SQL + Python · E-Commerce Analytics

Olist E-Commerce
Performance Analysis

9 source tables, 8 loaded into SQLite · 96,478 delivered orders · R$15.4M revenue across Brazil · 2016–2018

● 96,478 orders analyzed 27 Brazilian states Sep 2016 → Aug 2018 9 source tables · SQLite
Total Revenue
R$15.4M
across 96,478 orders
Avg Order Value
R$159.83
per delivered order
Avg Delivery Time
12.0 days
SP delivers in 8.3d
Top Category
9.2%
Health & Beauty share
Monthly revenue trend

Monthly Revenue — Full Time Series

Sep 2016 → Aug 2018 · annotated peak: Nov 2017 Black Friday spike (+53% MoM)
Revenue breakdown

Revenue by Category

Top 10 of 73 categories · underscores replaced with readable names

Top States by Revenue

SP alone = 37% of total · note delivery gap vs southern states
State Revenue Orders Avg del.
Operational insights

Delivery Speed vs Customer Rating

At order grain: fast (≤7d) average 4.41★, very slow (>21d) fall to 3.01★ — a 1.40-star gap
Fast (≤ 7 days)
4.41 ★
4.41
33,539
Normal (8–14 days)
4.29 ★
4.29
36,190
Slow (15–21 days)
4.10 ★
4.10
15,264
Very Slow (> 21 days)
3.01 ★
3.01
10,831
Score axis: 1.0 (worst) → 5.0 (best) · one row per ORDER (95,824 with a review), reviews averaged per order before joining. Adjusted for state, category and freight, each extra delivery day multiplies the odds of a ≤3-star review by 1.0854 (95% CI 1.083–1.088, n=95,824) — a one-week slip is 1.77×. Unadjusted it is 1.0770/day, so controlling for confounders makes the effect stronger, not weaker.

Payment Method Mix

Credit card dominates at 78.6% of revenue · boleto = Brazil's bank slip payment

Observed Pattern: Slower Deliveries Coincide With Lower Review Scores

São Paulo (SP) achieves an 8.3-day avg delivery and accounts for 37% of all revenue (R$5.77M). States with >15-day average delivery like Bahia (18.8d) and Espírito Santo (15.2d) sit well above the 12.0-day national average, and the review-score gap by delivery bucket below is the clearest lever in this dataset. This is an unadjusted comparison — delivery time is confounded with state, category mix and freight cost, so it is not yet a causal estimate.