diff --git a/cve_bin_tool/data_sources/nvd_source.py b/cve_bin_tool/data_sources/nvd_source.py
index 1ee0a4d34b..54f5395fdc 100644
--- a/cve_bin_tool/data_sources/nvd_source.py
+++ b/cve_bin_tool/data_sources/nvd_source.py
@@ -310,12 +310,15 @@ def format_data_api2(self, all_cve_entries):
                 )
                 cve["severity"] = re.sub(r"[\W]", "", cve["severity"])
 
-            # score should be numeric
-            try:
-                cve["score"] = float(cve["score"])
-            except ValueError:
-                self.logger.debug(f"Score for {cve['id']} is invalid: {cve['score']}")
-                cve["score"] = "invalid"
+            # score should be numeric, except when score is "unknown"
+            if cve["score"] != "unknown":
+                try:
+                    cve["score"] = float(cve["score"])
+                except ValueError:
+                    self.logger.debug(
+                        f"Score for {cve['id']} is invalid: {cve['score']}"
+                    )
+                    cve["score"] = "invalid"
 
             # CVSS_vector will be validated/normalized when cvss library is used but
             # we can at least do a character filter here