top of page

Olist Customer Insights Dashboard (AI-Enhanced)

1. The Mission (Project Context)

In a massive economy like Brazil, logistics isn't just a back-office function—it is the core of the customer experience. The mission of this project was to deconstruct the journey of an order from a seller's warehouse to a customer's doorstep. By synthesizing millions of data points, I aimed to identify the 'Logistics Tax'—the hidden cost of delivery delays—and provide Olist with a roadmap to improve regional retention and seller performance.

 

2. The Interactive Stack (The Visuals)

  • Screen 1: AI-Driven Delivery & Customer Experience Dashboard

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​​

​

​

​

​

​

  • Screen 2: Delivery & Fulfillment Performance Dashboard

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​​

​

​

​

​

​

​

​

​

​

  • Screen 3: Customer Satisfaction Overview

 

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​​

​

​

​

​

​

​

​

3. Technical Architecture

  • Data Modeling: Constructed a robust relational schema in Tableau, joining 8 CSV files. I utilized the Orders table as the central fact, creating one-to-many relationships with Order_Items, Payments, and Reviews.

  • LOD Expressions: Used Level of Detail (LOD) calculations to handle "double-counting" issues. For example, {FIXED [Order Id] : MIN([Payment Value])} ensured revenue remained accurate even when an order contained multiple items.

  • Geospatial Join: Integrated a custom Geolocation dataset, linking over 19,000 unique zip-code prefixes to calculate the physical distance between sellers and customers.

 

4. Key Business Insights (“The Discovery”)

  • The 14-Day Cliff: There is a critical psychological threshold for Brazilian consumers. Once delivery exceeds 14 days, the likelihood of a 1-star review increases by 400%, regardless of product quality.

  • The Power Categories: Health & Beauty and Watches & Gifts are the primary revenue engines ($2.4M combined). These categories also show the highest "Repeat Purchase" potential if delivered within 7 days.

  • The "Boleto" Delay: Using bank slips (Boleto) for payment adds an average of 48 hours to the "Order Approval" phase compared to Credit Cards, which ripples through the entire logistics chain and lowers final satisfaction scores.

 

5. Strategic Recommendations

  • Regional Hub Strategy: Olist should prioritize onboarding sellers in the North and Northeast regions. Currently, 70% of sellers are in the Southeast, forcing cross-country shipments that are the primary cause of the 7.9% late delivery rate.

  • Incentivize Credit over Boleto: Implement a "Fast-Track" approval badge for Credit Card purchases to encourage customers to move away from Boleto, effectively shaving two days off the total fulfillment time.

  • Dynamic ETA Logic: Update the "Estimated Delivery Date" algorithm to be more conservative for long-distance routes. The data shows that meeting a "Long" estimate results in a 4.2 rating, while missing a "Short" estimate results in a 1.8 rating.

​

6. Let’s Talk About the Data

The technical highlight of this project was the sheer scale of the Geolocation data. With over 1 million rows of coordinates, the initial dashboard was unusable due to lag. I solved this by performing Data Aggregation at the Zip-Code Prefix level. By clustering a million points into 19,000 localized hubs, I preserved the geographic story while increasing the dashboard's rendering speed by 85%. This ensured that an executive could filter from a national level down to a specific city without any performance 'spinning wheels'.

​

​

Olist Customer Insights Dashboard (AI-Enhanced) Build Guide

A. Data Model

  • Start with olist_orders_dataset.csv

  • Relate to reviews, payments, items, customers, sellers, and geolocation

  • Ensure relationships use Order ID, Customer ID, Seller ID, and Geolocation prefix

​

B. Core Calculated Fields

  • Avg Delivery Delay

  • Avg Review Score

  • Positive Review %

  • Total Reviews

  • Seller Order Count 

  • Dummy (1) for navigation

​

C. Visualizations Built

  • KPI cards

  • Map with seller coordinates

  • Scatterplot for delivery value/time

  • Category bar charts

  • Line charts for review trends

  • Stacked bars for score distribution

  • Navigation buttons using floating sheets

​

D. Dashboard Layout

  • Two with 1300 x 1500 layout 

  • One with 1300 x 1700 layout

  • Dedicated button zone

  • Sections for KPIs, main visuals, and insights

  • Consistent theme (blue/gray neutral)

​

E. Publishing

  • All dashboards published as a single workbook

  • Embedded into the website using Tableau Public's iframe code

  • Buttons automatically navigate inside the same workbook

​

​

​

​

​

bottom of page