From Data to Decisions: How Terno AI Analyzes Olist’s E-Commerce Supply Chain

Why E-Commerce Data Holds Hidden Gold

Imagine walking into the busiest shopping mall in Brazil. Thousands of people are coming and going, each browsing, purchasing, and checking out in different ways. Some pay in cash, some with cards, others with installment plans unique to the region. Now imagine trying to track every single transaction, delivery, and return in real time. That’s what the Brazilian e-commerce giant Olist represents—a bustling digital marketplace where millions of interactions happen across the country.

This blog is built on a real conversation I had with Terno AI, where I uploaded the Olist e-commerce dataset and asked step-by-step questions. Every chart, table, and insight you see here comes directly from that interactive analysis.

Introduction

Brazil is one of the fastest-growing e-commerce markets in Latin America. But with opportunity comes complexity. Deliveries often span thousands of kilometers—from São Paulo’s bustling urban centers to the remote Amazonian state of Roraima. Payments are just as diverse: while credit cards dominate, millions still rely on boleto bancário (bank slips). And like every marketplace, delays, cancellations, and refunds remain part of the story.

For business leaders and analysts, the challenge is clear: how do you turn this ocean of transactions into meaningful insights? Which patterns actually matter—faster deliveries, smoother payment experiences, or fewer cancellations?

That’s where Terno AI comes in. Think of it as a smart co-pilot for e-commerce data. Instead of spending weeks cleaning spreadsheets or writing SQL queries, you simply ask: “Which payment method generates the most revenue?” or “Where are deliveries consistently late?” Terno AI responds with clear visual answers in seconds.

In this blog, we’ll walk through the Olist dataset—a real Brazilian e-commerce dataset—to uncover stories hiding in the data. And more importantly, we’ll show how Terno AI transforms raw CSVs into business-ready insights for decision-makers in FMCG and retail.

By the end, you’ll know:

• Which payment methods dominate Brazil’s online shopping scene
• How delivery times vary dramatically by state
• Why cancellations happen—and which categories suffer the most
• How companies can act on these insights to improve customer experience and profitability

Let’s dive in.

The Terno Al Chat: Full, unedited conversation: (Click here to see every prompt, answer, and visualization as it happened.)

Use Cases Demonstrated with Terno AI

• Dataset Exploration & Data Quality Check — profile tables, data types, missing values, and numeric summaries to validate readiness.
• Entity-Relationship Mapping (ERD) — clarify how customers, orders, items, payments, products, and sellers connect.
• Payment Insights — analyze payment-type share, installment behavior, and revenue mix across methods.
• Delivery & Logistics Performance — measure delivery time, delays vs. estimates, and regional differences.
• Order Status Monitoring — track the distribution of delivered, shipped, canceled, and other statuses over time.
• Cancellations & Risk Management — identify categories and months with higher cancellation rates to prioritize fixes.

Product Overview

Before jumping into charts and insights, let’s first understand the “playground” we’re analyzing. The Olist dataset is one of the most widely studied e-commerce datasets from Brazil. It’s like a snapshot of how a digital marketplace operates, with tables covering everything from orders and payments to customer reviews, sellers, and products.

Think of it as an x-ray of the entire shopping journey:

• Customers – who they are and where they live (from São Paulo’s urban centers to remote Amazonian towns).
• Orders – what they purchased, when, and how the process unfolded.
• Payments – which methods were used (credit cards, “boleto,” vouchers, etc.).
• Products – details such as category, weight, size, and even translated names.
• Sellers – the businesses fulfilling these orders.
• Geolocation – zip codes tied to coordinates, allowing for delivery performance mapping.
• Reviews – customer satisfaction captured in ratings and comments.

In total:

• ~100k orders
• ~1M geo-coordinates
• ~112k items
• 30k+ products

Think of it as the black box of a marketplace—once opened, it tells the story of Brazilian e-commerce at scale.

Terno Insight:

Dataset Overview & Why It Matters

Before asking big questions like “Which payment method brings the most revenue?” or “Where are deliveries slow?”, we first need to know what kind of data we’re working with. Think of it as checking your ingredients before starting a recipe.

