Predictive Travel Search
Concur Technologies
My team and I worked with an external vendor on some strategy projects to help guide the vision of our products in the next 3-5 years. Of the many concepts we explored, the one that generated the most interest from our leadership team was the idea of full itineraries (air + hotel + car) presented within search results.
The motivation driving this exploration came from our data sciences team, who could predict with 30% accuracy what even a new Concur user will book for their business travel based on information in a very basic profile. We wanted to test whether or not a 30% accuracy would be enough to justify putting an entire itinerary recommendation in front of a user, or if it would be too off-base for them to find it useful.
The questions we aimed to answer through our usability sessions:
- Do users understand what a suggested trip is?
- Does showing the suggested trips accelerate or disrupt the user experience?
- Are suggested trips a quicker, simpler way to book full itineraries?
Testing uncovered that presenting three full itineraries did not resonate for users. We modified the design and instead presented only one full itinerary within the regular linear search results. With that change, comments from users were extremely positive:
"It's saving me the effort of having to pick all of that...Wow, that is a lot quicker than the way I'm doing it!"
"If it's assuming my preferences, then this looks good."
One of our other areas of focus was the way the screen animations enhanced the user experience: did they clarify, or confuse? Did the UI anticipate the user's next move, and guide them through the necessary steps? Our efforts were positively received by our users:
"Wow, it's just popping through this stuff automatically...it was easier to use. It seemed to anticipate my next move, and required less tapping."
Overall, participants in the study found the idea of a "suggested trip" to be useful, usable, and desirable. When they noticed the suggested itinerary, they found it to be very useful. They easily understood their preferences and company policy was being accounted for in what was presented. They also were more likely to prefer the suggested itinerary path to the idea of booking each piece of their trip separately. Having the complete itinerary displayed did not detract from the usability of booking flights and hotels separately for those user who preferred a more traditional approach.
The only downfall we found was the reliability and accuracy of our suggested itineraries. They worked best when, obviously, the predictions were relevant to the users. These explorations have been guiding further work on our data sciences team to make our recommendations solid so that we can start presenting itineraries this way in our UI. In the meantime, other concepts that became cornerstone assumptions of this project are being divided out into other project teams and implemented so that they are already in place when the team is ready to launch the idea of full itineraries as search results.