Chip-Retail-Uplift-Testing

Chip Retail Uplift Testing & Experimentation

📌 Project Overview

This project evaluates the impact of a retail trial intervention on chip sales using uplift testing and controlled experimentation. The goal is to determine whether trial stores experienced a statistically meaningful improvement compared to matched control stores, and to translate those findings into actionable business recommendations.

This analysis mirrors a real-world retail analytics workflow: data preparation → control selection → trial evaluation → uplift measurement → recommendation.

🧠 Key Questions

• Did the trial stores outperform comparable control stores?

• Is the observed uplift statistically significant?

• Which stores benefited from the trial, and which did not?

• Should the intervention be rolled out, modified, or discontinued?

📊 Data Description

The project uses transactional and customer behavior data from a grocery retail context.

Main datasets

• QVI_data.csv – aggregated sales and transaction metrics (large file, tracked with Git LFS)

• QVI_transaction_data.xlsx – detailed transaction-level data

• QVI_purchase_behaviour.csv – customer purchasing behavior and segmentation data

⚠️ Large files are tracked using Git LFS. If cloning this repository, ensure Git LFS is installed.

🔬 Methodology

1. Exploratory Data Analysis (EDA)

• Examined data structure, time coverage, and variable distributions

• Identified missing values, anomalies, and data quality issues

• Explored sales trends, seasonality, and customer segment behavior

• Informed downstream feature engineering and control-store selection

2. Data Cleaning & Feature Engineering

• Date normalization and temporal aggregation

• Construction of sales, volume, and transaction-level metrics

• Creation of customer segmentation and behavioral features

3. Control Store Selection

• Identification of suitable control stores based on historical similarity

• Visual and quantitative validation of pre-trial alignment

4. Trial vs Control Comparison

• Pre-trial vs trial-period performance analysis

• Time-series comparisons of key performance indicators

5. Uplift Measurement

• Estimation of incremental sales and performance uplift

• Store-level evaluation rather than pooled averages

6. Business Interpretation & Recommendation

• Clear recommendation for each trial store

• Practical implications for rollout, refinement, or discontinuation

📈 Key Outputs

-Trial vs control performance visualizations

• Store-level uplift analysis

• Reproducible Jupyter notebooks documenting the full workflow