diff --git a/connect-file-pulse-filters/src/main/java/io/streamthoughts/kafka/connect/filepulse/filter/AbstractDelimitedRowFilter.java b/connect-file-pulse-filters/src/main/java/io/streamthoughts/kafka/connect/filepulse/filter/AbstractDelimitedRowFilter.java
index 86725b805..963e07ee4 100644
--- a/connect-file-pulse-filters/src/main/java/io/streamthoughts/kafka/connect/filepulse/filter/AbstractDelimitedRowFilter.java
+++ b/connect-file-pulse-filters/src/main/java/io/streamthoughts/kafka/connect/filepulse/filter/AbstractDelimitedRowFilter.java
@@ -42,6 +42,8 @@ public abstract class AbstractDelimitedRowFilter<T extends AbstractRecordFilter<
 
     private StructSchema schema;
 
+    private String cachedHeaders;
+
     private final Map<Integer, TypedField> columnsTypesByIndex = new HashMap<>();
 
     /**
@@ -103,6 +105,11 @@ public RecordsIterable<TypedStruct> apply(final FilterContext context,
         if (schema == null || isSchemaDynamic()) {
             inferSchemaFromRecord(record, columnValues.length);
         }
+
+        if (schema != null && configs.extractColumnName() != null && shouldInferSchema(record)) {
+            inferSchemaFromRecord(record, columnValues.length);
+        }
+
         final TypedStruct struct = buildStructForFields(columnValues);
         return RecordsIterable.of(struct);
     }
@@ -115,12 +122,23 @@ public boolean isSchemaDynamic() {
                configs.isAutoGenerateColumnNames();
     }
 
+    private boolean shouldInferSchema(TypedStruct record) {
+        if (cachedHeaders == null) {
+            return false;
+        }
+        final String fieldName = configs.extractColumnName();
+        String field = record.first(fieldName).getString();
+        return cachedHeaders.length() == field.length() && !cachedHeaders.equals(field);
+    }
+
     private void inferSchemaFromRecord(final TypedStruct record, int numColumns) {
         schema = Schema.struct();
 
         if (configs.extractColumnName() != null) {
             final String fieldName = configs.extractColumnName();
             String field = record.first(fieldName).getString();
+            cachedHeaders = field;
+
             if (field == null) {
                 throw new FilterException(
                     "Cannot find field for name '" + fieldName + "' to determine columns names"
diff --git a/connect-file-pulse-filters/src/test/java/io/streamthoughts/kafka/connect/filepulse/filter/CSVFilterTest.java b/connect-file-pulse-filters/src/test/java/io/streamthoughts/kafka/connect/filepulse/filter/CSVFilterTest.java
index 36ceb836c..0792709bf 100644
--- a/connect-file-pulse-filters/src/test/java/io/streamthoughts/kafka/connect/filepulse/filter/CSVFilterTest.java
+++ b/connect-file-pulse-filters/src/test/java/io/streamthoughts/kafka/connect/filepulse/filter/CSVFilterTest.java
@@ -39,6 +39,100 @@ public void setUp() {
         configs.put(CSVFilter.PARSER_SEPARATOR_CONFIG, ";");
     }
 
+    @Test
+    public void should_extract_column_names_from_diff_order_headers() {
+        configs.put(READER_EXTRACT_COLUMN_NAME_CONFIG, "headers");
+        filter.configure(configs, alias -> null);
+
+        RecordsIterable<TypedStruct> output = filter.apply(null, DEFAULT_STRUCT, false);
+        Assert.assertNotNull(output);
+        Assert.assertEquals(1, output.size());
+
+        final TypedStruct record = output.iterator().next();
+        Assert.assertEquals("value1", record.getString("col1"));
+        Assert.assertEquals("2", record.getString("col2"));
+        Assert.assertEquals("true", record.getString("col3"));
+
+        final TypedStruct input1 = TypedStruct.create()
+                .put("message", "false;3;value2")
+                .put("headers", Arrays.asList("col3;col2;col1"));
+        RecordsIterable<TypedStruct> output1 = filter.apply(null, input1, false);
+        Assert.assertNotNull(output1);
+        Assert.assertEquals(1, output1.size());
+
+        final TypedStruct record1 = output1.iterator().next();
+        Assert.assertEquals("value2", record1.getString("col1"));
+        Assert.assertEquals("3", record1.getString("col2"));
+        Assert.assertEquals("false", record1.getString("col3"));
+
+        final TypedStruct input2 = TypedStruct.create()
+                .put("message", "4;false;value3")
+                .put("headers", Arrays.asList("col2;col3;col1"));
+
+        RecordsIterable<TypedStruct> output2 = filter.apply(null, input2, false);
+        Assert.assertNotNull(output2);
+        Assert.assertEquals(1, output2.size());
+
+        final TypedStruct record2 = output2.iterator().next();
+        Assert.assertEquals("value3", record2.getString("col1"));
+        Assert.assertEquals("4", record2.getString("col2"));
+        Assert.assertEquals("false", record2.getString("col3"));
+    }
+
+    @Test
+    public void should_extract_column_names_from_diff_order_headers_and_null_value() {
+        configs.put(READER_EXTRACT_COLUMN_NAME_CONFIG, "headers");
+        filter.configure(configs, alias -> null);
+
+        RecordsIterable<TypedStruct> output = filter.apply(null, DEFAULT_STRUCT, false);
+        Assert.assertNotNull(output);
+        Assert.assertEquals(1, output.size());
+
+        final TypedStruct record = output.iterator().next();
+        Assert.assertEquals("value1", record.getString("col1"));
+        Assert.assertEquals("2", record.getString("col2"));
+        Assert.assertEquals("true", record.getString("col3"));
+
+        final TypedStruct input1 = TypedStruct.create()
+                .put("message", "false;;")
+                .put("headers", Arrays.asList("col3;col2;col1"));
+        RecordsIterable<TypedStruct> output1 = filter.apply(null, input1, false);
+        Assert.assertNotNull(output1);
+        Assert.assertEquals(1, output1.size());
+
+        final TypedStruct record1 = output1.iterator().next();
+        Assert.assertNull(record1.getString("col1"));
+        Assert.assertNull(record1.getString("col2"));
+        Assert.assertEquals("false", record1.getString("col3"));
+    }
+
+    @Test
+    public void should_extract_column_names_from_diff_order_headers_and_diff_size() {
+        configs.put(READER_EXTRACT_COLUMN_NAME_CONFIG, "headers");
+        filter.configure(configs, alias -> null);
+
+        RecordsIterable<TypedStruct> output = filter.apply(null, DEFAULT_STRUCT, false);
+        Assert.assertNotNull(output);
+        Assert.assertEquals(1, output.size());
+
+        final TypedStruct record = output.iterator().next();
+        Assert.assertEquals("value1", record.getString("col1"));
+        Assert.assertEquals("2", record.getString("col2"));
+        Assert.assertEquals("true", record.getString("col3"));
+
+        final TypedStruct input1 = TypedStruct.create()
+                .put("message", "false;4;")
+                .put("headers", Arrays.asList("col3;col2"));
+        RecordsIterable<TypedStruct> output1 = filter.apply(null, input1, false);
+        Assert.assertNotNull(output1);
+        Assert.assertEquals(1, output1.size());
+
+        final TypedStruct record1 = output1.iterator().next();
+        Assert.assertEquals("false", record1.getString("col1"));
+        Assert.assertEquals("4", record1.getString("col2"));
+        Assert.assertNull(record1.getString("col3"));
+    }
+
     @Test
     public void should_auto_generate_schema_given_no_schema_field() {
         filter.configure(configs, alias -> null);