An Alexa skill that helps find recipes and gives easy-to-follow step-by-step cooking directions.
Voice Prototype created using Voiceflow
TARGET AUDIENCE:
This skills is for anyone who wants to prepare quick meals, and who is familiar with voice technology. The target group will be single men and women between 18-40 years old, which are young adults who have a busy professional life and live independently.
INTERVIEWS:
In order to learn more about the users' goals, current frustrations, and pain points in cooking and using recipes., I set out to do interviews with four users.
Due to COVID, these interviews were all conducted remotely via video chat.
INSIGHTS:
All participants cook everyday and at least twice a day.
Most participants only look for recipes right before cooking.
Participants decide what to cook based on what is in their fridge that day and depends on the freshness of the ingredient (use less fresh items first).
Generally participants want to use recipes to expand their repertoire, get good at some dishes, and try to make dishes that they tried in restaurants.
Participants wants the skill to remember their preferences and don’t mind hearing the same suggestions, but they also want variety in the mix.
Participants spend approximately 1-2.5 hours cooking every day.
Participants want to hear instructions that are not cumbersome and not so technical.
Participants want to know which ingredient is most important in making that particular dish and what she/he can substitute for less important ones.
Participants are busy professionals; they often cook food for many days at once, so food that store well is favored.
Participants find videos useful in demonstrating the cooking process/ certain techniques.
THE USER PERSONA:
The user is a 30-35 year old single working professional, who is often too busy to cook for herself. She can use some help in choosing good recipes, using the ingredients in her fridge, and being guided through simple cooking directions.
THE SYSTEM PERSONA:
Designing a voice interface is like conjuring a character in a play or dance performance. The personality of the voice assistant plays a big role in how a conversation with our target user will play out. I created Jun, a chef who is straightforward when giving cooking instructions but friendly and casual. She uses simple language and only offers details when she is asked. She also has a sense of humor and likes to creak jokes sometimes, but not during cooking!
USER FLOWS AND SAMPLE DIALOGUES:
With the help of insights from user interviews, I then moved onto creating sample dialogues and interaction flows that simulate how the input and output of the dialogue go. Based on what I learned from the interviews, I created five user stories with their corresponding sample dialogs to start mapping out the experience. I then map each intent to each of the user stories. I created 4 intents, along with Amazon's built in global intents, I have a total of 5 intents:
Overall Flow
I then wrote some user stories for each intent, then use the stories to create conversation flows.
CHOOSEMEALINTENT
As a busy professional, I want to cook a bulk of food that will last for 4-5 days, so I don’t have to cook every day.
U: Alexa, open minute recipe and find me something that I can keep for a few days.
A: Hi Cecilia. Stews, casseroles, and curries are good for storage, which would you like?
U: Stew
A: For stews, I recommend Beef Stew, Garden chicken Stew, or the Curried Lentil and Rice stew, what do you feel like?
U: How long does the garden chicken stew take?
A: The Garden chicken stew is the quickest dish to make - 30 minutes.
U: Sure, that sounds good
A: I’ve sent a list of ingredients to your phone. Take a look, do you need to order anything?
U: Just order orzo, I have the rest.
A: I added orzo to your shopping list. Do you want anything else?
U: No thanks.
A: No Problem.
CHOOSEINGREDIENTINTENT
When I’m deciding on what to make, I want to search by the ingredients available in my fridge so I can make use of the food I already have.
As a busy professional who lives alone, I want to be offered meal options with consideration for my balance of nutritions that week, so I get a balanced diet without eating too much of anything.
When I am cooking, I want to be provided the information on what substitutes are available for some ingredients that I don’t have on hand.
U: Alexa, open minute recipe for some dinner.
A: Hi Cecilia, wanna use ingredients in your fridge? Tell me what you have.
U: Sure. I have some eggs and chicken, what can I make ?
A: I recommend chicken parmesan then. Or do you prefer something else?
U: OK. Is it easy?
A: It’s a beginners level recipe. it only takes 20 mins and 4 ingredients.
U: Ok. I’ll try it.
A: I sent the list to your Alexa app to have a look.
A: Do you have all the ingredients?
U: Alexa, I don’t have buttermilk though…
A: You can add a tablespoon of lemon juice or white vinegar for every 1 cup of milk to make a buttermilk substitute.
U: Oh ok, that’s great.
A: OK. Let me know when you are ready to cook.
