diff --git a/src/tablib/formats/_ods.py b/src/tablib/formats/_ods.py
index 97cef201..f83b9e0b 100644
--- a/src/tablib/formats/_ods.py
+++ b/src/tablib/formats/_ods.py
@@ -110,22 +110,29 @@ def import_sheet(cls, dset, sheet, headers=True, skip_lines=0):
 
         dset.title = sheet.getAttribute('name')
 
-        def is_real_cell(cell):
-            return cell.hasChildNodes() or not cell.getAttribute('numbercolumnsrepeated')
+        def repeat(cell):
+            try:
+                n = int(cell.getAttribute("numbercolumnsrepeated"))
+            except (AttributeError, TypeError, ValueError):
+                n = 1
+
+            return [cls.read_cell(cell)] * n
 
         rows = (row for row in sheet.childNodes if row.tagName == "table:table-row")
 
         for i, row in enumerate(rows):
             if i < skip_lines:
                 continue
-            row_vals = [cls.read_cell(cell) for cell in row.childNodes if is_real_cell(cell)]
+            row_vals = [c for cell in row.childNodes for c in repeat(cell)]
             if not row_vals:
                 continue
             if i == skip_lines and headers:
-                dset.headers = row_vals
+                dset.headers = limit_row(row_vals)
             else:
-                if i > skip_lines and len(row_vals) < dset.width:
-                    row_vals += [''] * (dset.width - len(row_vals))
+                if i > skip_lines:
+                    if len(row_vals) < dset.width:
+                        row_vals += [''] * (dset.width - len(row_vals))
+                    row_vals = row_vals[:dset.width]
                 dset.append(row_vals)
 
     @classmethod
@@ -248,3 +255,18 @@ def detect(cls, stream):
             return True
         except Exception:
             return False
+
+
+def limit_row(headers, limit=5):
+    """Limits the number of trailing empty row values"""
+    limited = []
+    count = 0
+
+    for v in headers:
+        limited += [v]
+        count = count + 1 if v == "" else 0
+
+        if count == limit:
+            break
+
+    return limited
diff --git a/tests/files/columns_repeated.ods b/tests/files/columns_repeated.ods
new file mode 100644
index 00000000..3829da97
Binary files /dev/null and b/tests/files/columns_repeated.ods differ
diff --git a/tests/test_tablib.py b/tests/test_tablib.py
index 7a89eb7b..adc54ed4 100755
--- a/tests/test_tablib.py
+++ b/tests/test_tablib.py
@@ -1260,7 +1260,16 @@ def test_ods_import_set_ragged(self):
         ods_source = Path(__file__).parent / 'files' / 'ragged.ods'
         with ods_source.open('rb') as fh:
             dataset = tablib.Dataset().load(fh, 'ods')
-        self.assertEqual(dataset.pop(), (1, '', True, ''))
+        self.assertEqual(dataset[0], (1, '', True) + ('',) * 6)
+
+    def test_ods_import_columns_repeated(self):
+        ods = Path(__file__).parent / "files" / "columns_repeated.ods"
+
+        with ods.open("rb") as fh:
+            dataset = tablib.Dataset().load(fh, "ods")
+
+        self.assertEqual(dataset[0], ("a1", "", "", "d1"))
+        self.assertEqual(dataset[1], ("repeat",) * 4)
 
     def test_ods_unknown_value_type(self):
         # The ods file was trafficked to contain: