Wynn Nightlife Chatbot

Background

The Wynn Nightlife Chatbot is a chatbot designed to easily answer questions about the Nightlife at Wynn Las Vegas. Since this is an interaction bot, it features both text and UI visual elements for users to type and click on images, links, menus and buttons. Some example questions users can ask include what time is XS open, when is Marshmello playing, is Alesso playing on April 4, what’s the dress code for Encore Beach Club, how much are tickets, etc.

The Problem

Wynn’s Nightlife website lacks the information needed for nightlife customers, who are unable to find basic information such as ticket pricing, table pricing, baggage, and dress code. Compared to competitors in the nightlife industry, this information was difficult to find without contacting a live person.

Defining the Goals

Since it’s difficult to navigate and find information on the Wynn Nightlife website, this goal of this chatbot serves to provide less distractions and more personalized functions to help users find what they need. In addition, we also wanted to explore ways to provide another source of revenue for nightlife merchandise.

Research

In the discovery phase, I interviewed the Nightlife business stakeholders to understand their needs, goals, and what were the current pain points on the nightlife website. I helped gather the business and product requirements. Throughout this project, I managed the communication and collaboration with the chatbot vendor, Passage.ai.

Competitive Analysis

To better understand how chatbots functioned and what was offered, I compared popular chatbots such as Bank of America’s Erica, Domino’s Pizza bot, and Ebay bot. These bots lived on platforms such as Facebook Messenger, Apple Business Chat, and within their respective mobile apps.

User Interviews

During this phase, I interviewed direct customers (in person) from clubs such as Encore Beach Club, to understand the audience and their experience with chatbots.

Mind Maps

I began to create mind maps to map out user flows and provide lists of example intents (subject or topic that needs to be matched to the entered text) and utterances (any action on the intent that chatbot must respond to). Listing out the possible utterances paved the chat ux flow.

Working with the developer from the chatbot vendor, we created a beta version and conducted user interviews and user testing. We worked with our Nightlife team to provide content in a knowledge base format and utilize events/ticketing APIs for the chatbot to consume.

User Testing & MVP

After going through multiple internal iterations, updates to content and internal user testing, we launched an MVP on March 30, 2019. It was deployed first on the website and later on on Facebook messenger.

Metrics & Takeaways

From the period 3/3/19 – 4/12/19, we had:

• Unique users: 757
• Average Number of Messages per User: 3.69
• Total messages: 2.79k

Chatbot accuracy

Takeaway

Looking at the input provided by customers, we observed that the top theme of questions involved event detail, table and package pricing, bag storage, dress codes, and the different variations of ticket pricing (regular, expedited, general admission, vip, etc). The bot was able to effectively answer over 80% of questions. With continuous iterations and improvements, the goal is to get 93% accuracy.