Close Video


MY work at Expedia

Something Big
is About to Happen

It was with great excitement, in her body and voice, that our team lead told me about a new project. Call center agents will be able to see important information about the customer on the line before the start of the call. They’ll be able to greet the caller by her name and be ready with alternative hotels to offer because they’ll know which itinerary she’s calling about and the reason for her call.

Our whole team was excited. We were building ‘Voyager’, one tool that would replace the existing 14 tools Expedia agents had to master to assist calling customers. There used to be a different tool for almost every concern a caller had (tools for hotels, packages, changing dates, flights, and the list goes on…). Voyager was a single tool that could solve anything a customer is calling about. And this new feature, giving the agent a head start even before the call started, would take the project to an even higher level.

We believed exposing information about the caller would be a tremendous relief for agents. They’d no longer have to struggle with complicated customer names; they’d be able to make calls shorter, take more calls in less time, and make more customers happy. This would be a win-win situation for everyone - agents, callers, Expedia as a business. We were highly motivated, not yet aware of the many challenges we’d encounter along the way.

I was proud to be the lead designer for such greatness.

Expedia Searching for hotelsExpedia searching for hotels

Every Great Project Starts With Research

I needed to figure out how agents greet callers today. I visited two big call centers in the Philippines, and spent the days interviewing agents and listening to agent-customer conversations. I learned some fascinating things.

If you’re living in the Philippines, being an Expedia customer support agent is a great career choice. The offices are equipped with a gym, basketball court, yoga room and pools. Getting hired for this position isn’t easy, and the competition is rough. You need to have very good English, be computer savvy, know how to keep a conversation going, be creative and excellent in math.

Listening to Calls, Interviewing Agents

It seems like there are two types of people who call. First are those with urgent and complex issues. If their problem isn’t urgent or complex, they’d find answers some other way. The second are the lost and lazy. But no matter why people called, most callers already had a hotel in mind.

Expedia call centerExpedia call center

Our Agents Can't Fully Identify With Callers, and Neither Can I

Personally, I hate calling customer support and avoid it if I can. So I wanted to see if our agents could relate to the people they are servicing. I asked agents to tell me about a time they had a need to call support. But none of them ever did. This was really important to understand, and one of my biggest design challenges. I was designing tools for customer support agents, a role I was struggling to understand, and they were servicing Expedia customers, people they couldn’t fully understand.

First Prototypes


Back in Seattle, we worked in weekly sprints, building the product out, experimenting with different design directions. With each iteration we would get on a call with agents in different levels of seniority for usability sessions. We kept this process going for a few months, gradually refining the experience as we went.

Refining the experienceRefining the experience

We learned that the best design is giving agents just enough information about the call so that they can compose a nice opening themselves, without giving them exact word-by-word script, as that made them feel constrained. That balance of just the right amount of information let them feel empowered and in control of the conversation.

The project could of ended here. But agents are just one side of the equation.


What about the callers? We never thought about them

A quick look at Twitter reveals what customers think about customer service at Expedia.

Twitter ComplaintsTwitter complaints

My team is in charge of the company’s customer support. We always raved that user research is easy for us, we have access to every single one of our users (=the agents), and can contact them whenever we want. We don’t need to give them any compensation for their time, so it’s also very cheap. That’s how we’ve been operating for years. Until this project.

While working on this project, I attended the UIE conference organized by Jared Spool, and participated in Kim Goodwin’s workshop about scenarios. Suddenly I realized, we’ll never have a good design unless we test a full scenario, including what the caller feels about how the agent greets him. What if callers think it’s creepy the agent knows so much about them before starting the conversation? What if they feel Expedia is acting like a big brother?

Expedia as big brotherExpedia as big brother

My Biggest Challenge Was Not Even Part of the Project’s Scope

Coming back from the conference, convinced we must set time for research with end customers in addition to the research we already did with agents, I was about to face a huge challenge: getting upper management to spend time and money on research our group has never done before: with end customers. We’d need to hire a recruiting agency, design a research plan, invest in some new software, and use a lab.

