Long-term ROAS forecasting from early signals
Projecting future cohort value with limited maturity, accounting for bias, regularization, and uncertainty intervals to support early investment decisions.
Selected experience
Representative work on acquisition and growth decisions.
Examples of how the practice turns incomplete signals and real-world constraints into clearer acquisition and growth decisions.
Automated budget and ROAS target optimization
An online system that adjusts campaign-level budgets and ROAS targets using expected returns, saturation, operating minimums, scale limits, and commercial constraints. It can run continuously without human intervention.
Forecast bias controls and regularization
Designing controls that prevent volatile signals or small cohorts from producing extreme recommendations, with error monitoring over time.
Measurement under mobile privacy constraints
Combining aggregated signals, SKAdNetwork, cohorts, and experimentation to sustain acquisition decisions when individual attribution becomes unreliable.
Automated recommendations and reporting
Pipelines that turn performance data into reviewable reports, alerts, and recommendations, reducing repetitive work while keeping assumptions visible.
Marketing and forecast anomaly detection
Monitoring unexpected shifts in spend, revenue, tracking, and predictive error to separate real opportunities from data or execution issues.
Working principles
What stays consistent.
Decision first
The commercial question determines the granularity, horizon, and accuracy that actually matter.
Visible uncertainty
Assumptions and ranges sit beside the recommendation, avoiding a false sense of precision.
Real transfer
The result should be understood, challenged, and maintained by the teams that will use it.
A focused first conversation
Start with the decision creating the most uncertainty.
In an introductory conversation, we can frame the problem, review the available evidence, and decide whether there is a useful next step.
Discuss your challenge