diff --git a/csv_detective/detection/formats.py b/csv_detective/detection/formats.py
index 616cc82..a453e6a 100755
--- a/csv_detective/detection/formats.py
+++ b/csv_detective/detection/formats.py
@@ -40,6 +40,13 @@ def detect_formats(
     if len(formats) == 0:
         return analysis, None
 
+    # filter parent-children to only active formats (tags may exclude some)
+    active_parent_children = {
+        parent: [c for c in children if c in formats]
+        for parent, children in fmtm._children.items()
+        if parent in formats
+    }
+
     # Perform testing on fields
     if not in_chunks:
         # table is small enough to be tested in one go
@@ -49,6 +56,7 @@ def detect_formats(
             limited_output=limited_output,
             skipna=skipna,
             verbose=verbose,
+            parent_children=active_parent_children,
         )
         handle_empty_columns(scores_table_fields)
         res_categorical, _ = detect_categorical_variable(
@@ -67,6 +75,7 @@ def detect_formats(
             limited_output=limited_output,
             skipna=skipna,
             verbose=verbose,
+            parent_children=active_parent_children,
         )
     analysis["columns_fields"] = prepare_output_dict(scores_table_fields, limited_output)
 
diff --git a/csv_detective/format.py b/csv_detective/format.py
index 8080757..292c5fe 100755
--- a/csv_detective/format.py
+++ b/csv_detective/format.py
@@ -1,3 +1,4 @@
+from collections import defaultdict
 from typing import Any, Callable
 
 from csv_detective.parsing.text import header_score
@@ -15,6 +16,7 @@ def __init__(
         tags: list[str] = [],
         mandatory_label: bool = False,
         python_type: str = "string",
+        parent: str | None = None,
     ) -> None:
         """
         Instanciates a Format object.
@@ -39,6 +41,7 @@ def __init__(
         self.tags: list[str] = tags
         self.mandatory_label: bool = mandatory_label
         self.python_type: str = python_type
+        self.parent: str | None = parent
 
     def is_valid_label(self, val: str) -> float:
         return header_score(val, self.labels)
@@ -82,7 +85,7 @@ def __init__(
                 _test_values=module._test_values,
                 **{
                     attr: val
-                    for attr in ["labels", "description", "tags", "mandatory_label", "python_type"]
+                    for attr in ["labels", "description", "tags", "mandatory_label", "python_type", "parent"]
                     if (val := getattr(module, attr, None))
                 }
                 | {
@@ -100,6 +103,25 @@ def __init__(
             )
             for label in format_labels
         }
+        self._children: dict[str, list[str]] = defaultdict(list)
+        for name, fmt in self.formats.items():
+            if fmt.parent is not None:
+                if fmt.parent not in self.formats:
+                    raise ValueError(
+                        f"Format '{name}' declares parent '{fmt.parent}' which does not exist"
+                    )
+                self._children[fmt.parent].append(name)
+
+    def get_leaf_formats(self) -> dict[str, Format]:
+        return {name: fmt for name, fmt in self.formats.items() if name not in self._children}
+
+    def get_ancestors(self, format_name: str) -> list[str]:
+        ancestors = []
+        current = format_name
+        while (parent := self.formats[current].parent) is not None:
+            ancestors.append(parent)
+            current = parent
+        return ancestors
 
     def get_formats_from_tags(self, tags: list[str]) -> dict[str, Format]:
         return {
diff --git a/csv_detective/formats/code_epci.py b/csv_detective/formats/code_epci.py
index 11accb4..0b92f6f 100755
--- a/csv_detective/formats/code_epci.py
+++ b/csv_detective/formats/code_epci.py
@@ -1,6 +1,7 @@
 from csv_detective.formats.siren import _is as is_siren
 
 proportion = 0.9
+parent = "siren"
 description = "French EPCI (group of communes) code, subgroup of SIREN"
 tags = ["fr", "geo"]
 mandatory_label = True
diff --git a/csv_detective/formats/geojson.py b/csv_detective/formats/geojson.py
index e131804..d7e0d24 100755
--- a/csv_detective/formats/geojson.py
+++ b/csv_detective/formats/geojson.py
@@ -1,6 +1,7 @@
 import json
 