The Olist marketplace dataset gives us a rich snapshot of Brazil’s e-commerce activity:

• Orders – when purchases happened, their status, and delivery timelines.
• Payments – how people paid (credit card, boleto, voucher, debit).
• Order Items – what was bought, price, shipping costs, and which seller handled it.
• Products – categories, descriptions, photos, size and weight info.
• Customers & Sellers – where buyers and sellers are located across Brazil.
• Geolocation – latitude/longitude tied to postal codes to map delivery flows.
• Reviews – customer feedback scores and comments.

What we checked first:

• Completeness: Key fields like order IDs and payment info are intact (no missing IDs).
• Consistency: Timelines (purchase, ship, delivery) line up well for delivery analysis.

Connecting the Dots in the Marketplace

Behind every order in Olist, there’s a story. Each order starts with a customer, links to one or more products, comes from a seller, and is tied to a payment. Think of it like threads in a web—when you pull one (say, payments), the others move too (orders, customers, deliveries). This interconnected structure is what makes the dataset so powerful, because it lets us explore not just isolated numbers, but the relationships that drive real business outcomes.

How Do Brazilians Pay Online?

When you step into Brazil’s digital marketplace, one thing becomes immediately clear: credit cards dominate. Nearly three out of every four purchases are made with them. But unlike many countries, there’s a twist — Brazilians often split payments into installments. Instead of paying $400 upfront, a shopper might pay $40 over 10 months. It’s not just convenience; it’s cultural.

The second most popular method is the boleto, a bank slip that customers can print and pay at ATMs or convenience stores. This old-school method still makes up nearly 20% of online orders, showing how diverse payment habits really are. Debit cards and vouchers fill in the rest, each with their own niche role.

For businesses, this insight is gold. If you ignore installment options or boleto payments, you risk shutting out entire segments of customers. But by supporting these methods, platforms like Olist can unlock more sales and stronger trust with buyers across the country.

Credit cards dominate Brazil’s e-commerce payments (nearly 74%), while traditional “boleto” slips still account for a significant share at 19%. Vouchers and debit cards make up the rest.

Installment Preferences (Credit Card)

Here’s where Brazil gets unique. Many shoppers don’t just use credit cards—they split payments into installments.

• One-shot payments (33%) are most common.
• 2–4 installments (~40%) make up a big share of transactions.
• Extreme plans (10–24 installments) still exist, though rare.

Imagine buying a new iPhone. In the U.S., you’d likely pay all at once or maybe finance through Apple. In Brazil, it’s perfectly normal to split the cost into 10 monthly payments — not just for phones, but even for everyday items like furniture or electronics.

Installments are a cultural norm in Brazil — from 1–4× everyday payments to rare but real 10–24× splits.

Revenue Mix by Payment Type

When we switch from how people pay to which method brings in the most money, the picture gets even clearer.

• Credit cards dominate — almost 80% of total revenue comes from them. They’re the king of checkout, fueling Brazil’s online retail growth.
• Boleto (bank slips) still plays a serious role, contributing nearly 18% of revenue — showing how trust in traditional payment methods remains strong.
• Vouchers and debit cards together account for just a few percent, more like supporting actors than main players.

Think of it like a football match:

• Credit cards are the star striker, scoring most of the goals.
• Boleto is the steady midfielder, still critical for balance.
• Vouchers and debit cards? They’re the substitutes — useful, but not game-changers.

For businesses, this isn’t just trivia. Knowing which payment channels drive revenue helps prioritize checkout design, partnerships with banks, and fraud prevention strategies.

Revenue Mix by Payment Type: Credit cards aren’t just the most used payment method—they drive nearly 80% of total revenue in Brazil’s e-commerce, with boleto far behind and vouchers/debit cards playing only a minor role.

Once we understood which payment methods bring in the most revenue, the next logical question was: does this trend hold true across Brazil’s diverse regions?

This is where Terno AI really shines. Instead of manually joining tables or running SQL queries, I simply asked: “Show me the payment mix by state for delivered orders.” Within seconds, Terno AI generated both a table and a stacked bar chart—just like the screenshots you’ve seen above.

These visuals not only save hours of data wrangling but also tell a story at a glance: while credit cards dominate everywhere, the share of boleto payments rises significantly in certain states, reflecting cultural and regional preferences.

