Agoda (Booking Holdings) · UX research
The $5MM question nobody was asking.
Context
A routine analytics review at Agoda surfaced an anomaly worth chasing: a large, recurring volume of allotment rejects — where a hotel turns a guest away at check-in even though the room was booked and sold — concentrated in short-lead-time bookings across a handful of markets. The cost to the business was roughly $5MM a year, and it landed at the worst possible moment: the guest standing at the front desk.
The challenge
Every reject eroded customer loyalty at the point of highest friction, strained hotel-partner relationships, and burned support time — on top of the direct financial hit. But nobody could explain it. Most markets showed a steady baseline; one outlier market spiked with no data-level explanation, correlated only with short lead times. The mandate: find the real root cause, and validate a fix with hotel staff in the field before spending a single sprint building one.
What I did
I started in the data — three years of allotment-reject history — to separate signal from noise. The pattern was structural, not random: two of the three outlier markets were explained by partner-company bookings and ruled out; the third, highest-volume market had no data explanation at all, only a strong correlation with short lead times. That pointed away from a technical bug and toward something operational, happening at the hotel late in the process. I aligned the VP of Product, the VP of Partner Services, the Product Owner and Engineering on a contextual-inquiry approach — going to see it in person.
In the field, at the actual hotels where the rejects were happening, the cause became obvious in a way no dashboard could show: the OTA systems — Agoda's included — were only accessible to Sales and Reservations teams during business hours. The front-desk staff who handled late check-ins had their access restricted by hotel policy. So a short-lead-time booking arrived, the room was sold, but the person receiving the guest simply couldn't see it. With the root cause found, the field team ran rapid on-site design sprints and tested fixes with hotel staff in their own environment — landing on a validated solution polished to development-ready fidelity.
How I led
The leverage wasn't the research; it was the decision to go into the field at all. I built the business case on the $5MM figure to win VP alignment and a backlog slot, and framed the trip as risk management: flying a small team to the outlier market was a rounding error against the savings a validated fix would protect. Research only pays when it changes what the business does next — here it redirected engineering onto a problem that had been invisible from Bangkok.
Outcome
Root cause found, solution validated with the people who'd actually use it, and prioritised into the engineering backlog with a documented business case — $5MM in annual cost protected. All of it traced to one choice: go and watch, rather than guess from the data.
What it proves
This isn't a "we did research and found insights" story — it's research as a direct instrument of commercial value. The judgment that going to see it in context beats analysis from a distance is exactly what turned an unexplained data anomaly into a $5MM finding and a shipped fix.