diff --git a/build_canonical_layer.py b/build_canonical_layer.py
index 3df2caa..d16addf 100644
--- a/build_canonical_layer.py
+++ b/build_canonical_layer.py
@@ -1,4 +1,5 @@
 import click
+import re
 
 from layer_helpers import read_csv_rows, representative_value, stable_id, write_csv_rows
 
@@ -20,6 +21,8 @@ CANONICAL_FIELDS = [
     "updated_at",
 ]
 
+CANONICAL_DROP_TOKENS = {"CT", "COUNT", "COUNTS", "DOZ", "DOZEN", "DOZ.", "PACK"}
+
 LINK_FIELDS = [
     "observed_product_id",
     "canonical_product_id",
@@ -91,26 +94,24 @@ def auto_link_rule(observed_row):
             "high",
         )
 
-    if (
-        observed_row.get("representative_name_norm")
-        and not observed_row.get("representative_size_value")
-        and not observed_row.get("representative_size_unit")
-        and not observed_row.get("representative_pack_qty")
-    ):
-        return (
-            "exact_name",
-            "|".join(
-                [
-                    f"name={observed_row['representative_name_norm']}",
-                    f"measure={observed_row['representative_measure_type']}",
-                ]
-            ),
-            "medium",
-        )
-
     return "", "", ""
 
 
+def clean_canonical_name(name):
+    tokens = []
+    for token in re.sub(r"[^A-Z0-9\s]", " ", (name or "").upper()).split():
+        if token.isdigit():
+            continue
+        if token in CANONICAL_DROP_TOKENS:
+            continue
+        if re.fullmatch(r"\d+(?:PK|PACK)", token):
+            continue
+        if re.fullmatch(r"\d+DZ", token):
+            continue
+        tokens.append(token)
+    return " ".join(tokens).strip()
+
+
 def canonical_row_for_group(canonical_product_id, group_rows, link_method):
     quantity_value, quantity_unit = normalized_quantity(
         {
@@ -130,7 +131,10 @@ def canonical_row_for_group(canonical_product_id, group_rows, link_method):
     )
     return {
         "canonical_product_id": canonical_product_id,
-        "canonical_name": representative_value(group_rows, "representative_name_norm"),
+        "canonical_name": clean_canonical_name(
+            representative_value(group_rows, "representative_name_norm")
+        )
+        or representative_value(group_rows, "representative_name_norm"),
         "product_type": "",
         "brand": representative_value(group_rows, "representative_brand"),
         "variant": representative_value(group_rows, "representative_variant"),
diff --git a/tests/test_canonical_layer.py b/tests/test_canonical_layer.py
index 5b45d44..451c731 100644
--- a/tests/test_canonical_layer.py
+++ b/tests/test_canonical_layer.py
@@ -4,7 +4,7 @@ import build_canonical_layer
 
 
 class CanonicalLayerTests(unittest.TestCase):
-    def test_build_canonical_layer_auto_links_exact_upc_and_name_size(self):
+    def test_build_canonical_layer_auto_links_exact_upc_and_name_size_only(self):
         observed_rows = [
             {
                 "observed_product_id": "gobs_1",
@@ -81,6 +81,21 @@ class CanonicalLayerTests(unittest.TestCase):
                 "is_discount_line": "false",
                 "is_coupon_line": "false",
             },
+            {
+                "observed_product_id": "gobs_6",
+                "representative_upc": "",
+                "representative_retailer_item_id": "",
+                "representative_name_norm": "LIME",
+                "representative_brand": "",
+                "representative_variant": "",
+                "representative_size_value": "",
+                "representative_size_unit": "",
+                "representative_pack_qty": "",
+                "representative_measure_type": "each",
+                "is_fee": "false",
+                "is_discount_line": "false",
+                "is_coupon_line": "false",
+            },
         ]
 
         canonicals, links = build_canonical_layer.build_canonical_layer(observed_rows)
@@ -93,6 +108,11 @@ class CanonicalLayerTests(unittest.TestCase):
         self.assertEqual("exact_name_size", methods["gobs_3"])
         self.assertEqual("exact_name_size", methods["gobs_4"])
         self.assertNotIn("gobs_5", methods)
+        self.assertNotIn("gobs_6", methods)
+
+    def test_clean_canonical_name_removes_packaging_noise(self):
+        self.assertEqual("LIME", build_canonical_layer.clean_canonical_name("LIME  . / ."))
+        self.assertEqual("EGG", build_canonical_layer.clean_canonical_name("5DZ EGG / /"))
 
 
 if __name__ == "__main__":