Based on data analysis of 99,000+ orders from 2016–2018
Finding: Late deliveries have an average review score of 2.25 vs 4.21 for on-time deliveries — a statistically significant difference (p-value = 0.00).
Recommendations:
- Set stricter SLAs (Service Level Agreements) with sellers who have positive avg_delay
- Flag orders at risk of delay early using the ML predictor model
- Send proactive notifications to customers when delay is likely — reduces frustration even if delivery is late
Finding: This segment has the highest CLV (164.85 BRL) and highest avg spend (329.71 BRL) but hasn't purchased in ~395 days.
Recommendations:
- Send personalized win-back email campaigns with exclusive discounts
- Offer free shipping on next order — high spenders will respond
- Priority customer support for this segment
Finding: 73.77% of customers pay via credit card. Boleto users (19.45%) are cash-dependent.
Recommendations:
- Partner with major Brazilian credit card companies (Nubank, Itaú) for cashback offers — drives repeat purchases
- Offer 0% installment plans to convert Boleto users to credit card — increases average order value
- Debit card usage is only 1.44% — not worth investing here
Finding: SP has 48,816 orders (avg 152.76 BRL) vs RJ with 14,963 orders (avg 180.42 BRL). Difference is statistically significant (p-value = 0.00).
Recommendations:
- SP strategy — volume focused: bundle deals, flash sales, free shipping above threshold
- RJ strategy — premium focused: higher-value products, exclusive collections
- Low-order states (GO, DF, ES) — run awareness campaigns, currently untapped market
Finding: Top revenue seller earns 510,915 BRL but has only 3.40 avg review score. Top 4 sellers all have below 4.0 reviews.
Recommendations:
- Introduce a seller quality score combining revenue + review score
- Sellers below 3.5 review score should receive mandatory seller training
- Offer incentives to high-revenue sellers who improve their review scores
- Consider demoting chronic low-review sellers from featured listings
Finding: Frequency mean = 1.03, cohort retention drops to < 1% after Month 1. Olist is a one-time purchase platform.
Recommendations:
- Launch a loyalty/rewards program — points per purchase redeemable as discounts
- Personalized product recommendations via email after first purchase
- "Complete your collection" campaigns for categories like bed_bath_table, health_beauty
Finding: LDA topic modeling on negative reviews shows Product Quality is the #2 complaint topic (2,876 reviews) after Delivery Delay (3,248).
Recommendations:
- Implement mandatory product photo verification before listing
- Allow customers to flag "item not as described" — auto-flag sellers with high mismatch reports
- Introduce verified purchase badges for authentic reviews
| Priority | Issue | Impact | Action |
|---|---|---|---|
| 🔴 High | Delivery delays | Review score drops from 4.21 → 2.25 | Stricter seller SLAs |
| 🔴 High | Cant Lose Them churn | Highest CLV segment leaving | Win-back campaigns |
| 🟡 Medium | Low retention | Frequency = 1.03 | Loyalty program |
| 🟡 Medium | Product quality complaints | #2 negative topic | Seller quality score |
| 🟢 Low | State-wise targeting | RJ 18% higher AOV than SP | Segmented marketing |
| 🟢 Low | Credit card partnership | 74% already use credit card | Cashback program |