algo-social-engagement
計算並比較社群媒體的互動率與基準
中文翻譯
---
name: "algo-social-engagement"
description: "計算社群媒體互動率，並跨平台、跨計算變體做基準比較。當使用者需要計算互動指標、比較不同帳號或貼文的表現、或設定互動基準時使用，即使他們說的是「我的互動率是多少」、「互動率基準」或「社群媒體 KPI」也一樣。"
metadata:
category: "WP-38 社群演算法"
tags: ["social-media", "engagement-rate", "analytics", "benchmarking"]
---
# 互動率計算（Engagement Rate Calculation）
## 概述
互動率衡量的是受眾互動，相對於觸及人數或受眾規模的比例。公式：(reactions + comments + shares) / 分母 × 100%。分母的選擇（觸及、曝光、追蹤者）會大幅影響結果。每組貼文的計算複雜度是 O(n)。
## 何時使用
**觸發條件：**
- 要為社群媒體報告計算互動指標
- 要把帳號或貼文的表現，和產業平均做基準比較
- 要比較不同貼文或帳號之間的內容表現
**何時不要用：**
- 要整體評估影響力時（用影響力衡量）
- 要建模內容擴散的動態時（用病毒式傳播模型）
## 演算法
```
鐵律：互動率的分母「很重要」
用觸及、用曝光、用追蹤者算，會得出「不同」的數字：
- 依觸及的 ER = 互動數 / 觸及 × 100%（最準確，需要有分析權限）
- 依曝光的 ER = 互動數 / 曝光 × 100%（永遠比依觸及的低）
- 依追蹤者的 ER = 互動數 / 追蹤者 × 100%（公開資料，但被沒觸及到的追蹤者灌水）
回報或比較時，「一定要」註明用的是哪一種變體。
```
### 階段 1：輸入驗證
每則貼文要蒐集：按讚、留言、分享／轉推、收藏（依平台而定）、觸及或曝光或追蹤者數。
**關卡：** 所有要比較的貼文，分母要一致。
### 階段 2：核心演算法
1. 加總每則貼文的互動：按讚 + 留言 + 分享（+ 收藏、點擊，如果有的話）
2. 想的話可以替互動加權：分享=3×、留言=2×、按讚=1×（分享代表更高的投入度）
3. 除以選定的分母（優先用觸及，追蹤者當備案）
4. 計算：每則貼文的 ER、各貼文的平均 ER、ER 中位數、ER 隨時間的趨勢
### 階段 3：驗證
和平台基準比較。標記異常（ER > 20% 很可能是資料錯誤，或是爆紅的離群值）。
**關卡：** 結果在該平台的合理範圍內。
### 階段 4：輸出
回傳互動指標，並附上基準比較的脈絡。
## 輸出格式
```json
{
"metrics": {"avg_er_by_reach": 3.2, "avg_er_by_followers": 1.8, "median_er": 2.9, "top_post_er": 8.5},
"benchmark": {"platform": "instagram", "industry": "fashion", "benchmark_er": 2.5, "percentile": 72},
"metadata": {"posts_analyzed": 30, "period": "2025-Q1", "denominator": "reach"}
}
```
## 範例
### 輸入輸出範例
**輸入：** 貼文：150 個讚、20 則留言、5 次分享，觸及=5000
**預期：** 依觸及的 ER = (150+20+5)/5000 × 100% = 3.5%
### 邊界情況
| 輸入 | 預期 | 原因 |
|-------|----------|-----|
| 觸及 = 0 | 未定義，跳過該貼文 | 不能除以零 |
| 加碼／付費貼文 | 和自然流量分開 | 付費觸及會墊大分母，壓低 ER |
| 爆紅的離群值（10 倍平均） | 標記，另外分析 | 會扭曲平均值 |
## 常見陷阱
- **平台演算法變動**：Instagram 的演算法會定期改變。過往的 ER 基準會過時。要用滾動 90 天的基準。
- **虛榮指標陷阱**：高 ER 不代表有商業影響。迷因貼文的 1000 個讚 ≠ 產品貼文的 10 次連結點擊。要追蹤有意義的互動。
- **限時動態／Reel 的指標不同**：限時動態的互動（點擊、回覆）和 Reel 的互動（播放、分享），需要和動態貼文不同的公式。不要混用。
- **依追蹤者的 ER 雜訊多**：不是所有追蹤者都會看到每一則貼文（觸及 < 追蹤者）。依追蹤者算的 ER，會低估真正看到貼文的人的互動。
- **跨帳號規模比較**：小帳號依追蹤者算的 ER，天生就比較高。要做正規化，或依帳號規模分群，比較才公平。
## 參考資料
- 各平台的基準資料，見 `references/platform-benchmarks.md`
- 加權互動評分模型，見 `references/weighted-engagement.md`英文原文
---
name: "algo-social-engagement"
description: "Calculate and benchmark social media engagement rates across platforms and variants. Use this skill when the user needs to compute engagement metrics, compare performance across accounts or posts, or set engagement benchmarks — even if they say 'what is my engagement rate', 'benchmark engagement', or 'social media KPIs'."
