Agoda (Booking Holdings) · Consumer product · Experimentation
The booking form that never stopped improving.
Context
At Agoda, every B2C product is measured on one metric — Incremental Bookings Per Day (IBPD) — and every change is validated by A/B test at scale. In that culture, design decisions only count when an experiment says so, which means research has to be tightly wired to experiment hypotheses, not run as insight in isolation.
The problem
The booking form had been designed and led by a product owner with no UX background. Over time that produced a measurable conversion drag: a booking summary fragmented across columns, form fields buried below the fold, up to 35 unorganised payment options, and near-identical logged-in and logged-out experiences despite very different user needs. The mandate was to define a new direction the form could evolve toward, through research-informed experiments, to drive IBPD.
What I did
Before generating solutions, I ran four research streams in parallel — competitive benchmarking of leading OTAs, an analytics review of drop-off points, an audit of prior A/B tests, and user interviews plus usability testing across segments. A card-sort exercise gave us a ranked, evidence-based hierarchy of what users actually prioritise at checkout. Time-boxed workshops turned findings into hypotheses, paper prototypes were guerrilla-tested around the office before any build investment, and I designed dual-path experiment variants — one to run if the previous test won, one if it lost — so the programme kept moving regardless of outcome.
Outcome
The sequenced A/B programme delivered four wins — consolidating the split booking summary to push fields up the page, adding a property thumbnail for booking confidence, redesigning saved-card display for returning users, and categorising the 35 payment options. One variant — splitting the form into cards — lost, because it made the CTA less prominent. The loss was as useful as the wins.
What it proves
The experiments that had won in the past were exactly what sparked this project — because what wins at one point in a product's life can hurt conversion later. In a data-driven culture, research is never finished: it generates hypotheses, experiments validate them at scale, and results feed the next question. This is that closed loop, run deliberately.