 proportion = 1
+parent = "json"
 description = "JSON object in the [GeoJSON](https://fr.wikipedia.org/wiki/GeoJSON) format"
 tags = ["geo"]
 python_type = "json"
diff --git a/csv_detective/formats/latitude_l93.py b/csv_detective/formats/latitude_l93.py
index e18010c..b47d4ba 100755
--- a/csv_detective/formats/latitude_l93.py
+++ b/csv_detective/formats/latitude_l93.py
@@ -5,6 +5,7 @@
 from csv_detective.formats.latitude_wgs import SHARED_LATITUDE_LABELS
 
 proportion = 1
+parent = "float"
 description = "Latitude in the Lambert 93 format"
 tags = ["fr", "geo"]
 mandatory_label = True
diff --git a/csv_detective/formats/latitude_wgs.py b/csv_detective/formats/latitude_wgs.py
index 0df4a80..1008913 100755
--- a/csv_detective/formats/latitude_wgs.py
+++ b/csv_detective/formats/latitude_wgs.py
@@ -2,6 +2,7 @@
 from csv_detective.formats.int import _is as is_int
 
 proportion = 1
+parent = "float"
 description = "Latitude in the WGS format"
 tags = ["geo"]
 mandatory_label = True
diff --git a/csv_detective/formats/latitude_wgs_fr_metropole.py b/csv_detective/formats/latitude_wgs_fr_metropole.py
index a68dced..a130f6d 100755
--- a/csv_detective/formats/latitude_wgs_fr_metropole.py
+++ b/csv_detective/formats/latitude_wgs_fr_metropole.py
@@ -1,6 +1,7 @@
 from csv_detective.formats.latitude_wgs import _is as is_latitude, labels  # noqa
 
 proportion = 1
+parent = "latitude_wgs"
 description = "Latitude within the French metropole bounds in the WGS format"
 tags = ["fr", "geo"]
 mandatory_label = True
diff --git a/csv_detective/formats/longitude_l93.py b/csv_detective/formats/longitude_l93.py
index f3ad521..101aed8 100755
--- a/csv_detective/formats/longitude_l93.py
+++ b/csv_detective/formats/longitude_l93.py
@@ -5,6 +5,7 @@
 from csv_detective.formats.longitude_wgs import SHARED_LONGITUDE_LABELS
 
 proportion = 1
+parent = "float"
 description = "Longitude in the Lambert 93 format"
 tags = ["fr", "geo"]
 mandatory_label = True
diff --git a/csv_detective/formats/longitude_wgs.py b/csv_detective/formats/longitude_wgs.py
index 48f6a82..b2d262e 100755
--- a/csv_detective/formats/longitude_wgs.py
+++ b/csv_detective/formats/longitude_wgs.py
@@ -2,6 +2,7 @@
 from csv_detective.formats.int import _is as is_int
 
 proportion = 1
+parent = "float"
 description = "Longitude in the WGS format"
 tags = ["geo"]
 mandatory_label = True
diff --git a/csv_detective/formats/longitude_wgs_fr_metropole.py b/csv_detective/formats/longitude_wgs_fr_metropole.py
index 883829d..9323c24 100755
--- a/csv_detective/formats/longitude_wgs_fr_metropole.py
+++ b/csv_detective/formats/longitude_wgs_fr_metropole.py
@@ -1,6 +1,7 @@
 from csv_detective.formats.longitude_wgs import _is as is_longitude, labels  # noqa
 
