Marriott Hotels VUI Skill: A Case Study

With virtual assistants and smart speakers becoming a more integrated, everyday part of our lives, a voice skill and accompanying smart speaker for hotel guests is a very real possibility in the near future. This week’s design sprint focused on creating a voice skill for Marriott Hotel guests.

What’cha Got?

Our team started this sprint by surveying what information was available on Marriott’s website.

All Hotel Amenities

Hello Information

After doing a little research into what types of amenities and services Marriott has available for its guests, we realized the possibilities are nearly endless. One thing I’ve learned from these VUI design sprints is that while large pieces of information are easily readable and accessible from a visual standpoint, relaying the same information in a feasible way with voice is much more difficult.

Based on the wide array of amenities and services available to guests and our short time frame, we decided to focus on the key amenities which most guests would likely take advantage of and possibly have requests or questions about.

Use Cases

As a team, we chose to focus our user stories on questions and requests surrounding key amenities and basic requests a guest might have for hotel staff (typically what guests might ask for it they called the front desk).

Basic Requests

We tackled basic requests first. The whole idea of having a virtual assistant in every room would be to allow guests to have questions answered and requests met without having to call the front desk. We decided to try this out by discussing what a guest might need more of or might have forgotten when packing.

We started the user journey off by having a guest ask Alexa for more towels. But rather than have the conversation end after the request is met, why not see if the guest needs anything else?

Okay, I’ve got three towels on the way. Is there anything else I can bring you?
What do you have?
I’ve got travel toothbrushes and toothpaste, shampoo and conditioner, and a sewing kit. Would you like any of those?

We decided to have the user ask Alexa what she has rather than assume that the user knows what the hotel has available for guests. This allowed for Alexa to list out other possible options and give the user a feel for what is available just from a basic request for towels. This type of interaction not only meets guests needs but also adds a component of hospitality and personalization by letting guests know what is available to them.

Key Amenities

Moving forward, we began to play around with what guests might want to know about each of our key amenities.

We started by having our guest ask Alexa about hotel amenities and then having Alexa respond with each of the key amenities outline by Marriott.

Alexa, what are your amenities?
We offer high speed wifi, a fitness center, a pool, and a meeting event space. Would you like to hear about one of those?

This allowed users to get a feel for what the hotel offers and hear more about each option. We also expanded from just amenity information by asking guests if they’d like an item which corresponds with their specific question.

Tell me about your fitness center.
Our fitness center is open 24 hours each day and is located on the west side of the second floor. We offer treadmills and ellipticals for cardio, as well as weights and yoga equipment. Would you like us to bring you a water bottle?

With our basic scripting and user journeys completed, we set out to test our ideas with real people.

Testing and Challenges

Testing with our scripts verified some of our ideas about how and what to ask users regarding amenities and requests but it also showed us what needed work and revision.

Challenge 01: Natural Flow and Helpfulness

It was important to us as a team to make sure our skill made guests feel as comfortable as possible. In this case, Alexa is simply an extension of a hotel employee and making sure her responses were positive and helpful was paramount. In many cases, this simply meant changing a few words around but these small changes made a big difference when it came to the tone and flow of the conversation.

Example 01: 1st iteration

Alexa, Can I get more towels up here?
How many towels do you need?

Example 01: 2nd iteration

Alexa, Can I get more towels up here?
Sure, how many towels would you like?

These small changes made a huge difference with our testing. In our first example, by using a confirmation such as ‘sure’, and using ‘would you like’ rather than ‘do you need’, testers felt more at ease and the tone was one of positivity rather than just a run of the mill follow up question.

Example 02: 1st iteration

Sure, the pool is open every day from 8 AM to 10 PM and is located on the fourth floor. Do you need extra towels?

Example 02: 2nd iteration

Sure, the pool is open every day from 8 AM to 10 PM and is located on the fourth floor. Would you like me to bring you a pool towel?

We didn’t realize how the tone came across from our second example until we had begun testing. When we tested with our first user, the follow up question seemed abrupt and didn’t really fit with the flow of the informational statement before it. After our first tester, we changed the wording slightly to make it sound more helpful and positive. The flow and tone were much better with our subsequent tests.

Challenge 02: Introduction Fluidity

After our team had called it a day, I started thinking about introductions while on my drive home. The inclusion of smart speakers and virtual assistants for Marriott Hotels would be a surprise for guests. How would they hear about the Marriott skill? Would they be briefed at the front desk when they checked in or would there be an instructional card next to the smart speaker in their room? I wanted to take it a step further and have Alexa introduce herself. I knew this would need be a brief and informational introduction that would possibly open up a line of communication between guests and Alexa. I wanted to give guests an idea of what the Marriott skill could be used for and possible questions they could ask.

1st iteration:

“Hello and welcome to your room. I’m Alexa, and I’m here to answer any questions you may have about your stay. You can ask me things like, “Alexa, can I get more towels” and “Alexa, where is the fitness center” but you can also ask me things like, “Alexa, where is the nearest starbucks”. Let me know if there is anything I can do to make your stay more enjoyable.”

This introduction hit all the points I wanted to make guests aware of (make Alexa known and heard, and give guests an idea about what they could ask) but it was way, way too long. I wanted the introduction to be more of a quick discovery and ‘wow’ moment for guests rather than a ‘how do I shut this thing off?’ moment.

2nd iteration:

“Welcome to your room. I’m Alexa, your personal assistant during your stay. You can ask me about hotel amenities and services as well as local landmarks and information. Let me know if there is anything I can do to make your stay more enjoyable.”

This briefer intro seemed to fit better while still giving guests an idea about Alexa’s capabilities. While I wasn’t able to test these introductions with users, I did plug them into a voice reader just to hear them read back to me. It became obvious that the second intro sounded better and took less time to listen to while the first intro seemed to ramble on a little too long.

Based on the Basics

These responses and their iterations were chosen based on Grice’s Maxims of quantity and relation. These two principles strive to make VUI design and conversation flow as succinct, helpful, and relevant as possible. You can learn more about Grice’s Maxims here.

The Wrap Up

As this design sprint came to a close, there were a few things I took away from this experiment.

01. Information Design

When our team first started looking into what Marriott offered guests, we were overwhelmed with options and possibilities. Although we didn’t build out all of the possible amenities, we did pay attention to how information was laid out within the site. This gave us clues as to what was most important and what guests would most likely ask about.

02. Tone and Flow

The testing that accompanied this design sprint really spotlighted how just a few words can really affect overall tone and flow of a conversation. Understanding how we wanted our skill to sound and the interaction we wanted to facilitate with guests (kind and happy to help) gave us a basis on how to adjust our scripts for the better.