diff --git a/app/services/comparison.py b/app/services/comparison.py
index 0187a82..b5a1620 100644
--- a/app/services/comparison.py
+++ b/app/services/comparison.py
@@ -8,7 +8,8 @@
 MIN_SHARED_FOR_SCORE = 5
 RESULT_LIMIT = 5
 METHOD = (
-    'We compare rank positions directly. Smaller differences mean closer agreement.'
+    'Each gap is the absolute difference between your two rank positions. '
+    'Alignment is 100% minus the average gap as a share of the longer shelf’s rank span.'
 )
 
 
@@ -17,6 +18,13 @@ def _rank_gap(your_rank: int, their_rank: int) -> int:
     return abs(your_rank - their_rank)
 
 
+def _alignment_score(gaps: list[int], longer_shelf_count: int) -> int:
+    """Scale the average raw gap against the longer shelf's possible span."""
+    rank_span = max(1, longer_shelf_count - 1)
+    mean_gap = sum(gaps) / len(gaps)
+    return round(max(0.0, 1 - (mean_gap / rank_span)) * 100)
+
+
 def _item(catalog, **extra) -> dict:
     return {
         'id': str(catalog.id),
@@ -72,6 +80,11 @@ def compare_shelf(  # pylint: disable=too-many-locals
         getattr(row, shelf.join_col): row
         for row in db.query(tracker).filter(tracker.user_id == viewer.pk).all()
     }
+    viewer_ranked_count = sum(
+        1
+        for row in viewer_trackers.values()
+        if row.on_rankings and row.rank is not None
+    )
     base['recommendations'] = [
         _item(
             item,
@@ -106,8 +119,10 @@ def compare_shelf(  # pylint: disable=too-many-locals
         shared, key=lambda item: (item['gap'], item['title'].lower())
     )[:RESULT_LIMIT]
     if len(shared) >= MIN_SHARED_FOR_SCORE:
-        mean_gap = sum(item['gap'] for item in shared) / len(shared)
-        base['alignment_score'] = round(max(0.0, 100 - mean_gap))
+        base['alignment_score'] = _alignment_score(
+            [item['gap'] for item in shared],
+            max(viewer_ranked_count, len(target_rows)),
+        )
         base['alignment_status'] = 'ready'
 
     if watchlist_visible:
diff --git a/tests/integration/router_comparison_test.py b/tests/integration/router_comparison_test.py
index aeda991..08171b7 100644
--- a/tests/integration/router_comparison_test.py
+++ b/tests/integration/router_comparison_test.py
@@ -104,7 +104,7 @@ def test_comparison_scores_visible_rankings_and_marks_watchlist(
     )
     assert movies['shared_ranked_count'] == 5
     assert movies['alignment_status'] == 'ready'
-    assert isinstance(movies['alignment_score'], int)
+    assert movies['alignment_score'] == 60
     assert [item['title'] for item in movies['common_watchlist']] == ['Movie 8']
     assert [item['title'] for item in movies['recommendations']] == [
         'Movie 6',
diff --git a/tests/unit/services/comparison_test.py b/tests/unit/services/comparison_test.py
index 2ab5f72..e115715 100644
--- a/tests/unit/services/comparison_test.py
+++ b/tests/unit/services/comparison_test.py
@@ -1,7 +1,7 @@
 # pylint: disable=missing-function-docstring
 """The direct rank-distance calculation behind comparisons."""
 
-from app.services.comparison import _rank_gap
+from app.services.comparison import _alignment_score, _rank_gap
 
 
 def test_equal_ranks_have_no_gap():
@@ -11,3 +11,8 @@ def test_equal_ranks_have_no_gap():
 def test_gap_is_the_absolute_rank_difference():
     assert _rank_gap(88, 10) == 78
     assert _rank_gap(10, 88) == 78
+
+
+def test_alignment_scales_average_gap_against_longer_shelf_span():
+    assert _alignment_score([4, 4, 2, 2, 0], longer_shelf_count=7) == 60
+    assert _alignment_score([0, 0, 0, 0, 0], longer_shelf_count=100) == 100