 proportion = 1
+parent = "longitude_wgs"
 description = "Longitude within the French metropole bounds in the WGS format"
 tags = ["fr", "geo"]
 mandatory_label = True
diff --git a/csv_detective/parsing/columns.py b/csv_detective/parsing/columns.py
index f29d370..1e37448 100755
--- a/csv_detective/parsing/columns.py
+++ b/csv_detective/parsing/columns.py
@@ -75,17 +75,25 @@ def test_col(
     limited_output: bool,
     skipna: bool = True,
     verbose: bool = False,
+    parent_children: dict[str, list[str]] | None = None,
 ):
     if verbose:
         start = time()
         logging.info("Testing columns to get formats")
     return_table = pd.DataFrame(columns=table.columns)
-    for idx, (label, format) in enumerate(formats.items()):
+
+    if parent_children:
+        parent_names = set(parent_children.keys())
+        leaf_formats = {k: v for k, v in formats.items() if k not in parent_names}
+        parent_formats = {k: v for k, v in formats.items() if k in parent_names}
+    else:
+        leaf_formats = formats
+        parent_formats = {}
+
+    for idx, (label, format) in enumerate(leaf_formats.items()):
         if verbose:
             start_type = time()
             logging.info(f"\t- Starting with format '{label}'")
-        # improvement lead : put the longest tests behind and make them only if previous tests not satisfactory
-        # => the following needs to change, "apply" means all columns are tested for one type at once
         for col in table.columns:
             return_table.loc[label, col] = test_col_val(
                 table[col],
@@ -96,9 +104,38 @@ def test_col(
             )
         if verbose:
             display_logs_depending_process_time(
-                f'\t> Done with format "{label}" in {round(time() - start_type, 3)}s ({idx + 1}/{len(formats)})',
+                f'\t> Done with format "{label}" in {round(time() - start_type, 3)}s ({idx + 1}/{len(leaf_formats)})',
+                time() - start_type,
+            )
+
+    for label, format in parent_formats.items():
+        if verbose:
+            start_type = time()
+            logging.info(f"\t- Propagating to parent format '{label}'")
+        children = parent_children[label]
+        for col in table.columns:
+            child_scores = [
+                return_table.loc[child, col]
+                for child in children
+                if child in return_table.index
+            ]
+            max_child = max(child_scores) if child_scores else 0
+            if max_child > 0:
+                return_table.loc[label, col] = max_child
+            else:
+                return_table.loc[label, col] = test_col_val(
+                    table[col],
+                    format,
+                    skipna=skipna,
+                    limited_output=limited_output,
+                    verbose=verbose,
+                )
+        if verbose:
+            display_logs_depending_process_time(
+                f'\t> Done with parent format "{label}" in {round(time() - start_type, 3)}s',
                 time() - start_type,
             )
+
     if verbose:
         display_logs_depending_process_time(
             f"Done testing columns in {round(time() - start, 3)}s", time() - start
@@ -138,6 +175,7 @@ def test_col_chunks(
     limited_output: bool,
     skipna: bool = True,
     verbose: bool = False,
+    parent_children: dict[str, list[str]] | None = None,
 ) -> tuple[pd.DataFrame, dict, dict[str, pd.Series]]:
     def build_remaining_tests_per_col(return_table: pd.DataFrame) -> dict[str, list[str]]:
         # returns a dict with the table's columns as keys and the list of remaining format labels to apply
@@ -156,7 +194,10 @@ def build_remaining_tests_per_col(return_table: pd.DataFrame) -> dict[str, list[
         logging.info("Testing columns to get formats on chunks")
 
     # analysing the sample to get a first guess
-    return_table = test_col(table, formats, limited_output, skipna=skipna, verbose=verbose)
+    return_table = test_col(
+        table, formats, limited_output, skipna=skipna, verbose=verbose,
+        parent_children=parent_children,
+    )
     # mandatory_label formats are zeroed out at the end if the label doesn't match,
     # so there's no point running the expensive field tests on those columns
     mandatory_label_skip: dict[str, set[str]] = {
@@ -229,10 +270,12 @@ def build_remaining_tests_per_col(return_table: pd.DataFrame) -> dict[str, list[
         if not any(remaining_tests for remaining_tests in remaining_tests_per_col.values()):
             # no more potential tests to do on any column, early stop
             break
+        parent_names = set(parent_children.keys()) if parent_children else set()
         for col, fmt_labels in remaining_tests_per_col.items():
-            # testing each column with the tests that are still competing
-            # after previous batchs analyses
+            # testing each column with the leaf tests that are still competing
             for label in fmt_labels:
+                if label in parent_names:
+                    continue
                 batch_col_test = test_col_val(
                     batch[col],
                     formats[label],
@@ -240,12 +283,37 @@ def build_remaining_tests_per_col(return_table: pd.DataFrame) -> dict[str, list[
                     skipna=skipna,
                 )
                 return_table.loc[label, col] = (
-                    # if this batch's column tested 0 then test fails overall
                     0
                     if batch_col_test == 0
-                    # otherwise updating the score with weighted average
                     else ((return_table.loc[label, col] * idx + batch_col_test) / (idx + 1))
                 )
+            # propagate to parent formats
+            for parent_label in fmt_labels:
+                if parent_label not in parent_names:
+                    continue
+                children = parent_children[parent_label]
+                child_scores = [
+                    return_table.loc[child, col]
+                    for child in children
+                    if child in return_table.index
+                ]
+                max_child = max(child_scores) if child_scores else 0
+                if max_child > 0:
+                    return_table.loc[parent_label, col] = max_child
+                else:
+                    batch_col_test = test_col_val(
+                        batch[col],
+                        formats[parent_label],
+                        limited_output=limited_output,
+                        skipna=skipna,
+                    )
+                    return_table.loc[parent_label, col] = (
+                        0
+                        if batch_col_test == 0
+                        else (
+                            (return_table.loc[parent_label, col] * idx + batch_col_test) / (idx + 1)
+                        )
+                    )
         remaining_tests_per_col = build_remaining_tests_per_col(return_table)
         batch, batch_number = [], batch_number + 1
     analysis["nb_duplicates"] = sum(row_hashes_count > 1)
diff --git a/tests/test_fields.py b/tests/test_fields.py
index 0afbede..8a9962c 100644
--- a/tests/test_fields.py
+++ b/tests/test_fields.py
@@ -160,6 +160,34 @@ def test_early_detection(args):
         mock_func.assert_not_called()
 
