Airline Optimizes Online Booking Performance

The challenge

The airline's primary booking channel is their website, beginning with the search function on the homepage and continuing on to each step in the booking process. Any issue preventing customers from completing a step could significantly impact revenue within minutes. Taking a reactive approach by waiting to see a drop in conversion rates or waiting to receive bad reviews from customers risks losing significant revenue. But a complex website with hundreds of thousands of visitors every month inevitably has errors, and prioritizing a response to every error took an enormous amount of resources for the company.

The solution

The eCommerice team brought in Glassbox to detect errors automatically, prioritize them by business impact and collaborate with IT on efficient resolutions.

Autonomous monitoring

Once Glassbox was deployed on the site, the team activated the built in machine learning to receive alerts on spiking trends in errors. The algorithm used historical baselines to dynamically define the normal expectation for each error occurrence, saving the team time of manually setting thresholds and continuously tweaking them. And because Glassbox catches user behavior (client-side) and behind-the-scenes (server-side) errors with no need for pre-tagging, they were able to continue to focus on rapidly rolling out changes to their website without having to worry it would disrupt their ability to monitor errors.

Actionable alerts

Their decision bore fruit when they received an alert triggered by a rise in errors above the normal range. Drilling down into the out-of-the-box error report, they traced it to a specific API error code. Replaying sessions with this error code revealed a problem preventing users from completing their search for flights.

There was no visual indicator to users to show why they were unable to proceed. This meant users who did contact support could not explain the issue. However, the eCommerce team was able to capture the error code in session replay and share it directly with IT. This dramatically reduced the time it would have taken them to diagnose the problem themselves.

Prioritized decision making

On top of that, the team immediately set up an ad hoc funnel to measure how many sessions were impacted.  The funnel revealed that 50% of sessions in which customers were trying to use the search were impacted. Armed with this analysis, they escalated the issue immediately and the problem was fixed within 24 hours.

Results

By swiftly detecting, diagnosing and prioritizing the error in their search function the eCommerce team was able to save hundreds of at risk bookings and maintain the loyalty of their customers.

Automated monitoring gives them confidence to continue to innovate and improve their site experience. And by only escalating high priority issues to engineering, they ensure future issues will receive a quick response.