Nine years running operations, quality assurance and growth in high-volume international environments — now doing the same work with Python and SQL underneath it.
Most of my career has been about spotting where a process leaks: recurring error patterns in QA audits, workflow friction across remote teams, conversion drop-offs in a funnel. Analytics is the same instinct with better evidence. What I care about is the point where an analysis stops being a chart and becomes a decision someone can act on.
Coming from operations inside a large digital platform, the questions I find most interesting are the ones marketplaces and products actually run on — why customers churn, where an operation quietly breaks, and which lever is worth funding first.
Currently: Data Analytics at Ironhack · Operations & QA at TP, on Trust & Safety for a global social media platform · based in Lisbon, moving to Madrid
Looking for: Business Analytics · Product & Marketplace Analytics · Business Intelligence — open to remote, hybrid or on-site
Where should a marketplace with a limited budget actually spend? An analysis of 99,441 orders from a Brazilian e-commerce platform, testing three competing answers against the data.
The expected answer was seller acquisition. It didn't survive testing — acquisition channel showed no significant relationship with seller performance (ANOVA p = 0.59, confirmed non-parametrically at p = 0.38). What did hold up was delivery: 37.8% of one-star reviews involve a late delivery, against 3.0% of five-star reviews.
The recommendation changed once I stopped ranking states by their late-delivery rate and started ranking them by their share of the total problem. That moved the answer from a small state with a bad rate to the one state responsible for a fifth of every late delivery in the marketplace.
Python MySQL Tableau — hypothesis testing with assumption checks, SQL cross-validation of every core metric, a seven-sheet interactive dashboard, and a written account of what the analysis does not prove.
→ Read the findings · → Browse the repository
