Group Members: Annie, Damanpreet, Krishnan, Swastika
In this blog, we share the current status of our product Homefood!
1. Solution-agnostic title
College students can feel nostalgic when away from family. With Homefood we want to emulate the feeling of warmth, safety and community that many associate with home.
2. Solution-agnostic opportunity desciption
A local student might have an easier access to family; an off-campus student with kitchen access might be able to bring his family recipe on the table; a lucky student might run into a hometown dish at farmer’s market or a local restaurant one weekend. The list could go on, but at the cost of distance, time, skills, money, and even fortune. At Homefood, we want to sweep away all these obstacles by building a platform that connects these groups to exchange authentic, homemade food, building a unique sense of community/family that centers around food.
3. Potential benefits in the customer’s words
- As a college foodie, I love trying different cuisines, but I often have a hard time deciding what to order, especially when it comes to an unfamiliar culture. I am excited to try Homefood and see recommendations and try signature dishes from home chefs coming from those cultures.
- As an undergraduate student who does not have a kitchen in my dorm or a car, I am so glad to have Homefood delivering food to my door. It is so convenient that I would not need to commute half an hour to a restaurant that is away from campus but still able to have good food. It also saves me a lot of money.
- As an international student who has not gone back home for three years, I would scream if chefs of Homefood could cook authentic dishes from my hometown. I cannot find it in any local restaurant since those dishes are so regional, and I would love to connect with those chefs, especially if they come from the same place with me.
- As a home chef, I enjoy cooking for people around me. If I can make someone else happy with the food I make, I would definitely want to do it. And extra income with that, even better.
- As a home chef, I would love to build bonds with my customers. Unlike a restaurant, I would love to make dishes specifically for the person who is ordering the food.
4. Summarization of a mock customer case study involving this work (with quantifiable impacts)
Our product takes the stories of two people as they work to build a community around food. First, we have Jenny, an empty-nester living in Redwood City. Her kids have gone off to college and she has already retired. She found herself bored in her kitchen, feeling lonely as she reminisced about the times that she would cook for her children. For Jenny, cooking was a love language. It was her method of expressing care and making her kids feel safe. In order to fill the gap left by her kids, Jenny downloaded Homefood, curated her menu and profile and got ready to create a new family to cook for.
On the opposite side, we have Sebastian, a college-student who is away from home for the first time. After eating dining hall food for the past three months, he is feeling nostalgic for his mom’s cooking, something he realized that he took for granted. Feeling isolated from home and in need of a homemade meal, Sebastian downloaded Homefood. He eventually found Jenny, whose roots reminded him of his mother’s.
After ordering his meal, Jenny begins to converse with Sebastian to get a better understanding of how he likes his food. As they converse, the app encourages them to learn more about each other. Sebastian learns about Jenny’s work at Google while Jenny learns about Sebastian’s ambitions to be a doctor. After receiving his meal, Sebastian reaches out to Jenny to thank her. The next week, his craving for a homemade meal returns, and he knows exactly who to turn to: Jenny.
5, 6. Top three areas of uncertainty and plan to explore those areas
Our top three areas of uncertainty are:
- Is the value add of a sense of home/a homemade meal enough to encourage users to choose our product over well-established food-delivery services like Doordash? To navigate this uncertainty, we want to create some small scale tests that will evaluate whether users will order from our platform when giving presented with a doordash alternative. To do so, we will first allow users to browse through Doordash, and then a curated menu by homechefs accompanied by some chef profiles. All menu items will be priced at $15 (for consistency). After doing so, we will then give them the option between a $15 doordash gift card, or a free meal from one of the chefs. We will then see which choices students make and explore how they made that decision.
- Do students and chefs actually want to establish a connection with each other that goes beyond a transactional relationship? This uncertainty will be explored via our role prototype. We will connect a homechef with a couple of students. The homechef will curate a menu with our help along with a profile that the students can reference. Students will order by directly messaging the chef via text. The ordering will happen via text to help humanize the ordering process. We will then provide prompts the chef and consumer to find ways to get to know each other .This will be in the form of small fun facts about each other. We will first see if they engage with the fun facts at all, and then we will see if they use those prompts to facilitate an interaction with the other user. Our goal is to see if both parties are willing to work to create a connection.
