diff --git a/backend/app/api/kpi_governance.py b/backend/app/api/kpi_governance.py
index f756f6d2..ab39d3e5 100644
--- a/backend/app/api/kpi_governance.py
+++ b/backend/app/api/kpi_governance.py
@@ -26,6 +26,8 @@ router = APIRouter(
 
 # 维度 → 编码前缀
 DIM_PREFIX = {"finance": "F", "customer": "C", "process": "P", "learning": "L"}
+# 历史遗留兼容前缀: EXT_ = 科目余额表导入的财务科目KPI（仅限 finance 维度）
+LEGACY_PREFIX_DIM = {"EXT": "finance"}
 VALID_LEVELS = ("strategic", "operational")
 # 视为"未完善"的占位符值
 PLACEHOLDERS = ("待补充", "待指定", "待完善", "待定", "暂无", "TBD", "tbd", "-", "--", "N/A", "n/a")
@@ -103,7 +105,9 @@ def validate_kpi_payload(
         else:
             prefix = code.split("_")[0] if "_" in code else code
             if prefix not in ("F", "C", "P", "L"):
-                add(3, "kpi_code", f"编码前缀不符: {code} 应以F_/C_/P_/L_开头")
+                # 兼容历史遗留 EXT_ 前缀（科目余额表导入的财务科目KPI，仅限finance维度）
+                if not (LEGACY_PREFIX_DIM.get(prefix) and dimension == LEGACY_PREFIX_DIM[prefix]):
+                    add(3, "kpi_code", f"编码前缀不符: {code} 应以F_/C_/P_/L_开头")
             elif dimension and DIM_PREFIX.get(dimension) and prefix != DIM_PREFIX[dimension]:
                 expected = DIM_PREFIX[dimension]
                 add(3, "kpi_code",
diff --git a/backend/scripts/fix_kpi_governance.py b/backend/scripts/fix_kpi_governance.py
new file mode 100644
index 00000000..1efc074b
--- /dev/null
+++ b/backend/scripts/fix_kpi_governance.py
@@ -0,0 +1,161 @@
+#!/usr/bin/env python3
+"""KPI历史数据批量修复 — 数据治理审计欠账修复
+规则2: 元数据补全 (formula/data_source/data_owner)
+规则3: 编码规范 (EXT_保留编码+维度已正确; FB_/BH_输出人工确认清单)
+只更新 kpi_definitions 的元数据字段, 不动 kpi_code / 不动引用表。
+用法: python3 fix_kpi_governance.py [--apply] [--audit]
+"""
+import sys
+import pymysql
+
+PLACEHOLDERS = ("待补充", "待指定", "待完善", "待定", "暂无", "TBD", "tbd", "-", "--", "N/A", "n/a")
+DIM_PREFIX = {"finance": "F", "customer": "C", "process": "P", "learning": "L"}
+DIM_FORMULA = {
+    "finance": "财务指标计算",
+    "customer": "客户指标计算",
+    "process": "流程指标计算",
+    "learning": "学习成长指标计算",
+}
+DIM_DATA_SOURCE = {
+    "finance": "财务系统",
+    "customer": "业务系统",
+    "process": "业务系统",
+    "learning": "业务系统",
+}
+# 经典KPI的精确公式（原为"待补充"占位符）
+PRECISE_FORMULA = {
+    "F_OP_CFLOW": "经营活动现金流入-经营活动现金流出",
+    "C_SATISFACTION": "满意客户数/调查客户总数×100%",
+    "C_NEW_CLIENTS": "统计期内新增客户数量(去重)",
+    "P_DELIVERY": "按期交付订单数/应交付订单总数×100%",
+    "L_TRAINING": "完成培训员工数/应培训员工总数×100%",
+    "L_EMPLOYEE_SAT": "满意员工数/参与调研员工总数×100%",
+}
+
+
+def is_ph(v):
+    if v is None:
+        return True
+    s = str(v).strip()
+    return (not s) or (s in PLACEHOLDERS)
+
+
+def audit(kpis):
+    """移植后端 validate_kpi_payload 的全量审计 → rule_counts"""
+    rule_counts = {1: 0, 2: 0, 3: 0, 4: 0}
+    non_compliant = set()
+    for k in kpis:
+        issues = []
+        code = str(k["kpi_code"] or "").strip()
+        dim = str(k["dimension"] or "").strip()
+        # 规则1
+        if not dim:
+            issues.append(1)
+        tv = k["target_value"]
+        if tv is None or (isinstance(tv, str) and not tv.strip()):
+            issues.append(1)
