diff --git a/etl/customers/stonewater/Wave 3 Preparation.py b/etl/customers/stonewater/Wave 3 Preparation.py
index 5ebb06e26..974cd9084 100644
--- a/etl/customers/stonewater/Wave 3 Preparation.py	
+++ b/etl/customers/stonewater/Wave 3 Preparation.py	
@@ -1719,6 +1719,72 @@ def propsed_wave_3_sample():
     # Tier 2: We have a property in the same archetype that was surveyed and is below EPC D
     #
 
+    def match_property_to_surveyed(property, survey_results_with_original_features):
+        surveyed = survey_results_with_original_features[
+            (
+                survey_results_with_original_features["Property Type"] ==
+                property["Property Type"]
+            ) &
+            (
+                survey_results_with_original_features["Wall Type"] ==
+                property["Wall Type"]
+            ) &
+            (
+                survey_results_with_original_features["Roof Type"] ==
+                property["Roof Type"]
+            ) &
+            (
+                survey_results_with_original_features["Heating"] ==
+                property["Heating"]
+            )
+            ].copy()
+
+        if not surveyed.empty:
+            return surveyed
+
+        surveyed = survey_results_with_original_features[
+            (
+                survey_results_with_original_features["Property Type"] ==
+                property["Property Type"]
+            ) &
+            (
+                survey_results_with_original_features["Wall Type"] ==
+                property["Wall Type"]
+            ) &
+            (
+                survey_results_with_original_features["Roof Type"].str.split(":").str[0] ==
+                property["Roof Type"].split(":")[0]
+            ) &
+            (
+                survey_results_with_original_features["Heating"] ==
+                property["Heating"]
+            )
+            ].copy()
+
+        if not surveyed.empty:
+            return surveyed
+
+        surveyed = survey_results_with_original_features[
+            (
+                survey_results_with_original_features["Property Type"] ==
+                property["Property Type"]
+            ) &
+            (
+                survey_results_with_original_features["Wall Type"] ==
+                property["Wall Type"]
+            ) &
+            (
+                survey_results_with_original_features["Roof Type"].str.split(":").str[0] ==
+                property["Roof Type"].split(":")[0]
+            ) &
+            (
+                survey_results_with_original_features["Heating"].str.split(":").str[0] ==
+                property["Heating"].split(":")[0]
+            )
+            ].copy()
+
+        return surveyed
+
     results = []
     for region in tqdm(unique_postal_regions):
         # Take all of the properties in that region
@@ -1757,6 +1823,7 @@ def propsed_wave_3_sample():
             ][["Archetype ID", "Current EPC Band"]].drop_duplicates()
 
         if region_surveyed["Archetype ID"].duplicated().sum():
+
             region_surveyed = []
             for arch_id in archetypes:
                 for _, property in region_assets[region_assets["Archetype ID"] == arch_id].iterrows():
@@ -1765,6 +1832,12 @@ def propsed_wave_3_sample():
                         ].copy()
                     if archetype_data.empty:
                         continue
+                    if archetype_data.shape[0] > 1:
+                        # Look for an exact match, or as close as possible
+                        archetype_data_filtered = match_property_to_surveyed(property, archetype_data)
+                        if not archetype_data_filtered.empty:
+                            archetype_data = archetype_data_filtered
+
                     archetype_data["distance_meters"] = haversine(
                         lat1=property.latitude, lon1=property.longitude,
                         lat2=archetype_data["latitude"].values, lon2=archetype_data["longitude"].values
@@ -1899,28 +1972,15 @@ def propsed_wave_3_sample():
             # This means that this archetype was never surveyed and so we need to find a sufficiently similar property
         final_missed_matches = []
         for a_id in missed_addressids:
+
+            match_type = "3 - compared to similar properties"
+
             property = asset_list[asset_list["Address ID"] == a_id].squeeze()
 
-            surveyed = survey_results_with_original_features[
-                (
-                    survey_results_with_original_features["Property Type"] ==
-                    property["Property Type"]
-                ) &
-                (
-                    survey_results_with_original_features["Wall Type"] ==
-                    property["Wall Type"]
-                ) &
-                (
-                    survey_results_with_original_features["Roof Type"] ==
-                    property["Roof Type"]
-                ) &
-                (
-                    survey_results_with_original_features["Heating"] ==
-                    property["Heating"]
-                )
-                ].copy()
+            surveyed = match_property_to_surveyed(property, survey_results_with_original_features)
 
             if surveyed.empty:
+                match_type = "3 - compared to similar properties, relaxed"
                 # In this case, we do one additional check where we filter on everything the same apart from heating,
                 # where we do a slightly more rough match
                 surveyed = survey_results_with_original_features[
@@ -2026,14 +2086,12 @@ def propsed_wave_3_sample():
             expected_epc = sap_to_epc(expected_sap)
 
             if expected_epc in ["C", "B", "A"]:
-                tier = "5 - EPC C or above"
-            else:
-                tier = "3 - similar property, weighted on distance"
+                match_type = "5 - EPC C or above"
 
             final_missed_matches.append(
                 {
                     "Address ID": a_id,
-                    "Confidence Tier": tier,
+                    "Confidence Tier": match_type,
                     "Current EPC Band": expected_epc
                 }
             )
@@ -2197,22 +2255,9 @@ def propsed_wave_3_sample():
     # '2 - same archetype',
     # '3 - similar property, weighted on distance'
 
-    gain_columns = [
-        '1 - Archetype surveyed',
-        '1 - property was surveyed',
-        '2 - same archetype',
-        '3 - similar property, weighted on distance'
-    ]
-    #
-    # Loss is the sum of these columns:
-    # '4 - no similar property, needs survey to confirm',
-    # '5 - EPC C or above', '5 - property was surveyed'
+    gain_columns = sorted([x for x in results["Confidence Tier"].unique() if "1 - " in x or "2 - " in x or "3 - " in x])
+    loss_columns = sorted([x for x in results["Confidence Tier"].unique() if "4 - " in x or "5 - " in x])
 
-    loss_columns = [
-        '4 - no similar property, needs survey to confirm',
-        '5 - EPC C or above',
-        '5 - property was surveyed'
-    ]
     geographic_summary["Gain"] = geographic_summary[gain_columns].sum(axis=1)
     geographic_summary["Loss"] = geographic_summary[loss_columns].sum(axis=1)
 
@@ -2283,7 +2328,7 @@ def propsed_wave_3_sample():
 
     # Remaining loss allowed
     # remaining_loss_constraint = 230 - region_totals["Loss"]
-    remaining_loss_constraint = 250
+    remaining_loss_constraint = 220
     postcode_selected_rows, _ = optimise(
         gain=postcode_summary_unselected_regions["Gain"].values,
         loss=postcode_summary_unselected_regions["Loss"].values,