+        feature_slug="ab-test",
         name="experiment.variant_b",
         experiment="exp-456",
 ):
diff --git a/docs/12-user-org-tracking.md b/docs/12-user-org-tracking.md
index cda0c84..d88ca32 100644
--- a/docs/12-user-org-tracking.md
+++ b/docs/12-user-org-tracking.md
@@ -81,7 +81,7 @@ You can also set identity dynamically within a span using `observe.set_identity(
 ```python
 from basalt.observability import observe, start_observe
 
-@start_observe(name="workflow")
+@start_observe(feature_slug="data-workflow", name="workflow")
 def process_data(auth_token):
     # Authenticate and get user info
     user_info = authenticate(auth_token)
@@ -105,7 +105,7 @@ When using context managers, you can set identity on the span handle:
 from basalt.observability import async_start_observe
 
 async def workflow():
-    async with async_start_observe(name="process", feature_slug="task") as span:
+    async with async_start_observe(feature_slug="task", name="process") as span:
         # Set identity on the span handle
         span.set_identity({
             "user": {"id": "user-789"},
@@ -197,7 +197,7 @@ def get_identity_from_request():
     }
 
 @app.route("/api/orders/<order_id>")
-@start_observe(name="get_order", feature_slug="orders", identity=get_identity_from_request)
+@start_observe(feature_slug="orders", name="get_order", identity=get_identity_from_request)
 def get_order(order_id):
     # Identity automatically set from request
     order = fetch_order(order_id)
@@ -212,8 +212,8 @@ Identity works seamlessly with async/await:
 from basalt.observability import async_start_observe
 
 @async_start_observe(
-    name="async_workflow",
     feature_slug="background-jobs",
+    name="async_workflow",
     identity={"user": {"id": "user-123"}}
 )
 async def process_async(data):