QVI Trial Store Performance Analysis 📈

Project Overview

This project evaluates the effectiveness of a new in-store layout trial conducted across three selected stores. The primary objective was to determine if the new layout design led to a statistically significant increase in sales compared to a baseline. By identifying suitable control stores and analyzing performance lifts, the project delivers data-driven recommendations regarding a potential company-wide rollout.

The Trial

• Trial Stores: 77, 86, 88
• Trial Period: February 2019 – April 2019
• Pre-Trial Baseline: July 2018 – January 2019

Data Source

The analysis utilizes the full QVI transaction dataset.

Note: The QVI_data.csv file exceeds 25MB and is not included in this repository.

Methodology

1. Data Preparation: Monthly performance metrics (total sales, customer count, and transactions) were aggregated from the QVI dataset.
2. Control Store Selection: For each trial store, a matching control store was identified based on similarity in pre-trial performance. Selection was based on a composite score calculating the magnitude distance of sales, customer count, and transactions per customer.
3. Impact Assessment: The total sales of each trial store were compared against its respective control store during the trial period to quantify the percentage lift.

Repository Structure

.
├── trial_analysis.R               # R script for analysis and visualization
├── README.md                      # Project documentation
└── plots/                         # Generated visualization files

Tools & Requirements

• Language: R (version 4.5.x)
• Libraries: data.table, ggplot2, lubridate

How to Run

1. Clone this repository to your local machine.
2. Open trial_analysis.R in RStudio.
3. Set the working directory:

   setwd("path/to/your/repository")

4. Install required packages if necessary:

   install.packages(c("data.table", "ggplot2", "lubridate"))

5. Run the script. It will identify control stores, generate comparative visualizations, and output the performance lift for each trial store.

Key Findings

Control Store Matching

• Store 77 matched with Control Store 268
• Store 86 matched with Control Store 225
• Store 88 matched with Control Store 237

Performance Results

The trial layout yielded positive results across all locations:

Trial Store	Control Store	Sales Lift
77	268	+33.62%
86	225	+5.69%
88	237	+12.29%

Recommendation

Proceed with a full-scale rollout. The consistent and significant positive sales lift observed across all three trial stores provides strong evidence that the new layout effectively drives revenue. A company-wide implementation is recommended to maximize growth in the chip category.

Author

Siyamthanda Amahle Buthelezi