ingest: support-analytics-reporter.md
- Source page: Analytics Reporter Agent Personality - Concepts: RFM-Analysis, Marketing-Attribution - Updated: index.md (entry fix), overview.md (Support dept), log.md
This commit is contained in:
102
wiki/concepts/Marketing-Attribution.md
Normal file
102
wiki/concepts/Marketing-Attribution.md
Normal file
@@ -0,0 +1,102 @@
|
||||
---
|
||||
title: "Marketing Attribution"
|
||||
type: concept
|
||||
tags: []
|
||||
sources: [support-analytics-reporter]
|
||||
last_updated: 2026-04-21
|
||||
---
|
||||
|
||||
## Aliases
|
||||
- 营销归因模型
|
||||
- Multi-Touch Attribution（多触点归因）
|
||||
- Attribution Modeling
|
||||
|
||||
## Definition
|
||||
Marketing Attribution（营销归因）是一种数据分析方法，用于将客户的转化收入或转化行为按照用户旅程中各触点（渠道/广告位/活动）的贡献权重进行分配，从而量化不同营销渠道的真实价值，指导预算优化和 ROI 最大化。
|
||||
|
||||
## Attribution Models
|
||||
|
||||
### 1. Single-Touch（单触点归因）
|
||||
| 模型 | 归因逻辑 | 优点 | 缺点 |
|
||||
|------|---------|------|------|
|
||||
| First Touch | 100% 归因给首个触点 | 识别获客渠道 | 忽视转化路径上的其他渠道 |
|
||||
| Last Touch | 100% 归因给末触点 | 识别转化触点 | 忽视品牌建设类触点 |
|
||||
|
||||
### 2. Multi-Touch（多触点归因）
|
||||
| 模型 | 归因逻辑 | 适用场景 |
|
||||
|------|---------|---------|
|
||||
| Linear | 平均分配权重 | 各触点均等重要 |
|
||||
| Time Decay | 越接近转化时间权重越高 | 短转化周期（B2C电商） |
|
||||
| Position Based (U-Shaped) | 首+末各 40%，中间均分剩余 20% | 品牌+效果兼顾 |
|
||||
| Data-Driven | 基于 Shapley 值或机器学习模型 | 有足够转化数据支撑 |
|
||||
|
||||
### 3. Algorithmic Attribution（算法归因）
|
||||
基于 Shapley 值（博弈论）或 logistic 回归/马尔可夫链模型，从数据中自动学习各触点权重，是最精确但数据需求量最大的方案。
|
||||
|
||||
## Multi-Touch Attribution Implementation
|
||||
|
||||
```sql
|
||||
-- Multi-touch attribution with first/last/intermediate weights
|
||||
WITH customer_touchpoints AS (
|
||||
SELECT
|
||||
customer_id,
|
||||
channel,
|
||||
campaign,
|
||||
touchpoint_date,
|
||||
conversion_date,
|
||||
revenue,
|
||||
ROW_NUMBER() OVER (PARTITION BY customer_id ORDER BY touchpoint_date) as touch_sequence,
|
||||
COUNT(*) OVER (PARTITION BY customer_id) as total_touches
|
||||
FROM marketing_touchpoints mt
|
||||
JOIN conversions c ON mt.customer_id = c.customer_id
|
||||
WHERE touchpoint_date <= conversion_date
|
||||
),
|
||||
attribution_weights AS (
|
||||
SELECT *,
|
||||
CASE
|
||||
WHEN touch_sequence = 1 AND total_touches = 1 THEN 1.0 -- Single touch
|
||||
WHEN touch_sequence = 1 THEN 0.4 -- First touch
|
||||
WHEN touch_sequence = total_touches THEN 0.4 -- Last touch
|
||||
ELSE 0.2 / (total_touches - 2) -- Middle touches
|
||||
END as attribution_weight
|
||||
FROM customer_touchpoints
|
||||
)
|
||||
SELECT
|
||||
channel,
|
||||
campaign,
|
||||
SUM(revenue * attribution_weight) as attributed_revenue,
|
||||
COUNT(DISTINCT customer_id) as attributed_conversions
|
||||
FROM attribution_weights
|
||||
GROUP BY channel, campaign
|
||||
ORDER BY attributed_revenue DESC;
|
||||
```
|
||||
|
||||
## Campaign ROI Calculation
|
||||
|
||||
```sql
|
||||
SELECT
|
||||
campaign_name,
|
||||
SUM(spend) as total_spend,
|
||||
SUM(attributed_revenue) as total_revenue,
|
||||
(SUM(attributed_revenue) - SUM(spend)) / SUM(spend) * 100 as roi_percentage,
|
||||
SUM(attributed_revenue) / SUM(spend) as revenue_multiple,
|
||||
COUNT(conversions) as total_conversions,
|
||||
SUM(spend) / COUNT(conversions) as cost_per_conversion
|
||||
FROM campaign_performance
|
||||
GROUP BY campaign_name
|
||||
ORDER BY roi_percentage DESC;
|
||||
```
|
||||
|
