diff --git a/agent/negotiation/cards/domain/tactics.py b/agent/negotiation/cards/domain/tactics.py
index 7f652ae..9fdfe7c 100644
--- a/agent/negotiation/cards/domain/tactics.py
+++ b/agent/negotiation/cards/domain/tactics.py
@@ -114,7 +114,7 @@ def compute_counter(spec: TacticSpec, context: Dict[str, Any]) -> Optional[int]:
     if counter <= 0:
         return None
     counter = min(counter, target)          # 목표가 초과 제시 금지 (가드레일)
-    counter_i = int(round(counter))
+    counter_i = int(counter / 10 + 0.5) * 10  # 10원 단위 반올림 — 앵커가·목표가 산정과 표기 통일(IMK 요청)
     if counter_i >= price:
         return None                          # 제시가가 이미 카운터 이하 → 카운터 무의미
     return counter_i
diff --git a/agent/tests/test_card_tactics.py b/agent/tests/test_card_tactics.py
index 681af4c..fd27bf7 100644
--- a/agent/tests/test_card_tactics.py
+++ b/agent/tests/test_card_tactics.py
@@ -47,8 +47,8 @@ def test_counter_formulas_and_clamp():
     assert compute_counter(tactic_for("WC-05"), ctx) == 10000
     # 갑 직전 포지션이 있으면 그 기준: (9800+11000)/2 = 10400 → 역시 클램프 10000
     assert compute_counter(tactic_for("WC-05"), dict(ctx, prev_customer_price=9800)) == 10000
-    # 클램프 미발동 구간: (9900+10050)/2 = 9975 ≤ target
-    assert compute_counter(tactic_for("WC-05"), dict(ctx, input_price=10050)) == 9975
+    # 클램프 미발동 구간: (9900+10050)/2 = 9975 ≤ target → 10원 단위 반올림 9980
+    assert compute_counter(tactic_for("WC-05"), dict(ctx, input_price=10050)) == 9980
 
 
 def test_counter_meaningless_degrades_to_hold():