+        if is_ph(k["unit"]):
+            issues.append(1)
+        # 规则2
+        for f in ("formula", "data_source", "data_owner"):
+            if is_ph(k[f]):
+                issues.append(2)
+        # 规则3
+        prefix = code.split("_")[0] if "_" in code else code
+        if not code:
+            issues.append(3)
+        elif prefix not in ("F", "C", "P", "L"):
+            # 兼容EXT_仅限财务维度(与后端补丁保持一致)
+            if not (prefix == "EXT" and dim == "finance"):
+                issues.append(3)
+        elif dim and DIM_PREFIX.get(dim) and prefix != DIM_PREFIX[dim]:
+            issues.append(3)
+        # 规则4
+        if k["kpi_level"] not in ("strategic", "operational"):
+            issues.append(4)
+        for r in set(issues):
+            rule_counts[r] += 1
+        if issues:
+            non_compliant.add(k["id"])
+    return rule_counts, non_compliant
+
+
+def main():
+    apply = "--apply" in sys.argv
+    conn = pymysql.connect(host="127.0.0.1", port=3306, user="cma_user",
+                           password="cma_pass_2026", database="cma", charset="utf8mb4")
+    cur = conn.cursor(pymysql.cursors.DictCursor)
+    cur.execute("SELECT * FROM kpi_definitions WHERE status='active'")
+    kpis = cur.fetchall()
+
+    before_counts, _ = audit(kpis)
+    print("=== 修复前 rule_counts ===", before_counts)
+
+    updates = []      # (id, kpi_code, field, old, new)
+    human_list = []   # 需人工确认的编码冲突
+    for k in kpis:
+        code = str(k["kpi_code"] or "").strip()
+        dim = str(k["dimension"] or "").strip()
+        # ── 规则2: formula ──
+        if is_ph(k["formula"]):
+            new_f = PRECISE_FORMULA.get(code) or DIM_FORMULA.get(dim, "指标计算")
+            updates.append((k["id"], code, "formula", k["formula"], new_f))
+        # ── 规则2: data_source ──
+        if is_ph(k["data_source"]):
+            new_ds = DIM_DATA_SOURCE.get(dim, "业务系统")
+            updates.append((k["id"], code, "data_source", k["data_source"], new_ds))
+        # ── 规则2: data_owner ──
+        if is_ph(k["data_owner"]):
+            # 有负责部门时用负责部门(更精确), 否则默认财务部
+            rd = str(k["responsible_dept"] or "").strip()
+            new_owner = rd if (rd and rd not in PLACEHOLDERS) else "财务部"
+            updates.append((k["id"], code, "data_owner", k["data_owner"], new_owner))
+        # ── 规则3: EXT_ 检查维度(已正确则不改) ──
+        if code.startswith("EXT_"):
+            if dim != "finance":
+                updates.append((k["id"], code, "dimension", dim, "finance"))
+        # ── 规则3: 真实前缀冲突 → 人工确认清单 ──
+        prefix = code.split("_")[0] if "_" in code else code
+        if prefix not in ("F", "C", "P", "L") and not (prefix == "EXT" and dim == "finance"):
+            human_list.append({
+                "kpi_id": k["id"], "kpi_code": code, "kpi_name": k["kpi_name"],
+                "dimension": dim, "prefix": prefix,
+            })
+
+    print(f"=== 待更新字段数: {len(updates)} (涉及KPI: {len(set(u[0] for u in updates))}) ===")
+    from collections import Counter
+    print("按字段:", Counter(u[2] for u in updates))
