swiss Chalet Google Assistant
Introduction
In 2017, as part of a major marketing announcement of Google Assistant in Canada, Recipe Unlimited, was approached by Google to create a chatbot for making orders at Swiss Chalet. The CTO of Recipe Unlimited, working with one of his technical groups, put a call to the Lab to help with the development. Together with a technical team at Recipe Unlimited, the Lab team developed and helped launched the MVP of the Swiss Chalet ordering chatbot.
Who was involved
FairVentures Lab team: 2 Sr. Developers, 2 Co-op Developers, Product Designer (my role)
Recipe Unlimited: Product Manager, Developer, QA Assurance Engineer.
Google: Business development manager, Support developer.
Problem
How might we enable existing Swiss Chalet customers to order a dinner through a simple voice or text interaction using Google Assistant?
Solution
Swiss Chalet Google Assistant that allows Swiss Chalet customers to order their defined “Quick Order” and their “Last Order” through voice and text.
Process
Story mapping to explore the scope and identify priorities for the MVP.
Crazy-eights with Lab team explore solution ideas.
Workflow diagram to identify API needs with Swiss Chalet team
Conversation flow diagram (Figma) to evaluate design and specify language, voice and tone with development team.
In previous project the Lab would prototype chatbot functionality using Chatfuel because it was fast and easy to create a basic and functional chatbot with multiple interaction types like carousels, images, quick replies, and menus. In this project, because the controls of Chatfuel were not consistent with Google Assistant, and Google Dialiogflow was cumbersome and intent-based (not flow-based) we chose to prototype the conversation flow in Figma. This allowed us to indicate the entry points, the branching of the conversations, visualize the scope, and highlight usability breakdowns. The collaboration features of Figma also allowed the PM at Swiss Chalet to review the designs, understand any usability breakdowns and discuss trade-offs. Figma worked very well for bug-fixing by allowing developers to comment on screens where there were questions or issues to be addressed.
Usability testing with a paper prototype.
When we discovered that the scope of the initial launch was going to be limited and not include menu navigation, we needed to better understand how that was going to be received by potential customers and what voice utterances they would use to interact with the Google Assistant given different scenarios.
We tested with 2 Lab members that regularly order Swiss Chalet for their family and who were unaware of the direction of the project at its early stages. We learned that there was concern and disappointment that they would have to download the app and create an account to order. There was also some disappointment in the limitations to what they could order, they probably wouldn’t have a single quick order for their family and would want to customize each time. After a meal was selected and the assistant was moving through the payment and delivery process, there was much more confidence and positivity. Both users expressed they were impressed at the idea of being able to order through a Google Assistant.
We later conducted usability testing of the live voice system with 3 customers which helped us understand how users might naturally want to voice through key scenarios. We learned where our programmed responses from the Google Assistant were awkward and where we could make language, voice, and tone light, casual and efficient - like what you’d expect if you called into make an order from a human.
Design Challenges
Primary Interaction
The primary interaction was designed to take customers to order choices as quickly as possible. Due to scope limitations in full menu navigation, we displayed their Quick Order, their last order, a way to change their Quick Order through the Swiss Chalet app or online. Usability testing of this flow revealed that it was easy for users to understand and move through in voice interactions. This was strong “happy path”, but there would be challenges for the user to arrive at this state, which is explained below in First Use interaction.
First Use of Swiss Chalet Google Assistant
In the first use interaction, we try to temper high expectation with an opening statement expressing this is an “… early version of the app…” and it has a limited set of features. The user is immediately asked to link their Swiss Chalet account with Google Assistant. This message unfortunately could not be modified and if we could would have indicated they could create a Swiss Chalet account in the process. Very few Swiss Chalet customers have an account, so we believed this would be a big detractor for a lot of customers in the early version.
After moving through the account creation and linking, if the customer didn’t have an initial order we had to direct them to the app to make that initial order. This is another major barrier to adoption. We evaluated the ability to make a popular order and indicate the number of people, but there was no way to clearly manage the choice of that popular order through the APIs. We do our best to soften the disappointment and encourage follow-through by offering a coupon.
The Lab team communicated these barriers very clearly to the Swiss Chalet product team making it clear how it was going to affect adoption. We expected users without an initial account would not likely sign up and only those that have an Swiss Chalet account would be encouraged to use the assistant.
Outcomes of the Project
We launch the Swiss Chalet Google Assistant with Google’s announcement in Canada. Canadian Swiss Chalet customers can now order meals through Google Assistant (and or course, we ordered lunch for the team through the system!). We subsequently shared the outcomes of that project with the other retail and restaurant companies through a Fairfax Retail and Restaurant Day event in June 2019. Admittedly, the limited scope of the chatbot makes it hard to know how widely adopted it will be and we’re awaiting the annual update on the results of the number of users that used the assistant. Launch of the assistant made headlines through a number of outlets:
https://www.restobiz.ca/swiss-chalet-offer-voice-orders-google-assistant/
http://strategyonline.ca/2018/05/31/swiss-chalet-brings-ordering-to-google-assistant/