The screenshot below is directly generated by Terno AI, showcasing the state-level breakdown of payment preferences.Figure:Payment Mix by State (Top 10) — Credit card leads in every state (~72–78%). boleto rises notably in RS (~24%), while voucher usage is highest in BA (~8%). (Screenshot generated with Terno AI.)

Once we knew which payment methods bring in the most revenue overall, the next question was simple: does this hold true everywhere in Brazil? Using Terno AI, I asked for the payment mix by state (delivered orders only). In seconds, it produced the table and stacked bar chart below—no manual joins, no SQL.

What the regions tell us

• Credit cards lead everywhere — typically ~72–78% across all top states (SP, RJ, MG, PR, SC, RS…).
• Boleto rises in the South — especially Rio Grande do Sul (~24%), suggesting stronger bank-slip preferences there.
• Voucher usage stands out in Bahia — roughly ~8%, higher than most states.
• Debit cards remain tiny — around ~1% almost everywhere.

What I’d do next

• Localize checkout defaults: Keep credit card first everywhere; surface boleto more prominently in RS/SC/PR; highlight vouchers a bit more in BA.
• Targeted promos: Run boleto payment fee waivers or cashback in the South; test voucher-based bundles in Bahia.
• Smart risk rules: Keep fraud checks tighter on long installment plans, but don’t penalize one-shot payments.

All visuals and percentages in this section were generated directly inside Terno AI from the Olist dataset—no spreadsheets, just questions and answers.

Next up: we’ll step away from payments and look at the journey after checkout—how fast orders arrive and where customers wait the longest.

E-Commerce Delivery Times in Brazil: Regional Insights

After payments, the next question every customer asks is: “How fast will my order arrive?” Delivery speed is the heartbeat of e-commerce—and in Brazil, it’s especially tricky. Orders often travel from São Paulo or Rio de Janeiro across thousands of kilometers to reach remote states like Amazonas or Roraima.

Instead of manually calculating averages from millions of rows, I simply asked Terno AI: “Show me the average delivery time (in days) for delivered orders, broken down by state.” Within seconds, Terno AI returned both a clear table and an easy-to-read bar chart.

1. Delivery Speed by State

When we measured the average delivery days from purchase to arrival, the gap was eye-opening:

• São Paulo (SP): ~8 days — the fastest, thanks to being Brazil’s e-commerce hub with dense infrastructure.
• Minas Gerais (MG), Paraná (PR), Distrito Federal (DF): ~11–12 days — efficient, supported by strong logistics networks.
• Roraima (RR): ~29 days — nearly a full month, showing the logistical hurdles of serving remote northern states.

What the data shows

• Fast deliveries (3–5 days): São Paulo, Rio de Janeiro, Minas Gerais, Paraná—where logistics hubs are strongest.
• Medium delays (7–10 days): Northeastern states like Bahia and Pernambuco.
• Longest waits (15+ days): Amazonas and Roraima, reflecting distance and infrastructure challenges.

2. Delays vs. Promises

We then compared promised vs. actual delivery dates—and the results were telling:

• Remote North (RR, AP, AM): consistently late by 12–16 days.
• Southeast (RJ, SP, MG): often early, with deliveries arriving 7–10 days before the promised date.

Business insight: Overpromising hurts customer trust, but delivering earlier than promised delights customers. Retailers should tailor delivery estimates by region to strike the right balance.Figure: Comparing promised vs. actual delivery times by state. Generated with Terno AI.

3. Monthly Trends in Delivery Times

Zooming out over time, the story becomes even clearer:

• 2016: customers waited over 20 days on average.
• 2018: delivery times dropped below 9 days, a major improvement.
• Holiday spikes: November–December consistently saw longer waits due to the seasonal rush.

Figure: Delivery times improved sharply between 2016–2018. Screenshot generated with Terno AI.

Terno Insight:

Why Logistics Insights Matter

Delivery isn’t just a backend metric—it directly impacts business growth:

• Faster delivery = higher customer satisfaction.
• Regional gaps = expansion opportunities.
• Trend tracking = proof of operational improvement.

With Terno AI, business leaders don’t need SQL queries or endless spreadsheets. They can simply ask natural questions like: “Which states have the slowest deliveries?” or “How have delivery times improved in the last two years?” and get instant, visual answers.

Next up: not all orders arrive smoothly. Some get canceled, others refunded. Let’s see what the data reveals about cancellations and risks.