metadata:
category: "WP-38 社群演算法"
tags: ["social-media", "engagement-rate", "analytics", "benchmarking"]
---
# Engagement Rate Calculation
## Overview
Engagement rate measures audience interaction relative to reach or audience size. Formula: (reactions + comments + shares) / denominator × 100%. The denominator choice (reach, impressions, followers) significantly affects the result. Computes in O(n) per post set.
## When to Use
**Trigger conditions:**
- Computing engagement metrics for social media reporting
- Benchmarking account or post performance against industry averages
- Comparing content performance across posts or accounts
**When NOT to use:**
- When evaluating influence holistically (use influence measurement)
- When modeling content spread dynamics (use virality models)
## Algorithm
```
IRON LAW: Engagement Rate Denominator MATTERS
By reach, by impressions, and by followers produce DIFFERENT numbers:
- ER by Reach = engagements / reach × 100% (most accurate, requires analytics access)
- ER by Impressions = engagements / impressions × 100% (always lower than by reach)
- ER by Followers = engagements / followers × 100% (public data, but inflated by non-reaching followers)
ALWAYS specify which variant when reporting or comparing.
```
### Phase 1: Input Validation
Collect per post: likes, comments, shares/retweets, saves (platform-specific), reach or impressions or follower count.
**Gate:** Consistent denominator across all posts being compared.
### Phase 2: Core Algorithm
1. Sum engagements per post: likes + comments + shares (+ saves, clicks if available)
2. Weight engagements if desired: share=3×, comment=2×, like=1× (shares indicate higher commitment)
3. Divide by chosen denominator (reach preferred, followers as fallback)
4. Compute: per-post ER, average ER across posts, median ER, ER trend over time
### Phase 3: Verification
Compare against platform benchmarks. Flag anomalies (ER > 20% likely data error or viral outlier).
**Gate:** Results within plausible range for platform.
### Phase 4: Output
Return engagement metrics with benchmarking context.
## Output Format
```json
{
"metrics": {"avg_er_by_reach": 3.2, "avg_er_by_followers": 1.8, "median_er": 2.9, "top_post_er": 8.5},
"benchmark": {"platform": "instagram", "industry": "fashion", "benchmark_er": 2.5, "percentile": 72},
"metadata": {"posts_analyzed": 30, "period": "2025-Q1", "denominator": "reach"}
}
```
## Examples
### Sample I/O
**Input:** Post: 150 likes, 20 comments, 5 shares, reach=5000
**Expected:** ER by reach = (150+20+5)/5000 × 100% = 3.5%
### Edge Cases
| Input | Expected | Why |
|-------|----------|-----|
| Reach = 0 | Undefined, skip post | Can't divide by zero |
| Boosted/paid post | Separate from organic | Paid reach inflates denominator, deflates ER |
| Viral outlier (10x avg) | Flag, analyze separately | Skews averages |
## Gotchas
- **Platform algorithm changes**: Instagram's algorithm shifts regularly. Historical ER benchmarks become outdated. Use rolling 90-day benchmarks.
- **Vanity metric trap**: High ER doesn't mean business impact. 1000 likes on a meme ≠ 10 link clicks on a product post. Track meaningful engagements.
- **Story/Reel metrics differ**: Story engagement (taps, replies) and Reel engagement (plays, shares) need different formulas than feed posts. Don't mix.
- **Follower-based ER is noisy**: Not all followers see each post (reach < followers). ER by followers underestimates true engagement among those who saw the post.
- **Comparing across account sizes**: Smaller accounts naturally have higher ER by followers. Normalize or segment by account size for fair comparison.
## References
- For platform-specific benchmark data, see `references/platform-benchmarks.md`
- For weighted engagement scoring models, see `references/weighted-engagement.md`