Noah Joseph Manion

• 100% — Self-hosted Experimentation Stack Built & Shipped
• 15+ — Experiments in Active Backlog & Rotation
• $0 — Ongoing 3rd-party Testing Tool Cost

Google Optimize was dead, and the hosted alternatives were expensive, heavy, or sent experiment data through 3rd parties. Smaller clients needed a way to run rigorous tests without surrendering data ownership or paying enterprise pricing for features they would not use.

I architected and shipped Cadence, a self-hosted A/B testing platform, designing the stack and data model from scratch and driving the build through AI-assisted development. I shipped a visual editor, code injection, redirect testing, natural-language test creation, and an impact-attribution view that ties experiments to downstream outcomes.

Cadence runs live experiments for a paid field-service SaaS client and is dogfooded internally, with a backlog of 15+ tests. It gives clients full data ownership and removes recurring testing tool costs.

• 75% — Increase in Signup Conversion Rate
• 50% — Reduction in Churn Within 6 Months
• 30% — Reduction in CPA at Double the Volume

A seed-stage AI meeting tool was flying blind. Event tracking was inconsistent across the back end and 2 Angular front ends, the warehouse was full of unreliable data, and there was no working attribution. Marketing could not tell which channels or behaviors actually drove signups and retention.

I instrumented full-funnel event tracking with Segment across the back end and both front ends with merged pull requests, cleaned up the BigQuery warehouse, and built reverse-ETL syncs to ActiveCampaign and Attio. I layered in Clay and Apollo enrichment, rebuilt the attribution stack, then ran a structured A/B testing program on landing pages and onboarding, plus a behavioral-targeting paid rebuild.

Trustworthy data unlocked behavioral segmentation and real measurement. Signup conversion rose 75%, churn dropped 50% within 6 months, and CPA fell 30% while conversion volume doubled.

• 50% — CTR Lift from AI-assisted Creative
• 1 — Predictive LTV Model Deployed to Production
• 100% — Server-side Conversion Coverage via Meta CAPI

A DTC fundraising business needed sharper acquisition targeting and accurate conversion measurement as browser-side signal loss eroded ad performance. They lacked a forward-looking view of customer value and were leaking conversion data to the ad platforms.

I built a LightGBM predictive LTV and activation model to score and prioritize high-value cohorts, and implemented server-side Meta CAPI to restore conversion signal and improve match quality. I paired the modeling with AI-assisted creative testing.

Predictive scoring focused spend on cohorts most likely to convert and retain, server-side measurement recovered lost conversion signal, and AI-assisted creative drove a documented 50% CTR lift versus non-AI creatives.