diff --git a/CHANGELOG.md b/CHANGELOG.md
index 09a0112..3922f12 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -2,6 +2,11 @@
 
 All notable changes to this project are documented in this file.
 
+## Unreleased
+
+### Changed
+- **The release-mode minimum `number_of_cases` threshold for `applied_to_treat` edges is now 10 instead of 25.** `drop_low_number_of_cases` keeps edges whose case count is >= 10 (nulls kept, as before); release builds now ship `applied_to_treat` edges backed by 10-24 cases that earlier releases dropped. Gating (`--release` *and* `applied_to_treat`), null handling, non-numeric tolerance, and `significance`-phase placement are unchanged.
+
 ## 16.6.2 - 2026-09-04
 
 ### Fixed
diff --git a/src/tablassert/lib.py b/src/tablassert/lib.py
index 31f82d6..c0892c7 100644
--- a/src/tablassert/lib.py
+++ b/src/tablassert/lib.py
@@ -1058,7 +1058,7 @@ def drop_zero_effect_size(lf: pl.LazyFrame, col: str = "effect_size") -> pl.Lazy
     return lf.filter(pl.col(col).cast(pl.Float64, strict=False).ne_missing(0.0))
 
 
-def drop_low_number_of_cases(lf: pl.LazyFrame, col: str = "number_of_cases", threshold: float = 25.0) -> pl.LazyFrame:
+def drop_low_number_of_cases(lf: pl.LazyFrame, col: str = "number_of_cases", threshold: float = 10.0) -> pl.LazyFrame:
     """Drop release-mode ``applied_to_treat`` edges whose number of cases is below the threshold.
 
     Args:
diff --git a/tests/test_lib.py b/tests/test_lib.py
index 95b93c9..081037b 100644
--- a/tests/test_lib.py
+++ b/tests/test_lib.py
@@ -1309,11 +1309,11 @@ def test_drop_zero_effect_size_noop_without_column() -> None:
 
 
 def test_drop_low_number_of_cases_removes_below_threshold_keeps_at_threshold_and_nulls() -> None:
-    """drop_low_number_of_cases drops case counts under 25 while keeping 25+ and nulls."""
-    lf: pl.LazyFrame = pl.DataFrame({"subject": ["a", "b", "c", "d", "e"], "number_of_cases": [24, 25, 26, None, 0]}).lazy()
+    """drop_low_number_of_cases drops case counts under 10 while keeping 10+ and nulls."""
+    lf: pl.LazyFrame = pl.DataFrame({"subject": ["a", "b", "c", "d", "e"], "number_of_cases": [9, 10, 11, None, 0]}).lazy()
     result: pl.DataFrame = drop_low_number_of_cases(lf).collect()
     assert list(result["subject"]) == ["b", "c", "d"]
-    assert list(result["number_of_cases"]) == [25, 26, None]
+    assert list(result["number_of_cases"]) == [10, 11, None]
 
 
 def test_drop_low_number_of_cases_noop_without_column() -> None:
@@ -1328,7 +1328,7 @@ def test_drop_low_number_of_cases_keeps_non_numeric_cells() -> None:
     """drop_low_number_of_cases keeps non-numeric cells (a header row read as data) instead of crashing on a strict cast."""
     # The csv op reads sources with has_header=False, so a TSV's header row flows
     # through as a data row; the strict cast raised InvalidOperationError on it.
-    lf: pl.LazyFrame = pl.DataFrame({"subject": ["hdr", "a", "b", "c"], "number_of_cases": ["number_of_cases", "30", "10", None]}).lazy()
+    lf: pl.LazyFrame = pl.DataFrame({"subject": ["hdr", "a", "b", "c"], "number_of_cases": ["number_of_cases", "30", "9", None]}).lazy()
     result: pl.DataFrame = drop_low_number_of_cases(lf).collect()
     assert list(result["subject"]) == ["hdr", "a", "c"]