diff --git a/contrib/msggen/msggen/schema.json b/contrib/msggen/msggen/schema.json
index 15babcb2f..18a81d7ba 100644
--- a/contrib/msggen/msggen/schema.json
+++ b/contrib/msggen/msggen/schema.json
@@ -309,6 +309,12 @@
               "The bias, positive being good and negative being bad (0 being no bias).  Useful values are +/-1 through +/-10, though -100 through +100 are possible values."
             ]
           },
+          "description": {
+            "type": "string",
+            "description": [
+              "Description/annotation to display in askrene-listlayers(7)"
+            ]
+          },
           "relative": {
             "type": "boolean",
             "added": "v25.05",
@@ -316,12 +322,6 @@
             "description": [
               "The bias will be added to the previous value."
             ]
-          },
-          "description": {
-            "type": "string",
-            "description": [
-              "Description/annotation to display in askrene-listlayers(7)"
-            ]
           }
         }
       },
diff --git a/doc/schemas/askrene-bias-channel.json b/doc/schemas/askrene-bias-channel.json
index 1e3091336..73eddb5bf 100644
--- a/doc/schemas/askrene-bias-channel.json
+++ b/doc/schemas/askrene-bias-channel.json
@@ -33,6 +33,12 @@
           "The bias, positive being good and negative being bad (0 being no bias).  Useful values are +/-1 through +/-10, though -100 through +100 are possible values."
         ]
       },
+      "description": {
+        "type": "string",
+        "description": [
+          "Description/annotation to display in askrene-listlayers(7)"
+        ]
+      },
       "relative": {
         "type": "boolean",
         "added": "v25.05",
@@ -40,12 +46,6 @@
         "description": [
           "The bias will be added to the previous value."
         ]
-      },
-      "description": {
-        "type": "string",
-        "description": [
-          "Description/annotation to display in askrene-listlayers(7)"
-        ]
       }
     }
   },
diff --git a/plugins/askrene/askrene.c b/plugins/askrene/askrene.c
index 0e2f2f3fc..14a6e8964 100644
--- a/plugins/askrene/askrene.c
+++ b/plugins/askrene/askrene.c
@@ -1104,8 +1104,8 @@ static struct command_result *json_askrene_bias_channel(struct command *cmd,
 		   p_req("layer", param_known_layer, &layer),
 		   p_req("short_channel_id_dir", param_short_channel_id_dir, &scidd),
 		   p_req("bias", param_s8_hundred, &bias),
-		   p_opt_def("relative", param_bool, &relative, false),
 		   p_opt("description", param_string, &description),
+		   p_opt_def("relative", param_bool, &relative, false),
 		   NULL))
 		return command_param_failed();
 	plugin_log(cmd->plugin, LOG_TRACE, "%s called: %.*s", __func__,
diff --git a/tests/test_askrene.py b/tests/test_askrene.py
index 7404d4664..1c3787c0d 100644
--- a/tests/test_askrene.py
+++ b/tests/test_askrene.py
@@ -277,6 +277,23 @@ def test_layers(node_factory):
     with pytest.raises(RpcError, match="bias: should be a number between -100 and 100"):
         l2.rpc.askrene_bias_channel('test_layers', '1x1x1/1', 101, "bigger bias")
 
+    # We can make them relative.
+    l2.rpc.askrene_bias_channel('test_layers', '1x1x1/1', 1, 'adding bias', True)
+    expect['biases'] = [{'short_channel_id_dir': '1x1x1/1', 'bias': -4, 'description': "adding bias"}]
+    listlayers = l2.rpc.askrene_listlayers('test_layers')
+    assert listlayers == {'layers': [expect]}
+
+    l2.rpc.askrene_bias_channel(layer='test_layers', short_channel_id_dir='1x1x1/1', bias=-1, relative=True)
+    expect['biases'] = [{'short_channel_id_dir': '1x1x1/1', 'bias': -5}]
+    listlayers = l2.rpc.askrene_listlayers('test_layers')
+    assert listlayers == {'layers': [expect]}
+
+    # They truncate on +/- 100 though:
+    l2.rpc.askrene_bias_channel('test_layers', '1x1x1/1', -99, None, True)
+    expect['biases'] = [{'short_channel_id_dir': '1x1x1/1', 'bias': -100}]
+    listlayers = l2.rpc.askrene_listlayers('test_layers')
+    assert listlayers == {'layers': [expect]}
+
     # We can remove them.
     l2.rpc.askrene_bias_channel('test_layers', '1x1x1/1', 0)
     expect['biases'] = []