diff --git a/.travis.yml b/.travis.yml
index 3f285f6..773dde1 100644
--- a/.travis.yml
+++ b/.travis.yml
@@ -1,7 +1,8 @@
 language: python
 python:
-  - "2.6"
-  - "2.7"
+  - 2.6
+  - 2.7
+  - 3.5
 env:
   - DJANGO=1.5
   - DJANGO=1.6
@@ -10,12 +11,22 @@ env:
   - DJANGO=1.9
 matrix:
   exclude:
-    - python: "2.6"
+    - python: 2.6
+      env: DJANGO=1.6
+    - python: 2.6
       env: DJANGO=1.7
-    - python: "2.6"
+    - python: 2.6
       env: DJANGO=1.8
-    - python: "2.6"
+    - python: 2.6
       env: DJANGO=1.9
+    - python: 3.5
+      env: DJANGO=1.5
+    - python: 3.5
+      env: DJANGO=1.6
+    - python: 3.5
+      env: DJANGO=1.7
+
+
 install:
   - pip install -q Django==$DJANGO
   - pip install -r dev_requirements.txt
diff --git a/djqscsv/djqscsv.py b/djqscsv/djqscsv.py
index 034f689..27b85aa 100644
--- a/djqscsv/djqscsv.py
+++ b/djqscsv/djqscsv.py
@@ -147,7 +147,8 @@ def generate_filename(queryset, append_datestamp=False):
     Takes a queryset and returns a default
     base filename based on the underlying model
     """
-    base_filename = slugify(unicode(queryset.model.__name__)) + '_export.csv'
+    base_filename = slugify(six.text_type(queryset.model.__name__)) \
+        + '_export.csv'
 
     if append_datestamp:
         base_filename = _append_datestamp(base_filename)
@@ -167,17 +168,17 @@ def _validate_and_clean_filename(filename):
         else:
             filename = filename[:-4]
 
-    filename = slugify(unicode(filename)) + '.csv'
+    filename = slugify(six.text_type(filename)) + '.csv'
     return filename
 
 
 def _safe_utf8_stringify(value):
-    if isinstance(value, str):
+    if isinstance(value, six.string_types):
         return value
-    elif isinstance(value, unicode):
+    elif isinstance(value, six.binary_type):
         return value.encode('utf-8')
     else:
-        return unicode(value).encode('utf-8')
+        return six.text_type(value).encode('utf-8')
 
 
 def _sanitize_unicode_record(field_serializer_map, record):
@@ -188,7 +189,7 @@ def _serialize_value(value):
         if isinstance(value, datetime.datetime):
             return value.isoformat()
         else:
-            return unicode(value)
+            return six.text_type(value)
 
     obj = {}
     for key, val in six.iteritems(record):
diff --git a/test_app/djqscsv_tests/tests/test_csv_creation.py b/test_app/djqscsv_tests/tests/test_csv_creation.py
index 843c98c..f7ca0fb 100644
--- a/test_app/djqscsv_tests/tests/test_csv_creation.py
+++ b/test_app/djqscsv_tests/tests/test_csv_creation.py
@@ -12,12 +12,7 @@
 
 from djqscsv_tests.util import create_people_and_get_queryset
 
-from django.utils import six
-
-if six.PY3:
-    from io import StringIO
-else:
-    from StringIO import StringIO
+from django.utils.six import StringIO
 
 
 class CSVTestCase(TestCase):
diff --git a/test_app/djqscsv_tests/tests/test_utilities.py b/test_app/djqscsv_tests/tests/test_utilities.py
index cebd8a2..6ac4a84 100644
--- a/test_app/djqscsv_tests/tests/test_utilities.py
+++ b/test_app/djqscsv_tests/tests/test_utilities.py
@@ -51,16 +51,16 @@ def test_sanitize(self):
                   'nickname': u'\ufeffThe White Lady of Gont'}
         sanitized = djqscsv._sanitize_unicode_record({}, record)
         self.assertEqual(sanitized,
-                         {'name': 'Tenar',
-                          'nickname': '\xef\xbb\xbfThe White Lady of Gont'})
+                         {b'name': b'Tenar',
+                          b'nickname': b'\xef\xbb\xbfThe White Lady of Gont'})
 
     def test_sanitize_date(self):
         record = {'name': 'Tenar',
                   'created': datetime.datetime(1, 1, 1)}
         sanitized = djqscsv._sanitize_unicode_record({}, record)
         self.assertEqual(sanitized,
-                         {'name': 'Tenar',
-                          'created': '0001-01-01T00:00:00'})
+                         {b'name': b'Tenar',
+                          b'created': b'0001-01-01T00:00:00'})
 
     def test_sanitize_date_with_non_string_formatter(self):
         """
@@ -71,7 +71,7 @@ def test_sanitize_date_with_non_string_formatter(self):
         record = {'name': 'Tenar'}
         serializer = {'name': lambda d: len(d)}
         sanitized = djqscsv._sanitize_unicode_record(serializer, record)
-        self.assertEqual(sanitized, {'name': '5'})
+        self.assertEqual(sanitized, {b'name': b'5'})
 
     def test_sanitize_date_with_formatter(self):
         record = {'name': 'Tenar',
@@ -79,8 +79,8 @@ def test_sanitize_date_with_formatter(self):
         serializer = {'created': lambda d: d.strftime('%Y-%m-%d')}
         sanitized = djqscsv._sanitize_unicode_record(serializer, record)
         self.assertEqual(sanitized,
-                         {'name': 'Tenar',
-                          'created': '1973-05-13'})
+                         {b'name': b'Tenar',
+                          b'created': b'1973-05-13'})
 
     def test_sanitize_date_with_bad_formatter(self):
         record = {'name': 'Tenar',