diff --git a/pyproject.toml b/pyproject.toml
index f1bb921..d785419 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -42,7 +42,6 @@ dependencies = [
         "pyramid_fanstatic",
         "pyramid_mako",
         "pyramid_tm",
-        "SQLAlchemy-Utils",
         "waitress",
         "WebHelpers2",
         "WuttJamaican[db]>=0.19.2",
diff --git a/src/wuttaweb/grids/base.py b/src/wuttaweb/grids/base.py
index 3a3d4f5..c2f3156 100644
--- a/src/wuttaweb/grids/base.py
+++ b/src/wuttaweb/grids/base.py
@@ -32,7 +32,6 @@ from collections import namedtuple, OrderedDict
 
 import sqlalchemy as sa
 from sqlalchemy import orm
-from sqlalchemy_utils import get_columns
 
 import paginate
 from paginate_sqlalchemy import SqlalchemyOrmPage
@@ -1147,16 +1146,29 @@ class Grid:
         filters = filters or {}
 
         if self.model_class:
-            # nb. i first tried self.get_model_columns() but my notes
-            # say that was too aggressive in many cases.  then i tried
-            # using the *subset* of self.columns, just the ones which
-            # corresponded to a property on the model class.  and now
-            # i am using sa-utils to give the "true" column list..
-            for col in get_columns(self.model_class):
-                if col.key in filters:
-                    continue
-                prop = getattr(self.model_class, col.key)
-                filters[prop.key] = self.make_filter(prop)
+
+            # nb. i have found this confusing for some reason.  some
+            # things i've tried so far include:
+            #
+            # i first tried self.get_model_columns() but my notes say
+            # that was too aggressive in many cases.
+            #
+            # then i tried using the *subset* of self.columns, just
+            # the ones which correspond to a property on the model
+            # class.  but sometimes that skips filters we need.
+            #
+            # then i tried get_columns() from sa-utils to give the
+            # "true" column list, but that fails when the underlying
+            # column has different name than the prop/attr key.
+            #
+            # so now, we are looking directly at the sa mapper, for
+            # all column attrs and then using the prop key.
+
+            inspector = sa.inspect(self.model_class)
+            for prop in inspector.column_attrs:
+                if prop.key not in filters:
+                    attr = getattr(self.model_class, prop.key)
+                    filters[prop.key] = self.make_filter(attr)
 
         return filters