algo-hr-compensation
用市場資料做薪酬基準分析，建立薪資級距並檢查公平性
中文翻譯
---
name: "algo-hr-compensation"
description: "進行薪酬基準分析，把薪資和市場資料對照定位。當使用者需要評估薪資競爭力、建立薪資級距或分析薪酬公平性時使用，即使他們說的是『are we paying market rate』『salary benchmarking』或『compensation analysis』。"
metadata:
category: "WP-42 HR 演算法"
tags: ["hr", "compensation", "benchmarking", "salary-analysis"]
---
# 薪酬基準分析
## 概述
薪酬基準分析把內部薪資水準和外部市場資料比較，評估競爭力。使用 compa-ratio（實際薪資 / 市場中位數）和百分位定位。用來支援薪資級距設計、調薪和公平性分析。
## 何時使用
**觸發條件：**
- 評估目前的薪資在市場上有沒有競爭力
- 設計或更新薪資級距與薪酬結構
- 找出不同族群或職務之間的薪酬公平性缺口
**不該使用的情況：**
- 個別的績效調薪決策（用績效管理）
- 沒有市場資料時（至少要有薪資調查的基準）
## 演算法
```
IRON LAW: Benchmarking Is Only Valid With COMPARABLE Jobs
Matching by job TITLE alone is unreliable — "Senior Engineer" means
vastly different things at different companies. Match by: job content
(duties, scope), level (IC vs manager, experience band), industry,
geography, and company size. Poor job matching produces misleading
market rates.
```
### 階段 1：輸入驗證
收集：內部薪酬資料（底薪、獎金、股權）、市場調查資料（各職務的 P25、P50、P75）、內部職務與調查基準之間的職務對應。
**關卡：** 職務已正確對應，調查資料夠新（< 18 個月）。
### 階段 2：核心演算法
1. 依職務內容、層級和範圍，把內部職務對應到市場基準
2. 把調查資料推算到當前日期：套用預估的市場調幅
3. 計算每位員工的 compa-ratio：實際底薪 / 市場 P50
4. 計算百分位定位：實際薪資落在市場分布的哪裡
5. 分析：依部門、層級、年資、族群檢查公平性缺口
### 階段 3：驗證
檢查：compa-ratio 集中在 0.85-1.15（正常範圍）。標記離群值（< 0.80 薪資偏低，> 1.20 薪資偏高）。檢驗族群之間的公平性。
**關卡：** 分布合理，公平性分析已完成。
### 階段 4：輸出
回傳基準分析結果，附上級距建議。
## 輸出格式
```json
{
"summary": {"avg_compa_ratio": 0.97, "below_band_pct": 12, "above_band_pct": 8},
"by_role": [{"role": "Software Engineer", "market_p50": 1800000, "avg_actual": 1750000, "compa_ratio": 0.97}],
"equity_flags": [{"dimension": "gender", "gap_pct": 3.2, "statistically_significant": true}],
"metadata": {"employees": 500, "survey_source": "Mercer", "survey_date": "2025-H2"}
}
```
## 範例
### 輸入輸出範例
**輸入：** 50 位工程師，市場 P50=NT$1.8M，實際範圍 NT$1.5M-2.1M
**預期：** 平均 compa-ratio 約 0.97，部分低於級距的員工標記為需要調整。
### 邊界情況
| 輸入 | 預期 | 原因 |
|-------|----------|-----|
| 熱門市場（科技榮景） | 市場資料很快過時 | 套用較高的推算係數 |
| 混合遠端工作 | 需要依地點調整的級距 | 舊金山和台北的市場薪資差 2-3 倍 |
| 小公司，對不到調查資料 | 用較寬的產業代理資料 | 不完美，但總比沒有好 |
## 注意事項
- **總薪酬**：只做底薪的基準分析，會漏掉股權、獎金和福利。要比較總薪酬才能定位準確。
- **調查資料的落後**：已發布的調查反映的是 6-18 個月前收集的資料。市場變動快的時候，要把資料往前推算。
- **內部公平 vs 外部競爭力**：和市場對齊可能造成內部不公平（新人薪水比資深員工高）。兩邊都要顧。
- **地區差異**：遠端工作讓依地點定薪變複雜。要訂清楚的政策：依總部所在地、員工所在地，或混合。
- **薪酬公平的法律風險**：無法解釋的族群薪資缺口會帶來法律責任。要做迴歸式的公平性分析，並控制合理因素（經驗、績效、層級）。
## 參考資料
- 薪資級距設計方法論，見 `references/band-design.md`
- 薪酬公平的迴歸分析，見 `references/pay-equity.md`英文原文
---
name: "algo-hr-compensation"
description: "Conduct compensation benchmarking analysis to position salaries against market data. Use this skill when the user needs to assess pay competitiveness, build salary bands, or analyze pay equity — even if they say 'are we paying market rate', 'salary benchmarking', or 'compensation analysis'."