||||
## Key Metrics
|
||||
|
||||
| 指标 | 公式 | 业务含义 |
|
||||
|------|------|---------|
|
||||
| ROI | (归因收入 - 花费) / 花费 × 100% | 渠道盈利性 |
|
||||
| ROAS | 归因收入 / 广告花费 | 广告效率 |
|
||||
| CPA | 总花费 / 归因转化数 | 获客成本 |
|
||||
| Revenue Multiple | 归因收入 / 花费 | 收入倍数 |
|
||||
|
||||
## Connections
|
||||
- [[support-analytics-reporter]] — 使用多触点归因模型进行营销效果分析
|
||||
- [[Marketing-ROI]] — 归因分析是 ROI 计算的基础
|
||||
- [[Business-Intelligence]] — 属 BI 领域的营销分析子方向
|
||||
79
wiki/concepts/RFM-Analysis.md
Normal file
79
wiki/concepts/RFM-Analysis.md
Normal file
@@ -0,0 +1,79 @@
|
||||
---
|
||||
title: "RFM Analysis"
|
||||
type: concept
|
||||
tags: []
|
||||
sources: [support-analytics-reporter]
|
||||
last_updated: 2026-04-21
|
||||
---
|
||||
|
||||
## Aliases
|
||||
- RFM Segmentation
|
||||
- Recency, Frequency, Monetary Analysis
|
||||
- 客户价值分层分析
|
||||
|
||||
## Definition
|
||||
RFM Analysis 是一种三维客户价值分层方法，通过最近购买时间（Recency）、购买频率（Frequency）、消费金额（Monetary）三个维度对客户进行分群，从而识别高价值客户、流失风险客户和潜力客户，为精准营销和客户运营提供数据支撑。
|
||||
|
||||
## Core Metrics
|
||||
|
||||
| 维度 | 定义 | 计算方式 |
|
||||
|------|------|----------|
|
||||
| Recency | 最近一次购买距今天数 | `当前日期 - 最近购买日期`（越小越好） |
|
||||
| Frequency | 购买总次数 | `COUNT(order_id)`（越大越好） |
|
||||
| Monetary | 累计消费金额 | `SUM(revenue)`（越大越好） |
|
||||
|
||||
## Scoring Method
|
||||
|
||||
每个维度按分位数（通常5分位）打分：
|
||||
- R_Score：最近购买时间越近，分数越高（1-5分，5=最近）
|
||||
- F_Score：购买频率越高，分数越高（1-5分，5=最频繁）
|
||||
- M_Score：消费金额越高，分数越高（1-5分，5=最高金额）
|
||||
|
||||
组合 R+F+M 得 RFM Score（如 `555`、`311`）。
|
||||
|
||||
## Customer Segments
|
||||
|
||||
| RFM Score | 客户类型 | 策略建议 |
|
||||
|-----------|---------|---------|
|
||||
| 555/554/544/545/454/455/445 | Champions（冠军客户） | 奖励忠诚度，邀请推荐，升级销售高端产品 |
|
||||
| 543/444/435/355/354/345/344/335 | Loyal Customers（忠诚客户） | 培养关系，推荐新品，忠诚度计划 |
|
||||
| 553/551/552/541/542/533/532/531/452/451 | Potential Loyalists（潜力忠诚者） | 入会欢迎，早期参与，产品教育 |
|
||||
| 512/511/422/421/412/411/311 | New Customers（新客户） | 入职优化，早期参与 |
|
||||
| 155/154/144/214/215/115/114 | At Risk（流失风险客户） | 重新参与活动，特殊优惠，赢回策略 |
|
||||
| 其他 | Others（一般客户） | 常规触达，持续观察 |
|
||||
|
||||
## Implementation
|
||||
|
||||
```python
|
||||
import pandas as pd
|
||||
import numpy as np
|
||||
|
||||
def rfm_analysis(df, current_date=None):
|
||||
"""RFM Analysis implementation"""
|
||||
if current_date is None:
|
||||
current_date = df['date'].max()
|
||||
|
||||
rfm = df.groupby('customer_id').agg({
|
||||
'date': lambda x: (current_date - x.max()).days, # Recency
|
||||
'order_id': 'count', # Frequency
|
||||
'revenue': 'sum' # Monetary
|
||||
}).rename(columns={
|
||||
'date': 'recency',
|
||||
'order_id': 'frequency',
|
||||
'revenue': 'monetary'
|
||||
})
|
||||
|
||||
# Create quintile scores (1-5)
|
||||
rfm['r_score'] = pd.qcut(rfm['recency'], 5, labels=[5,4,3,2,1], duplicates='drop')
|
||||
rfm['f_score'] = pd.qcut(rfm['frequency'].rank(method='first'), 5, labels=[1,2,3,4,5], duplicates='drop')
|
||||
rfm['m_score'] = pd.qcut(rfm['monetary'], 5, labels=[1,2,3,4,5], duplicates='drop')
|
||||
|
||||
rfm['rfm_score'] = rfm['r_score'].astype(str) + rfm['f_score'].astype(str) + rfm['m_score'].astype(str)
|
||||
|
||||
return rfm
|
||||
```
|
||||
|
||||
## Connections
|
||||
- [[support-analytics-reporter]] — 使用 RFM 进行客户价值分层分析
|
||||
- [[Customer-Segmentation]] — RFM 是客户细分的核心方法之一
|
||||
- [[Business-Intelligence]] — 属 BI 领域的客户分析子方向
|
||||
Reference in New Issue
Block a user