diff --git a/litellm/integrations/focus/transformer.py b/litellm/integrations/focus/transformer.py
index c22be2e66fc..d4a29d4ccdc 100644
--- a/litellm/integrations/focus/transformer.py
+++ b/litellm/integrations/focus/transformer.py
@@ -129,7 +129,12 @@ class FocusTransformer:
                 & (pl.col("model_group").cast(pl.String) != "")
             )
             .then(pl.col("model_group").cast(pl.String))
-            .otherwise(pl.col("model").cast(pl.String))
+            .when(
+                pl.col("model").cast(pl.String).is_not_null()
+                & (pl.col("model").cast(pl.String) != "")
+            )
+            .then(pl.col("model").cast(pl.String))
+            .otherwise(pl.lit("unknown"))
             .alias("ServiceName"),
             pl.col("team_id").cast(pl.String).alias("SubAccountId"),
             pl.col("team_alias").cast(pl.String).alias("SubAccountName"),
diff --git a/tests/test_litellm/integrations/focus/test_transformer.py b/tests/test_litellm/integrations/focus/test_transformer.py
new file mode 100644
index 00000000000..53e54e83799
--- /dev/null
+++ b/tests/test_litellm/integrations/focus/test_transformer.py
@@ -0,0 +1,78 @@
+"""Tests for FocusTransformer."""
+
+from __future__ import annotations
+
+import json
+
+import polars as pl
+
+from litellm.integrations.focus.transformer import FocusTransformer
+
+
+def _make_row(**overrides) -> dict:
+    """Return a minimal valid spend-log row, with optional overrides."""
+    base = {
+        "date": "2026-03-17",
+        "spend": 0.05,
+        "api_key": "sk-test",
+        "api_key_alias": "my-key",
+        "model": "claude-sonnet",
+        "model_group": "sonnet-group",
+        "custom_llm_provider": "anthropic",
+        "team_id": "team-1",
+        "team_alias": "Engineering",
+        "user_id": "user-1",
+        "user_email": "test@example.com",
+        "prompt_tokens": 100,
+        "completion_tokens": 50,
+    }
+    base.update(overrides)
+    return base
+
+
+def test_service_name_uses_model_group_when_present():
+    row = _make_row(model_group="sonnet-group", model="claude-sonnet")
+    frame = pl.DataFrame([row])
+    result = FocusTransformer().transform(frame)
+    assert result["ServiceName"][0] == "sonnet-group"
+
+
+def test_service_name_falls_back_to_model_when_model_group_blank():
+    row = _make_row(model_group="", model="claude-sonnet")
+    frame = pl.DataFrame([row])
+    result = FocusTransformer().transform(frame)
+    assert result["ServiceName"][0] == "claude-sonnet"
+
+
+def test_service_name_falls_back_to_model_when_model_group_null():
+    row = _make_row(model_group=None, model="claude-sonnet")
+    frame = pl.DataFrame([row])
+    result = FocusTransformer().transform(frame)
+    assert result["ServiceName"][0] == "claude-sonnet"
+
+
+def test_service_name_defaults_to_unknown_when_both_blank():
+    row = _make_row(model_group="", model="")
+    frame = pl.DataFrame([row])
+    result = FocusTransformer().transform(frame)
+    assert result["ServiceName"][0] == "unknown"
+
+
+def test_tags_include_token_counts():
+    row = _make_row(prompt_tokens=100, completion_tokens=50)
+    frame = pl.DataFrame([row])
+    result = FocusTransformer().transform(frame)
+    tags = json.loads(result["Tags"][0])
+    assert tags["prompt_tokens"] == "100"
+    assert tags["completion_tokens"] == "50"
+
+
+def test_tags_include_all_metadata_keys():
+    row = _make_row()
+    frame = pl.DataFrame([row])
+    result = FocusTransformer().transform(frame)
+    tags = json.loads(result["Tags"][0])
+    assert "team_id" in tags
+    assert "model" in tags
+    assert "prompt_tokens" in tags
+    assert "completion_tokens" in tags