algo-social-influence
用互動加權指標評估網紅與帳號的影響力
中文翻譯
---
name: "algo-social-influence"
description: "用互動加權指標衡量社群媒體影響力，不只看追蹤者數量。當使用者需要評估網紅成效、比較不同帳號的影響力、或建立影響力評分系統時使用，即使他們說的是「誰比較有影響力」、「網紅排名」或「衡量社群影響」也一樣。"
metadata:
category: "WP-38 社群演算法"
tags: ["social-media", "influence", "influencer-marketing", "metrics"]
---
# 社群影響力衡量（Social Influence Measurement）
## 概述
影響力評分評估的是帳號驅動行動（互動、分享、轉換）的能力，不只是觸及而已。結合觸及（reach）、共鳴（resonance，互動深度）和相關性（relevance，主題權威度）。以加權綜合分數計算。
## 何時使用
**觸發條件：**
- 要為行銷活動評估並比較網紅
- 要建立影響力評分或排名系統
- 要評估品牌大使的成效
**何時不要用：**
- 要衡量內容的病毒式擴散動態時（用病毒式傳播模型）
- 只要計算基本的互動率時（用互動率計算器）
## 演算法
```
鐵律：追蹤者數量 ≠ 影響力
影響力需要「互動」。有 100 萬追蹤者、
互動率 0.01% 的帳號，影響力比不上
1 萬追蹤者、互動率 5% 的帳號。
衡量方式：觸及 × 互動率 × 相關性。
```
### 階段 1：輸入驗證
每個帳號要蒐集：追蹤者數、每則貼文的平均按讚／留言／分享、發文頻率、受眾輪廓、主題類別。
**關卡：** 至少 20 則近期貼文，指標才穩定。
### 階段 2：核心演算法
1. **觸及分數**：把追蹤者數正規化到對數尺度（邊際效益遞減）
2. **互動分數**：(平均互動數 / 追蹤者) × 100，依類型加權（分享 > 留言 > 按讚）
3. **相關性分數**：網紅內容和目標活動之間的主題重疊度
4. **綜合**：Influence = w₁×Reach + w₂×Engagement + w₃×Relevance（權重依活動目標調整）
5. 調整項目：受眾真實度（機器人追蹤者的 %）、發文頻率的一致性
### 階段 3：驗證
抽查：高分的帳號真的能驅動行動嗎？如果有的話，和歷史活動成效資料交叉比對。
**關卡：** 排名靠前的帳號，有可證明的互動歷史。
### 階段 4：輸出
回傳排名後的影響力分數，附上各成分的拆解。
## 輸出格式
```json
{
"rankings": [{"account": "@handle", "influence_score": 82, "reach": 75, "engagement": 90, "relevance": 85}],
"metadata": {"accounts_analyzed": 50, "weights": {"reach": 0.2, "engagement": 0.5, "relevance": 0.3}}
}
```
## 範例
### 輸入輸出範例
**輸入：** 帳號 A：50 萬追蹤者、0.5% 互動率。帳號 B：5 萬追蹤者、4.2% 互動率。相關性相同。
**預期：** B 分數較高，因為權重中互動佔主導。
### 邊界情況
| 輸入 | 預期 | 原因 |
|-------|----------|-----|
| 一炮而紅的帳號 | 近期互動高，但穩定度低 | 需要做時間一致性檢查 |
| 互動率低的名人 | 觸及高，但每塊錢的影響力低 | 只看觸及的策略，很貴 |
| 利基型的微網紅 | 相關性高 + 互動高 | 對精準鎖定的活動，投資報酬率最好 |
## 常見陷阱
- **假互動**：機器人的讚／留言會灌水指標。要用真實度工具（HypeAuditor 等）來偵測。
- **平台差異**：Instagram 上 2% 的互動率是平均；Twitter/X 上 2% 則很優秀。要依平台基準做正規化。
- **互動群組（engagement pods）**：一群網紅人為地互相互動。要檢查互動是否來自多元的來源。
- **影響力 ≠ 轉換**：高互動不保證有購買意願。要追蹤下游指標（連結點擊、優惠碼使用），才能算活動的 ROI。
- **時間衰減**：影響力會變。至少每季重新評估；變化快的類別，每月更好。
## 參考資料
- 受眾真實度的偵測方法，見 `references/authenticity-detection.md`
- 網紅 ROI 的衡量框架，見 `references/influencer-roi.md`英文原文
---
name: "algo-social-influence"
description: "Measure social media influence using engagement-weighted metrics beyond follower count. Use this skill when the user needs to evaluate influencer effectiveness, compare influence across accounts, or build an influence scoring system — even if they say 'who is more influential', 'influencer ranking', or 'measure social impact'."
