Every single day, restaurants throw away perfectly good food because it didn’t sell during breakfast, lunch, or dinner. In the US, up to 40% of food produced is wasted. A major culprit among this waste are restaurants, who throw out 43 billion pounds of food per year or New York City specifically, where over 650,000 tons of food waste is produced. In NYC, 20% of this trashed food is not spoiled or leftovers, but rather simply just unsold meals from these kitchens’ breakfast, lunch, and dinner times. In addition to wasted food, restaurants also waste their resources such as electricity and staff during non peak hours. We understand that managing resources efficiently is an important facet of Smart Cities, and so we built Munch.
What it does
Munch offers an Android App, Alexa Skill and Online Dashboard that connects restaurants with customers, allowing them to sell otherwise unsold food at a discounted price at specified times. It essentially facilitates a clearance sale but for food! It reduces food wastage at restaurants, lets consumers grab inexpensive meals and regulates customer flow at the restaurant to optimize their resources and maximize profit on otherwise thrown-out food.
The online dashboard lets restaurants create special “boxes” of items that they would like to sell at a cheaper price. Restaurants have access to a graph of current traffic at the venue and are provided suggested times to schedule a discount based on past data and predicted traffic. The customer volume data would be collected by overhead cameras at the restaurant entrance, tracking the number of people walking in and out using OpenCV.
The app lets users search for active discount boxes in their area, reserve a box, and then head over to the restaurant and collect their meal! In order to reach a wider audience, we also allow users to search directly from Amazon Alexa by simply asking “Alexa, ask Munch to show me discounts near 08536”.
We plan to charge restaurants a $1.00 royalty per box sold. By charging per box sold, we offer a no risk & high reward system as opposed to a monthly subscription or one time fee where restaurants would have to pay regardless of their sales. Also, since restaurants have the opportunity to sell food that would previously have been thrown out, there is simply no downside! By offering restaurants such an incentive, we can ensure that many restaurants will sign up for the service and of course, lead to profit!
Our service also has almost no cost associated with it. The only cost would be the Firebase backend as the service begins to scale. Even so, firebase can support 100k+ users for a monthly cost of under $50. The low cost of our business which only scales up with more users means that we only incur costs that are offset by profits.
How We Built It
Challenges we ran into
We ran into a few challenges over the course of this hackathon. One of the first was setting up OpenCV to recognize the contours and we had to learn about various processes to filter images such as erosion and dilation. We also struggled with figuring out a way to visualize the data (number of people at the restaurant in real time and patterns from past data). We ended up using a library called chart.js to plot the two graphs and then suggest discount times. Perhaps the biggest challenge that we faced was integrating the various aspects of the app : the web dashboard, app and alexa skill all had to push and retrieve data to and from each other and after getting the individual parts working, we had to make many modifications to the database structure.
Accomplishments that we're proud of
We built a functioning project with an app, website, people counter and Alexa Skill in 24 hours! We are proud of the way we planned and delegated the various parts of the projects and sacrificed our sleep!
What We learned
We learned some new tech skills such as using OpenCV and Firebase, but we also learned the importance of planning and delegating tasks in a team.
What's next for Munch
In the future, we plan to improve on our best-time-for-discount predictions by using Machine Learning to analyze past datasets and predict the best times to set discounts. Integrating machine learning will also let us suggest the best prices for the discount boxes, based on the profit driven by customer volume and pricing.