Designing a cooking VUI: What we have learned and what we still want to learn
By Abby Alepa, Helen Dominic, & Şeyma Toker Bradshaw
We are a group of three undergraduate and graduate linguists, UX researchers, and foodies, who set out to create a cooking VUI as a side hustle to help us better prepare for the UX and conversation design jobs in industry. A bright-eyed, bushy-tailed student trio dove into the world of AI and promptly realized how expansive this project had the potential to become.
We began without a refined vision for our cooking VUI design and have since made good progress in terms of visualizing what to build. A challenge thus far has been how to actualize this amazing skill we are conceptualizing. In this article, we share the journey that led us to our product vision through “research-integrated ideation workshops” in the early stages. We also share some of the questions we are still grappling with related to design feasibility.
What “should” we design?: How a research-integrated ideation workshop approach helped us
When we started our project, we had no clue what kind of VUI to develop. We knew we wanted to improve the overall experience of the Whole Foods customer but did not know how. It took us three ideation workshops integrated with a fast-paced generative UX research study to settle on creating a cooking VUI Alexa skill for Whole Foods recipes.
First, let’s talk about why we chose a research-integrated ideation workshop approach and how we went about it. In UX design, ideation workshops typically take place after the research team gathers sufficient insights from user research (see How To Run An Effective Ideation Workshop: A Step-By-Step Guide); the goal is to create as many judgment-free design decisions as possible. Rather than separating this workshop from UX research, we opted for a process-oriented, research-integrated ideation workshop approach. We all took on UX researcher roles during the first three weeks of the product development process.
We discussed how Whole Foods shoppers could use Amazon products to better their shopping experience and found that Whole Foods’ recipes feature offers a great resource for a cooking VUI Alexa skill. We next queried: Do Whole Foods shoppers want to cook these recipes with a VUI? What Alexa cooking skills are out there guiding users through recipes as we speak, and what can(‘t) they provide?
We designed a generative UX research study to answer these questions, conducting user interviews with Whole Foods shoppers and Alexa and Google Home device users and analyzing user reviews on Amazon (language data nerd alert!) for Alexa skills “All Recipes” and “Food Network Kitchen.”
We came to the second workshop with our findings: none of the Whole Foods shoppers or Alexa and Google Home device users we interviewed prepare recipes published by Whole Foods, but they would be more likely to cook Whole Foods recipes if such a skill existed. This proved to us that there is a real interest for a voice-first cooking experience for Whole Foods shoppers and recipe users.
To explore existing cooking VUIs, we analyzed the user reviews for other cooking skills to map out our product specifics. We discovered that “All Recipes” and “Food Network Kitchen,” rated 2.6 and 2 out of 5 respectively, do not offer satisfying cooking experiences due to (1) the VUI’s pace of speech, (2) failure to complete step-by-step instructions, and (3) the VUI’s inability to accommodate questions related to repeating a step, ingredients, and measurements. With these insights, we concluded that our voice VUI must excel in these areas.
Following a research-integrated and process-oriented approach to ideation workshops helped us maintain our open-mindedness and creativity as designers while also feeling increasingly grounded in understanding user insights and needs. Collectively taking on UX researcher roles has also enabled us to not rely on our limited experience with cooking VUIs and to consider a variety of user perspectives during the ideation process. We highly recommend this approach to novice researchers and designers who are developing similar products and may lack familiarity with the users and product market.
Know your user: Creating user journey maps
Next, we got into creating personas and user journey maps. After narrowing in on what specific problem we wanted to solve, one of the next choices we had to make was the target audience(s) we wanted to focus on. One of the most common pieces of advice we have received along this journey is this: know your user.
In the early conversations revolving around Cook with Whole Foods, we landed on two target users to focus on in our user journey maps. Since this is our first attempt at developing a low-fidelity prototype, we began by mapping out just one user persona. We are mapping out intents and utterances for the millennial, health-conscious Whole Foods shopper, River (see below). Happy paths have been populated and we are currently working on edge cases.
User Journey Map
Because we are taking this process in stride and learning as we go, we have not begun to tackle a second iteration of the user journey even though we do have one in mind. While conducting research about existing cooking Alexa skills, we found that there weren’t many suitable options for Alexa users with visual impairments. This became our second potential target user.
Our generative UX research study did not include any Alexa or Google Home users who have visual impairments. This is one of the reasons we are not adopting a second target user for now. While this may have been discouraging, having a second target user in mind that isn’t realistic to consider for this prototype taught us a lesson: though there may be a plethora of user types we want to consider, it is simply not within the scope of this project to do so, at least at this stage. We decided to stick to one user journey, for now, to maintain a clear vision moving forward.
So, yes, know your user — but also, sometimes you can’t.
At this point in our project, we are left with some questions: should we begin mapping for other user personas despite the small scale of our prototype? Would creating a user journey map for this second classification of target users result in a sequence of intents and utterances that are remarkably different from our original happy paths (in other words — is it worth it)? Would we be in over our heads if we tailored the product towards multiple personas?
Too few cooks in the kitchen: Developing a VUI without a developer
Despite these questions, we feel confident about our product vision. We are, however, still grappling with questions related to design feasibility. These unanswered questions have forced us to recognize a gap in our expertise. In order to gain insight into these things, we need a developer. Researchers, designers, and developers have to engage in iterative communication to procedurally inform one another in the same way an architect must consult an engineer’s pragmatic functionality concerns before designing a prototype for a skyscraper — definitely necessary!
As three UX researchers and conversation designers without a technical developer on the team, we are not entirely sure if we are asking the right questions about the execution of our prototype. Based on our competitor review research, we understand that elements essential to a positive user experience are lacking in existing skills, but we do not actually know if we are able to remedy those pain points. For example, we initially envisioned a multitasking product that can provide suggestions for ingredient substitutions and seamlessly return to reciting recipe steps; we are not even sure if it would be possible to technically input this in the prototype design.
We also assume that there are many other execution-related questions that we did not consider in the research stage due to our lack of a developer. This has sparked an interest in the process of research, design, and development: is it effective to follow a waterfall trajectory of research → design/execution or will an agile or otherwise constant iterative process of communication between researchers and developers, though time-consuming, produce a more well-rounded product?
What’s next?
This project has been nothing short of eye-opening, exciting, and full of potential. There have been some rude awakenings along the way, but each hurdle has taught us something new about the design process. As UX researchers and conversation designers, our focus is understanding users’ needs and wants and how to create a VUI that will address our findings. We hope that sharing our design progress thus far will shed some light on the questions we have moving forward, and maybe even inspire some wannabe UX and conversation designers to pursue their dreams.
For now, we will keep chugging along, creating dialogue through a conversation design platform (TBD which platform — suggestions encouraged!) to input our happy path as well as different cases including edge cases to create a low fidelity prototype to test on a small sample of users. We know there is likely a bumpy road ahead, but we are buckled up and ready to go. Wish us luck and feel free to share your suggestions with us!
Originally published at https://medium.com on March 16, 2022.