 
+def test_parent_score_propagation():
+    # latitude_wgs_fr_metropole values are valid floats,
+    # so float should get a propagated score without being tested directly
+    lat_values = fmtm.formats["latitude_wgs_fr_metropole"]._test_values[True]
+    table = pd.DataFrame({"lat": (lat_values * 100)[:100]})
+    parent_children = {
+        parent: [c for c in children]
+        for parent, children in fmtm._children.items()
+    }
+    returned_table = col_test(table, fmtm.formats, limited_output=True, parent_children=parent_children)
+    assert returned_table.loc["latitude_wgs_fr_metropole", "lat"] > 0
+    assert returned_table.loc["latitude_wgs", "lat"] > 0
+    assert returned_table.loc["float", "lat"] > 0
+
+
+def test_parent_tested_when_children_fail():
+    # float values that are NOT latitudes should still get float detected
+    table = pd.DataFrame({"val": ["999.99", "-500.5", "123456.789"] * 34})
+    parent_children = {
+        parent: [c for c in children]
+        for parent, children in fmtm._children.items()
+    }
+    returned_table = col_test(table, fmtm.formats, limited_output=True, parent_children=parent_children)
+    assert returned_table.loc["float", "val"] > 0
+    assert returned_table.loc["latitude_wgs", "val"] == 0
+    assert returned_table.loc["latitude_wgs_fr_metropole", "val"] == 0
+
+
 def test_all_proportion_1():
     # building a table that uses only correct values for these formats, except on one row
     table = pd.DataFrame(
diff --git a/tests/test_structure.py b/tests/test_structure.py
index 8bf65b9..3358483 100755
--- a/tests/test_structure.py
+++ b/tests/test_structure.py
@@ -52,6 +52,36 @@ def test_get_from_tags(tags):
             assert tag in fmt.tags
 
 
+def test_parent_references_valid():
+    for name, fmt in fmtm.formats.items():
+        if fmt.parent is not None:
+            assert fmt.parent in fmtm.formats, (
+                f"Format '{name}' declares parent '{fmt.parent}' which does not exist"
+            )
+
+
+def test_true_subset_invariant():
+    for name, fmt in fmtm.formats.items():
+        if fmt.parent is None:
+            continue
+        parent = fmtm.formats[fmt.parent]
+        for val in fmt._test_values[True]:
+            assert parent.func(val), (
+                f"'{val}' is valid for '{name}' but not for parent '{fmt.parent}'"
+            )
+
+
+def test_leaf_formats_excludes_parents():
+    leaves = fmtm.get_leaf_formats()
+    for parent_name in fmtm._children:
+        assert parent_name not in leaves
+
+
+def test_get_ancestors():
+    ancestors = fmtm.get_ancestors("latitude_wgs_fr_metropole")
+    assert ancestors == ["latitude_wgs", "float"]
+
+
 @pytest.mark.parametrize(
     "func, max_pos_args",
     (