Portfolio & Specialisations
Deep experience across measurement, lifecycle, predictive AI, and agentic workflow automation.
Google, Meta and Amazon Attribution Rebuild
Google was overcredited. Meta was underfunded. Amazon was invisible in reporting. I rebuilt the full attribution model from last-click to Markov chain across all three channels, redirecting spend toward the channels actually generating revenue. A/B testing on creative variants compounded the revenue impact of the reallocation.
Geo-Holdout Revealed Google ROAS Overstatement
Google Display was reporting strong ROAS. The true incremental return was far lower. I designed a 16-DMA geo-holdout to isolate actual causal lift from organic demand, then built a Bayesian MMM in PyMC to model how reallocation would affect total revenue. Shifting budget toward high-iROAS channels projected meaningful revenue uplift without adding a single dollar of total spend.
LTV-Led Meta Targeting Driving Online Revenue Growth
Meta prospecting was spending broadly on low-value audiences. RFM segmentation revealed a stark LTV gap between customer tiers. I shifted paid Meta budget toward Champions lookalikes, redesigned Klaviyo nurture flows for high-value cohorts, and realigned email send-timing based on cohort repurchase windows — driving revenue growth from smarter targeting, not higher spend.
Unified Revenue View Across Google, Meta and Email
Built a unified paid media performance system consolidating Google Ads, Meta Ads, and Klaviyo into a single blended revenue view. Identified that Email and Klaviyo significantly outperformed paid social in revenue efficiency, and that retargeting outperformed prospecting across both Google and Meta. Budget was redirected toward highest-revenue touchpoints accordingly.
AI Marketing Insights Copilot
Weekly reporting across Google, Meta, and Amazon was consuming 4.5 hours of analyst time before insights reached leadership. I built an agentic AI copilot that ingests raw channel performance data, applies structured reasoning, and delivers decision-ready revenue narratives automatically. Insights that took half a workday now arrive in 45 seconds. Leadership gets faster answers, analysts get time back for higher-order measurement work.
Executive Narrative Generator
An agentic system that takes raw Google Ads, Meta, and lifecycle performance data and produces stakeholder-ready executive summaries automatically. Eliminates the manual synthesis step between data and decision, so revenue insights reach leadership without delay.
Budget Pacing Monitor Agent
Automated budget pacing surveillance across Google, Meta, and Amazon. Real-time alerts flag overpace and underpace conditions with reallocation recommendations before revenue impact accumulates. Prevents wasted spend from compounding across a campaign flight.
Brand Sentiment Tracker
Sentiment monitoring pipeline that classifies brand mentions across social channels and surfaces volume trends and competitive share-of-voice in a live dashboard. Connects brand health signals to paid media context so teams understand how sentiment shifts correlate with channel spend and revenue performance.
Churn Prediction Protecting Subscription Revenue
Subscriber churn was eroding recurring revenue. I built a churn prediction model identifying missed payments as the single strongest signal, 5x more predictive than any demographic attribute. Model scores were translated into ranked retention lists and triggered intervention flows. The result was an 18% reduction in churn, directly protecting the recurring revenue base that paid media spend was being used to acquire.
RFM Segmentation Connecting Paid Media to Revenue Quality
Meta and Instagram spend was acquiring customers with a low average LTV while Champion-tier customers were generating significantly more. I used RFM segmentation to identify where paid media budget should actually flow, shifted Meta ad targeting toward Champions lookalikes, and redesigned Klaviyo nurture flows for high-value cohorts. Online revenue and repeat purchase rate both improved without increasing total ad spend.
Telco Lifecycle System Protecting Annual Recurring Revenue
Built an end-to-end customer lifecycle system on IBM Telco data connecting churn prediction, LTV modeling, and segmentation into five revenue-protecting flows. Month-to-month customers churned at dramatically higher rates than long-tenure subscribers. Model-triggered flows covering nearly all of the customer base were designed to move high-risk segments toward higher-tenure contracts and protect recurring revenue.