🛒 39,357	📦 38,765	👥 3,898	🏪 3	📊 9	🔬 IQR
Products catalogued	Transactions analysed	Customers segmented	Platforms compared	Dashboard modules	Price quality audit

---

📌 Business Questions Answered

• Which platform wins the price war by category?
• Why do 22.8% of customers churn — and what is the revenue risk?
• Which customer acquisition cohort has the highest lifetime value?
• What product category drives the most loyal customers?
• How much revenue can a targeted retention campaign recover?

---

🔍 Key Findings (from real data)

• Champions (20.8%, 809 customers) order every 58 days with 6.3× avg frequency
• Churned segment (22.8%, 899 customers) represents ₹310K+ annual revenue at risk
• Jun 2015 cohort achieved 26.3% Month-1 retention — 84% above the 24-cohort average
• Beverages-first customers show 24.9% churn rate — highest of any first-category
• Business simulator projects ROI of retention campaigns across all 5 RFM segments

Every number is traceable to a notebook cell. Assumptions are labelled. See docs/LIMITATIONS.md.

---

🛠️ Tech Stack

Layer	Technology
Dashboard	Streamlit
Data Processing	Python · Pandas · NumPy
Visualizations	Plotly
Statistical Analysis	SciPy (chi-squared, IQR) · mlxtend (Apriori)
Data Pipeline	Jupyter Notebooks (7-step)
Deployment	Streamlit Cloud

Pipeline outputs feed a dbt-compatible star schema (fct_orders, dim_products, dim_platforms) for downstream analytics engineering workflows.

---

📊 Dashboard Pages

📊 Overview — Platform KPIs & Intelligence Report	⚔️ Price Intelligence — Category Price Gap Heatmap
🛒 Market Basket — Apriori Association Rules	🌊 Customer Journey — Sankey Flow Diagram

Page	What It Shows
📊 Overview + Intelligence Report	Key metrics, auto-generated insights, platform comparison
⚔️ Price Intelligence	Price gap matrix across 15+ categories
⭐ Review & Rating	Rating distributions by platform and category
🛒 Market Basket	Apriori association rules with honest sparse-result disclosure
👥 Customer Segments	RFM treemap, recency vs frequency scatter, chi-squared test
📈 Cohort Retention	24-cohort heatmap, Month-1 retention trends
🌊 Customer Journey	Sankey: first category → segment → outcome
🎯 Business Simulator	Revenue impact projections for retention campaigns
🔬 Data Quality	IQR outlier detection · missing value audit · methodology map

---

🔢 Key Numbers — Notebook-Verified

39,357   products — Blinkit + Zepto + BigBasket catalog CSVs
38,765   grocery transactions (proxy dataset)              [nb 05, Cell 1]
 3,898   unique customers                                  [nb 05, Cell 1]
   809   Champion customers — 20.8% of base               [nb 05, Cell 4]
   899   Churned customers — avg 400 days since last order [nb 05, Cell 5]
   206   At-Risk customers — 5.3% of base                 [nb 05, Cell 4]

Jun-2015 cohort → 26.3% Month-1 retention (avg: 14.3%)   [nb 06, Cell 6]
Beverages first-buyers → 24.9% Churned segment rate       [nb 07, Cell 5]
Champion avg items 16.9 vs Churned 4.7  (3.6× ratio)     [nb 05, Cell 5]

₹350     avg order value                    [assumed — DPIIT/Redseer estimate]
 -2.0    price elasticity (Simulator)       [assumed — standard retail textbook]

---

🚀 Run Locally

git clone https://github.com/Yashaswini-V21/qc-pulse-india
cd qc-pulse-india
pip install -r requirements.txt
streamlit run app.py

To regenerate all pipeline outputs from raw CSVs:

python run_pipeline.py   # runs notebooks 01 → 07 in sequence

---

📁 Project Structure

QC_Pulse_India/
│
├── 📄 app.py                      ← Streamlit entrypoint — routing + CSS injection
├── 📄 run_pipeline.py             ← Notebook orchestrator (01 → 07, no silent substitutions)
├── 📄 requirements.txt            ← Pinned Python dependencies
├── 📄 CHANGELOG.md                ← Version history and integrity fix log
│
├── 📁 views/                      ← One module per dashboard page
│   ├── overview.py                  KPI cards + AI story cards
│   ├── price_intelligence.py        Price gap heatmap + platform scorecard
│   ├── review_rating.py             Rating distributions
│   ├── market_basket.py             Apriori rules + honest sparse-result disclosure
│   ├── customer_segments.py         RFM bubble chart + segment table + chi-squared
│   ├── cohort_retention.py          24-month retention heatmap
│   ├── customer_journey.py          Sankey diagram
│   ├── business_simulator.py        4-tab simulator + assumptions expander
│   └── data_quality.py              Live IQR outlier detection
│
├── 📁 utils/                      ← Shared backend utilities
│   ├── data_loader.py               @st.cache_data CSV loader + column validator
│   ├── simulator.py                 Business projection functions (typed params)
│   ├── story_generator.py           Auto-intelligence text — live from DataFrames
│   ├── styles.py                    Dark theme CSS tokens
│   ├── charts.py                    Plotly dark theme config
│   └── config.py                    File paths + constants
│
├── 📁 notebooks/                  ← Data pipeline (run in order)
│   ├── 01_data_load.ipynb
│   ├── 02_cleaning.ipynb
│   ├── 03_price_intelligence.ipynb
│   ├── 04_basket_analysis.ipynb     min_support=0.005 — as written, no substitution
│   ├── 05_rfm_segmentation.ipynb
│   ├── 06_cohort_retention.ipynb
│   ├── 07_sankey.ipynb
│   └── 08_data_quality.ipynb        IQR price audit
│
├── 📁 data/
│   ├── raw/                         Source CSVs (Blinkit, Zepto, BigBasket, Groceries)
│   └── clean/                       Processed outputs consumed by dashboard
│
├── 📁 docs/
│   ├── LIMITATIONS.md              ← ⭐ L1–L5 data caveat disclosure
│   ├── data_schema.md
│   └── DEVELOPER_GUIDE.md
│
└── 📁 tests/
    └── test_data_loader.py

---

👤 Author

Yashaswini V · LinkedIn · GitHub

---

       

_{QC Pulse India  ·  MIT Licensed  ·  Made with Python, Streamlit & honest data work  ·  2026}