Belli Data Analytics Cut Airline Booking Times by 47%
A look into how Belli leverages data analytics and targeted UI enhancements to eliminate workflow friction and reduce median booking times to 1.6 minutes.
Here's a look into how we use data and analytics to generate more revenue for our airline partners. Following the complete rollout of our enhanced booking platform, a recent performance analysis revealed a 47% like-for-like improvement in median booking times, dropping from an August baseline of 4.5 minutes to just 1.6 minutes. By refining our event tracking and implementing targeted UI enhancements,such as earlier error detection, local autosave, and eliminating dead-ends in duplicate flows,our team successfully removed hidden workflow friction. These optimizations save agents up to 20 minutes daily, ensuring our partners save valuable time and drive higher revenue.
To achieve this acceleration, our engineering team resolved several operational hurdles throughout August. We integrated a wallet configuration dialog directly into the duplicate wizard, preventing dead-ends for agents with unconfigured wallets. Validation errors like currency mismatches now surface earlier at step 2, saving agents from completing forms that would ultimately fail. We also introduced local autosave, allowing agents to recover their exact place if they accidentally refresh or navigate away. Concurrently, we rolled out advanced route and session-level tracking, establishing the data foundation necessary for true like-for-like performance comparisons.
In September, optimizations focused on system reliability and unblocking critical flows. We made submission requests idempotent, allowing agents to confidently retry after network errors without risking duplicate bookings. The team also strengthened origin and destination validation and resolved an unhandled edge case in the Special Handling Code (SHC) field. For multi-leg interline bookings, we corrected leg handling logic that had previously stalled the review step, fully restoring the end-to-end flow. Finally, we eliminated remaining wizard states that trapped agents mid-flow with no valid next action, ensuring smooth navigation and allowing for continuous drop-off monitoring moving forward.