Commercial outcomes across recent engagements
A selection of verified results — framed in revenue, cost, and risk terms.
Additional ad revenue generated by a sale-probability model identifying optimal inventory for third-party advertising — at 2.66× ROI, with a 4× increase in eligible inventory.
Real-time price forecasting model surfacing estimated sale prices at listing time, increasing completed listing rate and total gross merchandise value.
Bayesian experimentation across acquisition touchpoints drove a 15% increase in ad click rate and 5% subscriber growth, followed by an epsilon-greedy bandit removing manual A/B scheduling.
Purchase likelihood classifier triggering targeted CRM notifications. New user return rate doubled; early churn reduced by 5%.
Feature engineering improvements to an existing fraud detection classifier — a 23% lift in PR-AUC before model retraining — then deployed to production.
Retrained product embedding model for interaction-based recommendations, reducing inference runtime by 20% — directly lowering serving cost at scale.
Causal survey design demonstrating 17% incremental brand equity impact — a board-level measurement framework, not correlation analysis.
Demand prediction model across 1,000+ products enabling price point optimisation; extended to a reinforcement learning approach for continuous improvement.
Clients include
