diff --git a/scripts/check_schema_drift.py b/scripts/check_schema_drift.py
index 09a7bcf..27e9b8a 100644
--- a/scripts/check_schema_drift.py
+++ b/scripts/check_schema_drift.py
@@ -1,36 +1,107 @@
 import os
-from sqlalchemy import text
 import sys
+
+from sqlalchemy import text
+
 from ngm.database.models import get_engine
 
+# A blank case_type only signals selector/schema drift for *litigation* cases.
+# Non-litigation registrations (marriage, power-of-attorney, guardianship,
+# adoption, death-declaration) are filed under a numeric case-number namespace
+# (e.g. 082-02-0001) and legitimately carry no case_type. Litigation cases use
+# an alphabetic namespace (CP, FN, CR, WO, C1, ...). We therefore only count
+# blanks among alpha-coded rows.
+LITIGATION_ONLY = "split_part(case_number, '-', 2) ~ '[A-Za-z]'"
+
+# Only look at rows ingested recently: drift means *this* scrape stopped
+# capturing case_type, so a ratio over the whole table history would dilute a
+# fresh failure into insignificance (and needlessly scan millions of rows).
+RECENT_WINDOW_DAYS = 3
+
+# Fail if more than this fraction of recent litigation rows lack a case_type.
+DRIFT_RATIO_THRESHOLD = 0.02
+
+# Don't fail on a handful of rows: a tiny recent sample makes the ratio noisy.
+MIN_SAMPLE = 20
+
+
+def _validate_courts(conn, courts):
+    """Error out on unknown court identifiers so a typo can't silently pass."""
+    known = {
+        r.court_identifier
+        for r in conn.execute(
+            text(
+                "SELECT DISTINCT court_identifier FROM court_cases "
+                "WHERE court_identifier = ANY(:courts)"
+            ),
+            {"courts": list(courts)},
+        )
+    }
+    unknown = sorted(set(courts) - known)
+    if unknown:
+        print(f"ERROR: unknown court identifier(s): {', '.join(unknown)}")
+        sys.exit(1)
+
 
-def check():
+def check(courts=None):
     db_url = os.getenv("DATABASE_URL")
     if not db_url:
         print("DATABASE_URL not set")
         sys.exit(1)
 
+    where = [
+        LITIGATION_ONLY,
+        "created_at >= now() - (:days * interval '1 day')",
+    ]
+    params = {"days": RECENT_WINDOW_DAYS}
+    if courts:
+        where.append("court_identifier = ANY(:courts)")
+        params["courts"] = list(courts)
+    scope = " AND ".join(where)
+
     engine = get_engine(db_url)
     try:
         with engine.connect() as conn:
-            # Check if we have cases with null case_type or other expected fields
-            # suggesting a selector failure (schema drift)
-            result = conn.execute(
+            if courts:
+                _validate_courts(conn, courts)
+
+            row = conn.execute(
                 text(
-                    "SELECT count(*) FROM court_cases WHERE case_type IS NULL OR case_type = ''"
+                    f"SELECT count(*) AS total, "
+                    f"count(*) FILTER (WHERE case_type IS NULL OR case_type = '') "
+                    f"AS missing FROM court_cases WHERE {scope}"
+                ),
+                params,
+            ).one()
+            total, missing = row.total, row.missing
+
+            scope_label = f"courts={list(courts)}" if courts else "all courts"
+            window_label = f"last {RECENT_WINDOW_DAYS}d, {scope_label}"
+            if total < MIN_SAMPLE:
+                print(
+                    f"Schema check skipped: only {total} recent litigation rows "
+                    f"({window_label}), below sample floor {MIN_SAMPLE}"
                 )
-            )
-            count = result.scalar()
-            if count > 10:  # Threshold for drift detection
+                return
+
+            ratio = missing / total
+            if ratio > DRIFT_RATIO_THRESHOLD:
                 print(
-                    f"ERROR: Detected {count} cases with missing case_type. Possible schema drift!"
+                    f"ERROR: {missing}/{total} ({ratio:.1%}) recent litigation cases "
+                    f"missing case_type ({window_label}), above "
+                    f"{DRIFT_RATIO_THRESHOLD:.0%} threshold. Possible schema drift!"
                 )
                 sys.exit(1)
-            print("Schema check passed")
+            print(
+                f"Schema check passed: {missing}/{total} ({ratio:.2%}) recent "
+                f"litigation cases missing case_type ({window_label})"
+            )
     except Exception as e:
         print(f"Schema check failed: {e}")
         sys.exit(1)
 
 
 if __name__ == "__main__":
-    check()
+    # Optional positional args scope the check to specific court identifiers
+    # (e.g. `python scripts/check_schema_drift.py supreme special`).
+    check(courts=sys.argv[1:] or None)