- Will consumers trust algorithmic recommendations based on their preferences/community references when exploring unfamiliar cuisines? Our current plan to explore this uncertainty is through two prototypes. The first would be a google form, where consumers can input some cuisines they are interested in exploring, along with food preferences (likes, dislikes, allergies, etc.). We would then provide some guidance for the types of dishes they might like, along with some reviews scraped off the internet for that dish. We would then see how satisfied they are with that dish. We also wish to test this with our technical implementation prototype. This will be a review-based recommendation system where users can still input their preferences, but the algorithm will use home-made data to provide recommended dishes. We will see if users trust our recommendations or wish to do more research on the cuisine they are wanting to explore.
7. Prior research related to this opportunity
Most food delivery services deliver food that is made in a “production line” and reduce the exchange of food down to a transaction and rating . Some companies sell homemade food (eg. Shef), but overlook the subsequent step of setting up the connection between chefs and customers. Thus, only a sense of familiarity around the food may be created, but no sense of home or community. Some companies (eg. Shef and Woodspoon) try to bring more insight into the lives and backgrounds of chefs, but making the connection between chefs and customers one-sided.
8. Leading signals we might observe if this is working (and not working)
Leading signals we might observe if Homefood is working include good number of downloads and customer retention, more home chefs apply to be a member of our platform, increasing number of orders placed by students over time, active usage of chat feature (indicating interaction between chefs and college students).
9. Other options considered?
- More advanced filtering when it comes to food ordering: preferences, allergies, vegan/non-vegan, etc
- Group ordering system
- Pickup Option
- Partnership with shared kitchen
10. Why now, compared to alternatives? Cost of delay
COVID and visa issues have caused more homesickness in students. International students barely get a chance to visit home and meet their families. In such a scenario, Homefood can fill the void by acting as a ‘home away from home’.
Doordash has taken advantage of the COVID period and became the delivery app with more than 50% market share. Doordash was successful largely due to the fact that it incorporated local restaurants and targeted demographics that other apps like UberEats and GrubHub missed. Also, Doordash showed great growth even though it started from scratch and targeted Stanford college students.
Further, California recently passed a bill allowing the sale of meals cooked by home chefs. We have this unique selling point, catalyzed by a promising set of events, and our MVP should be ready in 1-3w. So, the cost of delay trumps over the addition of new/extra features (or quality), and so we should start rolling out the product soon.
11. Summary of technologies
We will implement a review-based dish recommendation system. For the prototype, we will design a Python code that takes inputs and gives outputs in terminals. Initially, we feed the chef profiles, dish varieties, and user profiles into the system. The database is updated every time the users order some dish, and the average rating of the dish/ chef is recalculated based on the reviews given by users. These reviews will be used for ‘recommending’ dishes to the users, sorted with respect to time of delivery, most ordered dish, best-reviewed dish, etc. Other features like ‘Show Menu’, ‘Filter by cuisines’, and ‘Filter by allergies’ will be included as well.
In the future, we would like to use ML for recommendations. Specifically, we will use Reinforcement Learning, where the states will represent the order history of the customer and chef ratings. The action will be a mapping from the customer to a chef/ dish, i.e., suggesting a dish to the customer. The rewards will be the weighted average of the number of dishes/ gross order total by the user, and the ratings of the dishes after order completion. We can use the Keras-RL framework for this implementation.
12. Rough time frame
- 1-3 w: Deliver end-to-end MVP in Python with aggregate review-based recommendation system
- 1-3 m: Implement RL-based recommendation on MVP
- 1-3 q: Add extra features and more scope of personalization, allow more cookies, and improve security of platform
- 1-3 y: Improve scaling and market the product further
Role Prototype: Connecting Students & Home Chefs
For our role prototype, we will have a home chef (Damanpreet’s mother, Kamaljeet) and a student interested in home-cooked Indian food (Damanpreet’s friend, Komal) interact with each other via text in a way to simulate the envisioned chat feature on HomeFood and test our assumption that consumers do want to engage with home chefs and cultivate a connection.
Damanpreet will facilitate this conversation by giving his mother a heads up about the conversation so she knows to expect it, but otherwise, it will be as close to an organic conversation as possible. Komal will text Kamaljeet with the presumption that Komal has already placed an order for one Saag Paneer and two Lachedar Parothay. We will provide Komal with a graphic that will emulate what pop-up instructions that would show up on the HomeFood app before using the chat feature (i.e. how to use it, what to ask chefs/can talk about, terms and conditions).
Komal and Kamaljeet will have their conversation and we will meet with each of them to get their take on the experience via individual interviews.
We will ask Komal and Kamaljeet:
- What did you enjoy most about chatting with Kamaljeet and why?
- What felt most challenging during that experience?
- Is there anything you feel as if the chat experience could have included or excluded?
