Sales Performance Analysis & Revenue Forecasting (Retail Case Study)

Overview

This project analyses 12 months of retail sales data (≈185k transactions) to understand what drives revenue and to build a simple, transparent forecasting approach for a mid-size retailer. The focus is on identifying high-value products, understanding pricing behaviour, and uncovering time patterns that can inform promotions, staffing and inventory decisions.

Although the dataset is based on the well-known Superstore sample, it is treated here as a proxy for a real client: a retail business looking for clear, data-backed answers to commercial questions.

---

Business Context & Key Questions

From the perspective of a Head of Sales / Commercial Manager, this analysis is designed to answer:

1. Which products and categories drive most of our revenue and margin?
2. How do price and discounting affect demand? Are there price points where elasticity changes noticeably?
3. When do we actually sell the most? (days of the week, times of day, seasonality over the year)
4. What can we reasonably expect future sales to look like, assuming no major structural changes?
5. Which commercial levers should we pull (promotions, pricing, assortment, inventory focus) to improve performance?

---

Data

• Source: Superstore-style transactional dataset (orders, customers, products, categories, regions, order dates, quantities, revenue, discounts).
• Grain: One row per order line.
• Time horizon: ~12 months of data.
• Size: ≈185,000 rows after basic cleaning and filtering.

Key fields used:

• Order Date, Ship Date
• Product ID, Product Name, Category, Sub-Category
• Region, City, Customer Segment
• Sales, Quantity, Discount, Profit

---

Approach

The project is implemented in Python using Jupyter Notebooks and follows a business-first workflow:

1. Data Cleaning & Preparation

   • Handle missing values and data types.
   • Create derived variables (e.g. gross revenue, weekday/weekend flags, hour-of-day buckets).
   • Ensure consistent date/time handling.

2. Descriptive & Diagnostic Analysis

   • Basic KPIs: total revenue, orders, units, average basket size.
   • Revenue/margin breakdown by:
     • Product, Sub-category, Category
     • Region and Customer Segment
   • Pareto analysis to identify the small set of products generating the majority of revenue.

3. Pricing & Discount Behaviour

   • Explore distribution of prices and discounts.
   • Identify price bands / discount bands where demand behaves differently.
   • Examine how heavy discounting interacts with margin and volume.

4. Time-Based Patterns

   • Daily and weekly seasonality (day of week, week of year).
   • Hour-of-day patterns to understand when customers actually buy.
   • Identify peaks such as mid-week (Tue–Thu) daytime sales concentration.

5. Simple Time-Series Forecasting

   • Aggregate sales to a regular time-series (e.g. daily/weekly revenue).
   • Fit baseline forecasting models (e.g. moving average, naïve, simple time-series models).
   • Compare forecast performance and produce short-term projections.

6. Recommendations

   • Translate analytical findings into concrete actions across promotions, assortment and inventory planning.

---

Repository Structure

• data/
  • Sample - Superstore.csv – transactional sales dataset.
• notebooks/
  • sales_analysis.ipynb – main notebook with sales performance analysis and visualisations.
• README.md – project overview and instructions.

---

Tools & Libraries

• Language: Python 3.x
• Core libraries:
  • pandas – data wrangling and aggregation
  • numpy – numerical operations
  • matplotlib, seaborn – visualisation
  • datetime / dateutil – date/time handling
• Environment: Jupyter Notebook

---

How to Run the Analysis

1. Clone the repository:

   git clone https://github.com/Kimuyu-Charles/Sales-Analysis.git
   cd Sales-Analysis