top of page

Olist — Retail Demand Signals and Projections

Understanding regional demand concentration, category drivers, and short term order projections using transactional retail data

Python  • Tableau

Overview

This project analyzes retail demand patterns using the Olist e-commerce dataset, with a focus on understanding how demand behaves across product categories, regions, and time.

The analysis transforms transactional data into practical demand signals that could support operational planning decisions, such as inventory allocation, logistics capacity, and short term demand anticipation.

Rather than relying on a single aggregated demand metric, the project emphasizes segmented demand behavior to reflect the heterogeneous nature of large scale e-commerce operations.

Business Problem

Retail operations rely on accurate demand visibility to support inventory planning, fulfillment efficiency, and regional resource allocation.
Without a clear understanding of where demand concentrates and how it evolves over time, planning decisions tend to be reactive and inefficient.

This analysis addresses the following business questions:

  • Which regions concentrate the highest order demand?

  • Which product categories consistently drive volume?

  • How does daily demand behave over time, and what short-term demand signal does it suggest?

Key Metrics (KPIs)

  • Orders Volume (primary KPI): total number of orders, used as the main indicator of demand behavior over time and across regions

  • Orders Growth Rate: relative change in order volume across periods, used to identify acceleration, deceleration, and seasonality patterns

  • Projected Orders Volume: estimated future order volume based on historical demand signals, used to explore short term demand trends

  • Geographic Distribution of Orders: regional distribution of order volume, used to identify concentration patterns and regional demand dynamics

  • Category Level Orders Distribution: distribution of orders across the selected top product categories, used to analyze demand composition

Analytical Approach

  • Loaded and validated transactional and reference tables from the Olist dataset, including orders, products, customers, and geolocation data

  • Performed data integrity checks to confirm time coverage, missing values, and duplicate records

  • Constructed demand metrics and analyzed order behavior across product categories, regions, and time

  • Interpreted demand behavior using segmented time series analysis rather than relying on aggregate averages

  • Generated short term demand signals that could support operational planning and projections

Demand Analysis

Daily demand patterns vary significantly across categories and are driven more by short, category specific pressure windows than by a single broad seasonal peak. One extreme outlier stands out across all categories on November 24, 2017, which corresponds to Black Friday and should be treated as an exceptional event rather than a recurring seasonal pattern. Outside of this date, demand behavior is more fragmented and category dependent.

Weekly seasonality is consistent across categories, with lower average order volumes on Saturdays and Sundays. This makes weekday demand more predictable and operationally relevant for planning purposes. Month level seasonality is more nuanced than expected. Instead of a generalized end of year surge, September through December tends to be softer overall, with November acting as a relative local high within that weaker period. Garden Tools is a notable exception, showing a clear November peak that suggests category specific seasonal drivers.​

Interactive Dashboard
dashboard preview olist.png

While the category level analysis provides depth, the interactive dashboard complements the findings by enabling dynamic exploration across regions, product categories, and time periods.

The dashboard allows stakeholders to navigate demand patterns spatially and temporally, supporting exploratory analysis without relying on static aggregate charts.

Key Findings

  • Demand behavior varies significantly across product categories and regions, reinforcing the importance of segmented analysis

  • A limited number of categories account for a disproportionate share of order volume

  • Category level demand exhibits volatility and seasonality that would be masked in aggregate views

  • Segmented demand signals provide more actionable guidance for short term operational planning

Tools & Technologies

• Python (Pandas, NumPy, Matplotlib, Seaborn)
• Data Visualization: Tableau

Full Project Repository

The complete project repository includes the full data preparation pipeline, category level demand analysis, projection logic, and supporting documentation.

All analytical steps, assumptions, and intermediate outputs are available in the GitHub repository to ensure transparency and reproducibility.

bottom of page