metadata:
category: "WP-38 社群演算法"
tags: ["social-media", "influence", "influencer-marketing", "metrics"]
---
# Social Influence Measurement
## Overview
Influence scoring evaluates an account's ability to drive actions (engagement, sharing, conversions) beyond mere reach. Combines reach, resonance (engagement depth), and relevance (topical authority). Computes as weighted composite score.
## When to Use
**Trigger conditions:**
- Evaluating and comparing influencers for marketing campaigns
- Building an influence scoring or ranking system
- Assessing brand ambassador effectiveness
**When NOT to use:**
- When measuring content virality dynamics (use viral spread models)
- When computing basic engagement rates (use engagement rate calculator)
## Algorithm
```
IRON LAW: Follower Count ≠ Influence
Influence requires ENGAGEMENT. An account with 1M followers and
0.01% engagement rate has less influence than one with 10K followers
and 5% engagement. Measure: reach × engagement rate × relevance.
```
### Phase 1: Input Validation
Collect per account: follower count, avg likes/comments/shares per post, posting frequency, audience demographics, topic categories.
**Gate:** Minimum 20 recent posts for stable metrics.
### Phase 2: Core Algorithm
1. **Reach score**: Normalize follower count to log scale (diminishing returns)
2. **Engagement score**: (avg engagements / followers) × 100, weighted by type (share > comment > like)
3. **Relevance score**: Topic overlap between influencer content and target campaign
4. **Composite**: Influence = w₁×Reach + w₂×Engagement + w₃×Relevance (weights tuned per campaign goal)
5. Adjust for: audience authenticity (bot follower %), post frequency consistency
### Phase 3: Verification
Spot-check: do high-scoring accounts actually drive actions? Cross-reference with historical campaign performance data if available.
**Gate:** Top-ranked accounts have demonstrable engagement history.
### Phase 4: Output
Return ranked influence scores with component breakdown.
## Output Format
```json
{
"rankings": [{"account": "@handle", "influence_score": 82, "reach": 75, "engagement": 90, "relevance": 85}],
"metadata": {"accounts_analyzed": 50, "weights": {"reach": 0.2, "engagement": 0.5, "relevance": 0.3}}
}
```
## Examples
### Sample I/O
**Input:** Account A: 500K followers, 0.5% engagement. Account B: 50K followers, 4.2% engagement. Same relevance.
**Expected:** B scores higher due to engagement dominance in weighting.
### Edge Cases
| Input | Expected | Why |
|-------|----------|-----|
| Viral one-hit account | High recent engagement, low stability | Need temporal consistency check |
| Celebrity with low engagement | High reach, low influence per dollar | Reach-only strategy, expensive |
| Micro-influencer niche | High relevance + engagement | Best ROI for targeted campaigns |
## Gotchas
- **Fake engagement**: Bot likes/comments inflate metrics. Use authenticity tools (HypeAuditor, etc.) to detect.
- **Platform differences**: 2% engagement on Instagram is average; 2% on Twitter/X is excellent. Normalize by platform benchmarks.
- **Engagement pods**: Groups of influencers artificially engaging with each other's content. Check if engagement comes from diverse sources.
- **Influence ≠ conversion**: High engagement doesn't guarantee purchase intent. Track downstream metrics (link clicks, promo code usage) for campaign ROI.
- **Temporal decay**: Influence changes. Quarterly reassessment is minimum; monthly is better for fast-moving categories.
## References
- For audience authenticity detection methods, see `references/authenticity-detection.md`
- For influencer ROI measurement framework, see `references/influencer-roi.md`