Order Outcomes & Cancellations: Reading Between the Status Lines

Every online order has a story. Some arrive safely at the customer’s doorstep, others get delayed, and a few never make it at all. By analyzing Olist’s order data with Terno AI, we uncovered what really happens after customers click “place order.”

Here’s the big picture:

• 60% of all orders are delivered successfully.
• 20% are still in “shipped” status.
• ~12% are canceled — meaning 1 in 8 customers doesn’t complete the journey.

Even small improvements in the pre-purchase experience could reduce cancellations and unlock growth.

Category-Level Cancellations: Which Products Are Riskier?

Not all product categories face the same risk of cancellation. With Terno AI, we found:

• Books (General Interest): ~1.4% cancellations – niche demand and supply mismatches.
• Small Appliances & Musical Instruments: ~1.3%.
• Toys & Consoles: <1% – more stable demand and fulfillment.

For retailers, this means high-risk categories deserve extra attention. The issue could be unreliable suppliers, stockouts, or customer second thoughts.

Terno Insight:

Monthly Trends: Are Cancellations Improving?

When we tracked cancellation rates over time, a clear story emerged:

• 2016 (early months): Unrealistic spikes of 50–100% — caused by very few recorded orders.
• 2017–2018: Rates stabilized below 1%, showing reliable fulfillment.
• Feb 2018: A small seasonal spike (~1.1%) likely linked to promotions or stockouts.

At the same time, delivery times improved from 20+ days in 2016 to under 9 days by mid-2018. This shows that operational efficiency was increasing while cancellations stayed low.

Why Cancellations Matter

• Hidden costs: refunds, re-stocking, lost sales.
• Trust at stake: repeated issues drive customers to competitors.
• AI-powered fixes: spot risks early (certain SKUs, suppliers, or months) and act fast.

With Terno AI, you don’t need a data science squad to figure this out. You can simply ask:

“Which categories have the highest cancellation rates?”
“Is our cancellation trend improving month by month?”

And instantly see tables and charts — just like the ones in this blog.

Key Insights Recap

• 60% of orders succeed — but 40% face delays or issues (shipped, canceled, or other).
• Niche categories like books or small appliances see more cancellations than toys or electronics.
• Cancellations stayed below 1% in recent years, a strong sign of reliability.

For decision-makers, the lesson is simple: measuring outcomes is the first step to improving them.

 Curious which of your products risk cancellation most? Let Terno AI uncover the patterns in seconds.

Conclusion: From Data to Decisions with Terno AI

This journey through Brazil’s Olist dataset shows how raw data becomes real business decisions when explored the right way:

• Payments reveal how Brazilians prefer to shop — from credit cards to multi-month installments.
• Deliveries uncover how geography shapes customer experience — fast in São Paulo, slower in remote states like Roraima.
• Cancellations highlight where trust can be lost — or earned — depending on category and timing.

The best part? You don’t need weeks of SQL queries, pivot tables, or manual joins to get here. With Terno AI, all it takes is natural-language questions like:

“Which payment type brings the most revenue?”
“Where are deliveries consistently late?”
“Which categories are most at risk of cancellation?”

In seconds, you get back the kind of visual insights and trends that usually take entire analyst teams to prepare.

Key Takeaways

• Credit cards dominate Brazilian e-commerce, but regional habits like “boleto” and installments still matter.
• Delivery times improved massively between 2016–2018, yet remote areas remain challenging.
• Cancellations remain low (<1%), but niche categories require closer monitoring.

For decision-makers, the message is clear: ask better questions, get faster answers, and turn data into action.

Ready to Unlock Insights in Your Own Data?

Using Terno AI, advanced analytics and predictive modeling become simple, fast, and actionable — even for non-technical leaders. Think of it as a shield for your database:

• Neutralizing risky SQL queries from Large Language Models (LLMs)
• Keeping your data secure while still delivering instant answers
• Turning CSV files into business-ready dashboards and forecasts in minutes

Book a Terno AI demo today and see how easily your business can move from raw data to real-world decisions.

Powered by Terno AI — Data made human, insights made simple.

Explore the Terno AI here:Terno AI Link

The Terno Al Chat: Full, unedited conversation: (Click here to see every prompt, answer, and visualization as it happened.)