GETINSTRUCTIONINTENT
When I am cooking, I want to be able to control the voice instruction with play, pause, or stop the recipe, so that I can follow along.
U: Alexa, start recipe.
A: Do you want to start cooking roasted tomato soup?
U: Yea.
A: First, lay tomato halves cut side up on cookie sheet. You can ask me any questions about this step, or let me know when you're done.
U: Done.
A: Next, brush olive oil over each half (top and sides).
U: OK. Next.
A: Next, add a sliver of garlic to each tomato and sprinkle seasonings over them. Any questions about this step?
U: No, next.
A: Next, roast in oven until charred, about 20-30 minutes.
A: Would you like me to set a timer for 20 minutes?
U: Yup.
A: Timer set for 20 minutes.
A: Bipbipbip…Check on your tomatos, if they are not charred yet, I’ll set a timer for another 10 minutes.
U: They are done!
A: Great. Your tomato is ready to be blended in blender and pour into large pot. Enjoy your meal!
SAVERECIPEINTENT
When I’m deciding on what to make, I might prefer to stick with what I like, but I also want variety from time to time.
U: Alexa, save Chicken Parmesan as my favourite.
A: Sure. Chicken Parmesan is now added to your favourite dishes.
A: To hear all of your favourites, say my favourites.
U: My favourites.
A: Chicken parmesan, sweet and sour chicken, and curry beef is in your favorite. Would you like to try one of these for dinner tonight?
U: hmmm.. Chicken sounds good.
A: How about the Sweet and sour chicken? You only need chicken, flour, and soy sauce. You can also mix in whatever vegetables you like. Wanna have that for dinner?
U: Sure, that sounds good.
A: Let me know when you are ready.
WRITING VOICE SCRIPT
LAUNCH PROMPT FOR NOVICE
STANDARD LAUNCH PROMPT
TAPERED LAUNCH PROMPT (FREQUENT USER)
TAPERED LAUNCH PROMPT (INFREQUENT USER)
CONTEXT (INFERENCE)
CONTEXT (MAINTAINING STATE)
MULTIMODAL CONSIDERATIONS
Many users expressed in the interview that they found video instructions the most useful during cooking. Sometime they don't know how to cut something or would like to know what the progress should look like, videos are great to get them up to speed quickly.
Images are also extremely useful in helping users make decisions on whether to make a dish or not. When being told a long list of ingredients, it is also helpful to have a visual reference in order to reduce cognitive load as much as possible. Users who owns a screen-equipped device would definitely benefit from a visual component along with the voice.
Examples of accompanying visual interface:
USABILITY TESTING
With the script of the major interactions written, I then tested the skill with 5 users from the target audience using the Wizard of Oz method.
The test aimed to access the following:
- Do the conversations feel organic/natural?
- Can users easily find the recipes? Discover the learnability of the skill.
- Does the participants feel it easy to navigate the recipe instructions using stop, pause, next?
- Measure the amount of time they need to find a recipe through voice.
- Are there additional utterances users might say?
- Does the system personality feel "right" to users?
KEY FINDINGS
Issue 1: an overview of their options
High Priority
Users wanted all the choices right "on the desk." The current system just gives 3 options without telling users how many options she/he has in total, causing confusion.
Before:
A: For stew, I have beef stew, would you like that?
Suggested Change:
A: I have 3 stew recipes in total, including Beef Stew, Garden chicken Stew, or the Curried Lentil and Rice stew. Which would you like?
Issue 2: Establish credibility
High Priority
Users felt that they don't know where the recipes are coming from, so they find it hard to trust that the recipes suggested by this skill will be good. They'd like to hear that the recipes are from famous chefs or websites.
Before:
Welcome back Cecilia. I can find you recipes for breakfast, lunch, dinner, or snacks, what do you feel like?
Suggested Change:
Welcome back Cecilia. All my recipes are pulled from the New York Times Cooking, Jamie Oliver, Julia child and more. I can find you recipes for breakfast, lunch, dinner, or snacks. what do you feel like?
Issue 3: users would like to know how many details/ingredients they can include in their request.
Mid-Priority
When users ask for recipe by the ingredients they have, they got confused how many ingredients they can list. They would like to be told that information so they won't break the system.
Before:
A: Hi Cecilia, wanna use ingredients in your fridge? Tell me what you have.
Suggested Change:
A: Hi Cecilia, wanna use ingredients in your fridge? I can use up to 3 ingredients in a recipe. Tell me what you have.
