Expedia | Travel booking platform
Using behavioral psychology to increase revenue by $4.7M.
After a shift in language almost cost us millions, this simple message based on social psychology and data analysis increased upsell by 3.12%.
Role
Content Designer
Team
UX Researcher and Product Manager
Background
Fixing a multi-million dollar mistake
In 2020, our design team changed how we displayed room options to provide more transparency around hotel room prices. Previously, hotel room choices were bundled with amenities such as parking, breakfast, and free cancellation.
This bundling is industry-standard, but it obscures the fact that users are paying extra for the included amenities. We, therefore, changed the design into an a-la-cart experience, allowing users to pick which amenities they’d like to add to their base rate for the room.
But we created a problem.
What was previously displayed as “free cancellation” was now a “fully refundable” amenity upsell. And we quickly saw a 2.24% decrease in refundable bookings.
I was brought in to solve this problem with content.
Business problem: Increased price transparency and changing the phrase “free cancellation” to “fully refundable” on our booking site accidentally resulted in a 2.24% decrease in refundable bookings.
Customer problem: As a traveler booking a hotel room, why would I pay extra for the ability to cancel when it’s free on other websites?
Design constraints
Solving with only content
Because our product designers were busy, I was asked to solve this problem using only components in our design system.
The component that best fit this use case was a small messaging card with a short header, one line of copy, and two CTAs.
User research
Testing psychology principles in the lab
In this small amount of copy, I needed to convey the value of paying more for the flexibility to cancel. Traditionally, Expedia used social proof and risk aversion to convey value, but I wanted to explore additional options.
Since I had access to our research lab at Expedia, I decided to test multiple copy strings based on different psychology principles.
In the research lab, users were asked to rank the following six copy strings based on helpfulness:
Generic – Stay flexible with a fully refundable option.
Numeracy bias – Not sure if plans will change? Add the fully refundable option for $80.
Perceived understanding – We know flexibility matters to you. Book a fully refundable option.
Authority – Plans can change. We recommend booking a fully refundable option.
Social proof – Most people have been booking fully refundable options.
Loss aversion – Don’t risk it. Book a fully refundable option so you can cancel if plans change.
Users preferred transparency, especially with numbers
The string based on numeracy bias, “Not sure if plans will change? Add the fully refundable option for $80,” was the clear winner from the research lab.
Quotes from user testing:
"This sounds more genuine. Like Expedia is looking to help instead of ordering me around. I'm more likely to listen."
"I like how the fee is clear so I can judge if it's worth it or not."
A/B Testing
Two header variants inspired by research winner
I learned two things from our user research—users liked numbers, and they didn’t like being told what to do.
I used this feedback to write two header variants we would A/B test on our site.
Heading 1: Not sure if plans will change?
Heading 2: Peace of mind, starting at $X
Making it personal with dynamic copy
The messaging card component had room for a short sentence underneath the header, and I wanted to use this space to explain why we were recommending a fully refundable room.
This project was completed in 2020, during the height of quarantine and travel cancellations. And our data showed that travelers were more likely to cancel their trips the further in the future they were scheduled.
And since we knew the travelers’ booking window, I could add dynamic copy, making the recommendation seem personalized.
Dynamic copy: We recommend booking a fully refundable room because your trip is X weeks out.
results
Both variants were winners (compared to control)
Variant A: 1.2% increase in upsell ($2.1M)