Enlarging the budget was not something any executive wanted to hear about, especially as they thought things were working well as they were. So I built a prototype and created presentations, trying to role play and demonstrate how proper user research could lead to not only a better experience, but also save time and money. I invited stakeholders to silent meeting sessions to gather and share feedback, and eventually convinced the head of our division to allocate resources and give it a try.

Silent MeetingSilent Meeting

I recruited the whole UX team for this study and gave each person a role. I wanted them to also realize the importance of true end to end research and get them to be ambassadors of this plan for future projects. This experience and the findings that followed gave me credibility within my group (and broader) as the go-to person for any data related to agent - customer relationships.

Designing the Conversation

We invited customers who called support at least once in the last year to our lab, and tested scenarios in which the agent knew who was calling and why, and scenarios in which the agent knew nothing. We never told any of the participants to call support. We chose scenarios in which eventually they’d have to call, but we let them figure that out on their own.

We discovered customers felt weird when the agent went right into the conversation knowing everything about them. But on the other hand, they loved the fact that the call was so quick. So what overrules - having a quick call or a weird call?

Speed or WeirdSpeed or Weird

What We Learned

Customers, just like agents, want to be in control. They get confused and anxious when things get too “automated” without warning. This is interesting because when we first started this project, we always assumed agents need to have as much control as possible so that the customer can sit back and relax, while having his concern answered. Turns out that is not accurate. Customers want control on their part as well.

Some people felt insecure about the agent knowing all their information right away. They assumed it was based on the number they were calling from and feared anyone can just pick up their phone and change their account details.

Although people felt weird, it was clear the call was fast and painless when the agent knew ahead of time what it was about, and people loved that. Not having to read out loud a long itinerary number could be really handy, especially if the customer is in his car, or at the airport.

People were less surprised about the ‘I already know you’ greeting if they were return callers. Each participant experienced 4 calls, 2 of them were ‘I know you’ calls. Once people knew the “personal thing” is what’s going on, they expected it and liked that it saved time. And, the younger the participants, the less suspicious they felt about the agent knowing who they were.

Translating these findings into a design principle, we knew we want our design to allow a sense of control for both the agent and the customer, and, we want the call to be as short as possibly needed.

Desired ExperienceDesired Experience

“Can I Access Your Account To Expedite Our Call?”

We changed one single thing which proved to be a winning decision. Instead of the agent going straight into the conversation, they started with a simple question - ‘Is it OK for me to access your account?’ This question removed the feeling that the conversation starts like a big brother that knows what you’ve been doing. We were surprised at how much of a difference those few words made. Even though this question adds a bit more length to the call, it creates confidence for callers and freedom for agents.

Agent asking for permission to access customer's accountAgent asking for permission to access customer's account

The new research method we experimented with in this project led to a whole new design approach, that proved to be more satisfying for both agents and customers, reduced call times, and saved a lot of money for the company. We went from 7.6% call resolution to 80% call resolution. After 3 months I was rewarded ‘Product Hero’ and Expedia still uses this design approach today.

A New Look To Expedia Agent Tools

Taking the findings a step further, we made some additional improvements to Voyager. One of them is presenting just one hotel as the main search result instead of many.

It is clear most people call because they have an open question about a hotel they’ve been researching. The old agent tool showed many hotel results, giving the agent many options to choose from and present to the customer. Many options seemed like a good thing, an agent should have all the answers in front of him, and be able to just pull something out his sleeve when a customer wants it. But in fact many options, when presented in random order, creates clutter and more time the agent needs to spend to locate what he really needs.

There should be an easy way to make a connection between search results. There should be an easy way for an agent to value one result over the other based on a customer’s preferences: ‘compared to the hotel you were thinking about, here is a cheaper / closer / fancier one with all other similar amenities’.