Refund Protect & AI Dynamic Pricing

I audited how Refund Protect was actually performing across partner sites — UX placement, conversion patterns, data completeness — to find where value was leaking. One finding stood out: pricing was static. The same price applied regardless of demand, seasonality, route, or event type, which meant we were leaving revenue on the table every time demand spiked and pricing didn't move with it. I dug into the transactional data to quantify the gap — modelling what revenue would look like if price flexed with real-time criteria instead of sitting fixed — and used that analysis to build the case for a dynamic pricing initiative.

With the opportunity quantified, I scoped what an AI dynamic pricing model would need: training on large volumes of historical transactional data to learn the criteria that actually move conversion and revenue — demand, seasonality, route/event type, and more. I worked through how the model should weigh those criteria, and scoped an internal UI that let the team tweak and tune the model's parameters directly, rather than pricing logic being a black box only engineers could touch.

I drove the AI Dynamic Pricing initiative end-to-end — from the trained model through to the partner-facing product — paired with a UX placement guide and a repeatable conversion audit process, so implementation quality stopped depending on which partner happened to build the integration. I also redesigned the refund application journey itself, cutting friction out of the form users hit when something actually went wrong.

Every change — pricing, placement, the new refund form — went out against real partner traffic, not a lab test. The refund form redesign shows up directly in the numbers: fewer people abandoning mid-form.

I built the reporting layer needed to keep improving: the Pulse Reporting Dashboard turns partner data into a live feedback loop for pricing, placement, and commercial conversations — instead of one-off bespoke reports built on demand.