Customer Segmentation & Campaign Analysis

An RFM (Recency, Frequency, Monetary) segmentation study on 3,745 customers from the UCI Online Retail II dataset, layered with marketing channel performance analysis to identify where budget should go.

Key Findings

KPI	Value
Customers	3,745
Total Revenue	$882K
Avg Order Value	$70
Total Invoices	12,580
Segments	8
Channels	5
Period	Dec 2010 -- Dec 2011

1. Champions Are the Business

Champions make up 20.8% of customers but generate 61.3% of all revenue ($540K). Losing even a small fraction of Champions would hurt more than losing most other segments combined.

2. 922 Dormant Customers Represent a Win-Back Opportunity

Dormant is the largest segment by headcount (24.6%) but contributes only $31K in revenue. These customers have bought before but haven't returned -- a targeted re-engagement campaign could recover an estimated $4,600.

3. Email Is the Highest-Revenue Channel

Channel	Customers	Revenue	Avg Order	Cost/Customer	Avg ROI
Email	1,199	$287,133	$240	$2.50	9,479%
Paid Social	739	$186,923	$253	$18.00	1,305%
Referral	619	$178,385	$288	$7.50	3,742%
Organic Search	827	$136,062	$165	$5.00	3,191%
Direct	361	$93,486	$259	$1.00	25,797%

Email dominates Champion acquisition. Paid Social disproportionately brings in Dormant customers, explaining its weaker ROI despite high revenue volume.

4. Full Segment Breakdown

Segment	Customers	Avg Recency	Avg Orders	Avg Spend	Total Revenue
Champions	779	16 days	8.9	$694	$540,322
Needs Attention	432	169 days	2.0	$312	$134,864
Loyal Customers	706	33 days	3.0	$135	$94,928
At Risk	169	126 days	4.2	$302	$51,022
Dormant	922	193 days	1.1	$34	$30,893
Potential Loyalists	572	38 days	1.4	$48	$27,545
Lost	132	272 days	1.0	$15	$1,921
New Customers	33	28 days	1.0	$15	$495

RFM Methodology

Each customer receives three independent scores from 1 to 5:

Dimension	What It Measures	Score Logic
Recency	Days since last purchase	Score 5 = bought recently, Score 1 = long time ago
Frequency	Number of unique orders placed	Score 5 = orders frequently, Score 1 = ordered once
Monetary	Total money spent	Score 5 = highest spenders, Score 1 = lowest spenders

The combined RFM score (3--15) determines which of the 8 behavioural segments a customer belongs to.

Strategic Recommendations

1. Protect Champions at All Costs -- Build a dedicated loyalty programme, early product access, and personalised email communication for this group. Losing 10% of Champions costs more revenue than losing 50% of Dormant customers.
2. Launch a Win-Back Campaign for Dormant Customers -- A targeted re-engagement email sequence with personalised discounts and past-purchase recommendations could convert 15% back into active buyers, recovering ~$4,600.
3. Reduce Paid Social Spend, Invest in Referral -- Paid Social costs $18/customer and attracts lower-value customers. Referral costs $7.50 and delivers the highest avg order value ($288). Reallocating 30% of Paid Social budget would improve ROI significantly.
4. Nurture Potential Loyalists Early -- 572 customers have bought recently but infrequently. A triggered email sequence after their second purchase with a loyalty incentive could move them into the Loyal segment within 90 days.

Project Structure

.
├── marketing_analysis.py                # RFM segmentation + channel performance analysis
├── retail_sample.csv                    # Sampled dataset (from UCI Online Retail II)
├── marketing_dashboard.html             # Interactive HTML dashboard with Chart.js
├── marketing_analytics_portfolio.pdf    # 9-page PDF report with charts and findings
└── README.md

File Descriptions

• marketing_analysis.py -- End-to-end pipeline: data cleaning, RFM score calculation, customer segmentation into 8 groups, simulated campaign channel assignment, ROI computation, and 5 Matplotlib visualisations (donut chart, revenue bars, channel performance, RFM scatter plot, channel mix).
• retail_sample.csv -- Sampled subset of the UCI Online Retail II dataset with columns: Invoice, StockCode, Description, Quantity, InvoiceDate, Price, Customer ID, Country.
• marketing_dashboard.html -- Self-contained interactive dashboard featuring: KPI strip, segment donut chart, revenue by segment bar chart, full segment breakdown table, channel revenue vs ROI, channel mix by segment, channel performance table, and 4 strategic recommendations.
• marketing_analytics_portfolio.pdf -- 9-page PDF report with executive summary, RFM methodology, segmentation results, customer behaviour map, channel performance analysis, and strategic recommendations.

Dataset

• Source: UCI Online Retail II
• Citation: Chen, D. (2015). Online Retail II [Dataset]. UCI Machine Learning Repository.
• Licence: CC BY 4.0
• Period: December 2010 -- December 2011
• Records: ~500K transactions (sampled for this analysis)

Tools & Technologies

• Python 3 -- Core analysis language
• Pandas & NumPy -- Data manipulation, RFM computation
• Matplotlib & Seaborn -- Static chart generation
• ReportLab -- PDF report generation
• Chart.js -- Interactive browser-based visualisations
• HTML/CSS -- Dashboard layout and styling

How to Run

# Install dependencies
pip install pandas numpy matplotlib seaborn

# Run the analysis
python marketing_analysis.py

# View the interactive dashboard
open marketing_dashboard.html    # macOS
xdg-open marketing_dashboard.html  # Linux

Author

Gbolahan Akande -- Data Analyst

---

Portfolio case study | Marketing Analytics | RFM Segmentation