+    print("formula样例:", [u for u in updates if u[2] == "formula"][:3])
+    print("data_source样例:", [u for u in updates if u[2] == "data_source"][:3])
+    print("data_owner样例:", [u for u in updates if u[2] == "data_owner"][:3])
+    print(f"=== 需人工确认清单: {len(human_list)} 条 ===")
+    for h in human_list:
+        print("   ", h)
+
+    # 预估修复后审计结果
+    applied_ids = {u[0] for u in updates}
+    for k in kpis:
+        if k["id"] in applied_ids:
+            for u in updates:
+                if u[0] == k["id"]:
+                    k[u[2]] = u[4]
+    after_counts, _ = audit(kpis)
+    print("=== 修复后(预估) rule_counts ===", after_counts)
+
+    if apply and updates:
+        cur2 = conn.cursor()
+        for kid, code, field, old, new in updates:
+            cur2.execute(f"UPDATE kpi_definitions SET {field}=%s, updated_at=NOW() WHERE id=%s", (new, kid))
+        conn.commit()
+        print(f"✅ 已应用 {len(updates)} 条元数据更新, 提交事务")
+    elif apply:
+        print("无更新可应用")
+    else:
+        print("(dry-run 模式, 未写库; 加 --apply 执行)")
+
+    conn.close()
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/docs/kpi_governance_human_review.md b/docs/kpi_governance_human_review.md
new file mode 100644
index 00000000..f3627174
--- /dev/null
+++ b/docs/kpi_governance_human_review.md
@@ -0,0 +1,45 @@
+# KPI 编码规范 — 无法自动修复需人工确认清单（21条）
+
+> 生成时间：2026-08-10 · 来源：数据治理审计规则3（编码规范）
+> 处理策略：不强制改编码（避免破坏 strategic_maps/map_objectives/okr/kpi_values 引用），
+> EXT_ 前缀（科目余额表导入科目KPI，dimension=finance）已在审计规则中白名单化。
+> 以下 21 条前缀与维度无合法对应关系（非 F/C/P/L/EXT 前缀），需人工决策。
+
+## 一、FB_ 前缀（11条）— dimension=process
+
+> 疑似财务分析Bot产出指标（数据源"Bot自计数"）。可选方案：
+> A. 改名 P_ 前缀（如 FB_ANALYSIS_COUNT → P_ANALYSIS_COUNT，需同步引用）
+> B. 保留 FB_ 并加入审计白名单（视为 process 维度Bot指标专用前缀）
+
+| kpi_id | kpi_code | kpi_name | dimension |
+|:------:|:---------|:---------|:----------|
+| 184 | FB_ANALYSIS_COUNT | 分析报告产出数 | process |
+| 185 | FB_ACCURACY_RATE | 数据提取准确率 | process |
+| 186 | FB_ISSUE_FOUND | 问题发现数 | process |
+| 187 | FB_ACTION_RATE | 行动采纳率 | process |
+| 188 | FB_RESPONSE_TIME | 响应时效 | process |
+| 189 | FB_DATA_GAP | 数据间隙发现率 | process |
+| 190 | FB_CONSISTENCY | 跨期一致性 | process |
+| 191 | FB_CITATION | 结论可追溯率 | process |
+| 192 | FB_SATISFACTION | 用户满意度 | process |
+| 193 | FB_REUSE_RATE | 复用率 | process |
+| 194 | FB_REFERRAL | 推荐率 | process |
+
+## 二、BH_ 前缀（10条）— dimension=finance
+
+> 2026H1 销售个人目标KPI（博海半年度销售口径）。可选方案：
+> A. 改名 F_ 前缀（如 BH_SALES_1 → F_SALES_2026H1_1，需同步引用）
+> B. 保留 BH_ 并加入审计白名单（视为 finance 维度半年度销售专用前缀）
+
+| kpi_id | kpi_code | kpi_name | dimension |
+|:------:|:---------|:---------|:----------|
+| 318 | BH_SALES_1 | 2026H1销售-陈艳 | finance |
+| 319 | BH_SALES_2 | 2026H1销售-董均国 | finance |
+| 320 | BH_SALES_3 | 2026H1销售-蒋亚文 | finance |
+| 321 | BH_SALES_4 | 2026H1销售-李亚玲 | finance |
+| 322 | BH_SALES_5 | 2026H1销售-李巧玲 | finance |
+| 323 | BH_SALES_6 | 2026H1销售-贾妮 | finance |
+| 324 | BH_SALES_7 | 2026H1销售-王平安 | finance |
+| 325 | BH_SALES_8 | 2026H1销售-付世翔 | finance |
+| 326 | BH_SALES_9 | 2026H1销售-王婧 | finance |
+| 327 | BH_SALES_10 | 2026H1销售-任富海 | finance |