diff --git a/examples/remote_scorer/README.md b/examples/remote_scorer/README.md
index da98ec92d4c2..66cb6b21fdc3 100644
--- a/examples/remote_scorer/README.md
+++ b/examples/remote_scorer/README.md
@@ -40,19 +40,48 @@ The endpoint must return HTTP 200 with a JSON object. The required top-level
 contract fields are:
 
 - `schema_version`: required integer; must be `1`.
-- `result`: required; can be any JSON-serializable shape you choose.
+- `result`: required structured scorer output.
 
-For example:
+The simplest structured result is one score object:
+
+- `value`: required; either a tag string, max 36 characters, or a numeric rating
+  from `0.0` to `1.0`.
+- `reason`: optional string explaining the score.
+- `confidence`: optional numeric confidence from `0.0` to `1.0`.
+
+For a single-score response:
 
 ```json
 {
   "schema_version": 1,
   "result": {
-    "message_length": 32
+    "value": 1.0,
+    "reason": "The response is clear and concise.",
+    "confidence": 0.9
   }
 }
 ```
 
+This sample returns one numeric rating and one tag:
+
+```json
+{
+  "schema_version": 1,
+  "result": [
+    {
+      "value": 1.0,
+      "reason": "Message is 32 characters; concise messages score best.",
+      "confidence": 1.0
+    },
+    {
+      "value": "concise",
+      "reason": "Message length category is concise.",
+      "confidence": 0.9
+    }
+  ]
+}
+```
+
 Non-200 responses are treated as scorer failures by Weave. The `result` value is
 the scorer output that Weave records as feedback.
 
@@ -191,5 +220,6 @@ adopting this exact FastAPI app. The important production requirements are:
 - HTTPS endpoint reachable from Weave.
 - Host allowlist configured if the Weave deployment requires it.
 - Bearer-token validation implemented with your identity/security standards.
-- HTTP 200 response body shaped as `{"schema_version": 1, "result": ...}`.
+- HTTP 200 response body shaped as
+  `{"schema_version": 1, "result": {"value": 0.9, "reason": "...", "confidence": 1.0}}`.
 - Optional dedupe uses `Idempotency-Key`.
diff --git a/examples/remote_scorer/scoring_logic.py b/examples/remote_scorer/scoring_logic.py
index 38f911ac8f74..eaf48a3ac7b5 100644
--- a/examples/remote_scorer/scoring_logic.py
+++ b/examples/remote_scorer/scoring_logic.py
@@ -5,9 +5,11 @@
 from typing import Any
 
 REMOTE_SCORER_SCHEMA_VERSION = 1
+CONCISE_MESSAGE_LENGTH = 120
+TOO_LONG_MESSAGE_LENGTH = 500
 
 
-def score_remote_call(request_body: dict[str, Any]) -> dict[str, Any]:
+def score_remote_call(request_body: dict[str, Any]) -> list[dict[str, Any]]:
     """Score one Weave remote scorer request.
 
     Replace this function with your real policy, model, or business logic. It is
@@ -25,7 +27,32 @@ def score_remote_call(request_body: dict[str, Any]) -> dict[str, Any]:
     message = inputs.get("message", "") if isinstance(inputs, dict) else ""
     if not isinstance(message, str):
         message = ""
-
-    return {
-        "message_length": len(message),
-    }
+    message_length = len(message)
+
+    if message_length <= CONCISE_MESSAGE_LENGTH:
+        conciseness_rating = 1.0
+        length_tag = "concise"
+    elif message_length >= TOO_LONG_MESSAGE_LENGTH:
+        conciseness_rating = 0.0
+        length_tag = "too-long"
+    else:
+        conciseness_rating = 1 - (
+            (message_length - CONCISE_MESSAGE_LENGTH)
+            / (TOO_LONG_MESSAGE_LENGTH - CONCISE_MESSAGE_LENGTH)
+        )
+        length_tag = "verbose"
+
+    return [
+        {
+            "value": round(conciseness_rating, 2),
+            "reason": (
+                f"Message is {message_length} characters; concise messages score best."
+            ),
+            "confidence": 1.0,
+        },
+        {
+            "value": length_tag,
+            "reason": f"Message length category is {length_tag}.",
+            "confidence": 0.9,
+        },
+    ]