Saved by the Batphone
Helping agents provide rebooking support for last-minute cancelled guests
Introduction
In 2021, My team shipped a rebooking tool for Aircover Booking Protection that allows agents to proactively reach out to guests who are cancelled on last-minute, and help them rebook, which has brought the rebooking rate from 31% to 69% (1.3 million cases/year), and NPS went up from 22 to 44.
Team
Jessica Chang (Agent XD, Desktop)
Kristenelle C (Guest XD, Mobile)
Jian Z (Product Manager)
Stephanie (Content Manager)
Steven H (Engineer)
Aleck (Engineer)
Cassie (ML Engineer)
Anastasia (Content Designer)
Timeline
3 months design
3 months eng
Background
In early 2021, Brian Chesky approached the Community Support Platform team with the prompt What can we do with a Batphone? and How can we build brand loyalty?
The team identified key points in the guest's experience where they can increase brand loyalty. The most common issue they found was last-minute cancellations, which occur approximately 10,000 times per week or 1.9 million trips per year. In these cases, 86% of guests do not file a complaint, instead they choose to book on a different platform.
In response, the company decided to dedicate resources for creating a new Agent team called Batphone whose sole purpose is to proactively call guests who are cancelled on last-minute.
Problem
Guests are cancelled on last-minute. Out of frustration, they self-book outside of Airbnb 69% of the time.
Breakdown of the problem
1. Agents are not assigned to the case in time
By the time agents start working on the case, guests are already booking on third-party platforms. It takes more than 30 minutes for an agent to respond up to several days. Most tickets get resolved >4 hours.
2. Rebooking options are not ideal
When a guest is cancelled on last minute, there are often little or no options in the area with the specific amenities that the guest needs.
3. Long ticket resolve time
Agent usually has other tickets to work on and often doesn't see new messages from Kim. Old agent tools did not have a priority system.
4. No resolution
Agent usually "resolves" the case by giving the guest a coupon. The guest detracts and does not want to book on Airbnb again. The guest is no longer a loyal customer.
Challenge & principles
How might we design a rebooking tool for agents that is flexible, scalable, efficient, and intelligent?
Flexibility & scalability
The design needed to be flexible and scalable. We were operating in a tight time frame, and the tool needed to be flexible to live as its own microsite, and eventually within the Agent Tools redesign.
Below are a few explorations for how the main components (map and listings) might flex and scale.
Map + List, side by side
Table
Map/List Switch
Map + List, Side by side
After a round of user studies, the Map + List, side by side view was most flexible and scalable, and offered the most familiar user experience. Agents were already used to the card UI on airbnb.com, as well as the map view is fixed to the right half of the screen.
The table view was a strong contestant because it enables column sorting of filters, but the ML algorithm was accurate enough that it unnecessitated the advantage of column sorting.
Efficiency & Intelligence
- Nova guides
- Canned responses
- Smart filters
- Compare old vs new
Challenge
Aleck, lead engineer on Batphone, declares that it is non-trivial to spin up a microsite as the rebooking tool for the March 2022 release
My response
Weeks of back and forth between modifying airbnb.com's wishlist feature and spinning up a new microsite endured. I wanted to expedite the decision making process by illustrating the entire happy paths for both approaches.
I felt that we were getting caught up in debating between two subpar experiences. I wanted to make sure we were hitting the quality bar of providing high-quality proactive assistance to guests.
If we were to build the rebooking tool as a separate microsite, it would likely either get destroyed shortly after the release and require more resources to build a similar tool into the Agent Tools redesign, or move onto other high impact opportunities. Additionally, the system would not be able to properly remember previously sent listings.
I took a step back to create design artifacts for the entire happy path of the rebooking tool built into the Agent Tools redesign, since that was the eventual goal and as a reminder to the team that we were working backwards from the end state.
The result
With these four approaches clearly laid out, we brought them to a meeting with the lead PM, lead engineers, lead designers, and was able to convince the PM and engineers that hitting straight at the target and building the rebooking tool right into Agent Tools would be the most scalable and optimal short and long term solution.
Final design documentation
Final Figma spec
Filters
Rebooking cards
Full agent rebooking experience
Impact
Learnings
I learned that it was important to use design artifacts not only to illustrate the ideal state, but also to illustrate and prove that we might be headed down the wrong path. It was hard for me to articulate that until it was expressed in prototype form.