metadata:
category: "WP-42 HR 演算法"
tags: ["hr", "compensation", "benchmarking", "salary-analysis"]
---
# Compensation Benchmarking
## Overview
Compensation benchmarking compares internal pay levels against external market data to assess competitiveness. Uses compa-ratio (actual pay / market midpoint) and percentile positioning. Informs salary band design, pay adjustments, and equity analysis.
## When to Use
**Trigger conditions:**
- Evaluating whether current salaries are competitive with the market
- Designing or updating salary bands and pay structures
- Identifying pay equity gaps across demographics or roles
**When NOT to use:**
- For individual performance-based pay decisions (use performance management)
- When no market data is available (need at least survey benchmarks)
## Algorithm
```
IRON LAW: Benchmarking Is Only Valid With COMPARABLE Jobs
Matching by job TITLE alone is unreliable — "Senior Engineer" means
vastly different things at different companies. Match by: job content
(duties, scope), level (IC vs manager, experience band), industry,
geography, and company size. Poor job matching produces misleading
market rates.
```
### Phase 1: Input Validation
Collect: internal compensation data (base, bonus, equity), market survey data (P25, P50, P75 by role), job matching between internal roles and survey benchmarks.
**Gate:** Jobs properly matched, survey data current (< 18 months).
### Phase 2: Core Algorithm
1. Match internal jobs to market benchmarks by content, level, and scope
2. Age survey data to current date: apply projected market movement rate
3. Compute compa-ratio per employee: actual base / market P50
4. Compute percentile positioning: where does actual pay fall in market distribution
5. Analyze: by department, level, tenure, demographics for equity gaps
### Phase 3: Verification
Check: compa-ratios cluster around 0.85-1.15 (normal range). Flag outliers (< 0.80 underpaid, > 1.20 overpaid). Test demographic equity.
**Gate:** Distribution reasonable, equity analysis completed.
### Phase 4: Output
Return benchmarking results with band recommendations.
## Output Format
```json
{
"summary": {"avg_compa_ratio": 0.97, "below_band_pct": 12, "above_band_pct": 8},
"by_role": [{"role": "Software Engineer", "market_p50": 1800000, "avg_actual": 1750000, "compa_ratio": 0.97}],
"equity_flags": [{"dimension": "gender", "gap_pct": 3.2, "statistically_significant": true}],
"metadata": {"employees": 500, "survey_source": "Mercer", "survey_date": "2025-H2"}
}
```
## Examples
### Sample I/O
**Input:** 50 engineers, market P50=NT$1.8M, actual range NT$1.5M-2.1M
**Expected:** Avg compa-ratio ~0.97, some below-band employees flagged for adjustment.
### Edge Cases
| Input | Expected | Why |
|-------|----------|-----|
| Hot market (tech boom) | Market data rapidly outdated | Apply higher aging factor |
| Remote work mixed | Location-adjusted bands needed | SF vs Taipei market rates differ 2-3x |
| Small company, no survey match | Use broader industry proxies | Imperfect but better than nothing |
## Gotchas
- **Total compensation**: Base salary benchmarking alone misses equity, bonuses, and benefits. Compare total comp for accurate positioning.
- **Survey data lag**: Published surveys reflect data collected 6-18 months ago. In fast-moving markets, age the data forward.
- **Internal equity vs external competitiveness**: Aligning with market may create internal inequities (new hire paid more than tenured employee). Balance both.
- **Geographic differentials**: Remote work complicates location-based pay. Define a clear policy: pay by HQ location, employee location, or hybrid.
- **Pay equity legal risk**: Unexplained demographic pay gaps expose legal liability. Conduct regression-based equity analysis controlling for legitimate factors (experience, performance, level).
## References
- For salary band design methodology, see `references/band-design.md`
- For pay equity regression analysis, see `references/pay-equity.md`