Mindful Subscriptions

Inspiration

Today, we all are subscribed to a plethora of services. There is also an industry trend. According to a Forbes report, the global subscription e-commerce market is expected to grow from $72.91 billion in 2021 to $120.04 billion in 2022. On average, one in two online platforms you use probably has a subscription option [1]. With so many options and new streaming platforms entering the market, consumers are actively making decisions on which streaming service is the best value proposition for them. A survey conducted in 2021, showed that consumers were excited about new streaming services, such as Disney +. "And nearly two-thirds of consumers intended to terminate or downgrade one or more of their current subscriptions to make room for a new service" [2]. To make matters worse, free trial marketing strategies have exploded since the pandemic, and inflation rates in recent months have spiked to 10% or more in some countries, putting pressure on consumer spending in ways that have been unseen since the 1970s. Streaming giants (such as Netflix) and other subscription services (such as Audible) have hiked their monthly rates in 2023 and with a stressed economy, shown through clear economic trends from high inflation and layoffs in many countries, consumers will want to reevaluate their entertainment options [3]. How do we keep track of all these different subscriptions and how can we evaluate where to cut costs in a currently stressed economy?

In order to help users manage their subscriptions, we decided to build Mindful Subscriptions with the help AI, using Cohere's APIs! With Cohere's API XXX to , we created a classifier to classify bank statement transactions vendors such as from vendors who use a Transaction Model (you pay as you use the products) vs those who use a Subscription Model (a periodic (monthly, yearly, or seasonal) payment to gain access to products or services). We also integrated Cohere's API / SERP API powered contextualized factual bot namely, the Grounded Question Answering bot to offer additional information on existing all Pricing Plans for the services that are users are subscribers for.

What it does

The Mindful Subscriptions application helps users manage their subscriptions by providing them a list of their active subscriptions and their associated cost. They are also grouped based off of their category such as streaming, food, entertainment, etc.

This project offers users transparency and simplicity into spendings on the various subscription services they are subscribers to. Using our application, they can explore the costs of their paid subscriptions and explore the ways to cut costs. The application dashboard allows users to view the possible the savings they can incur by removing certain subscription from their recurring expenses or by seeing alternative subscription tiers that they could subscribe to instead. On this application, users can freely explore their options on how to meet their needs as well as their financial goals.

Features:

1. Dashboard
   • bankstatement upload: upload bank statement csv
   • subscriptions: add/delete/edit(manage)
2. Toggle Pricing Intel
   • Get more information on the subscription and its pricing plans so that you can re-evaluate the best Pricing Plan to meet your personal needs and financial goals -- Cohere's API / SERP API powered contextualized factual bot namely, the Grounded Question Answering
3. AI Subscription Parser/Classifier
   • AI powered categorization of Transaction Model vs Subscription Model (subscription classification) vendors from historical bank transaction data (bank statement) (Beta)
4. AI Category Classifier:
   • AI powered classifier of the subscription categories (e.g. Education, Entertainment) (Beta)
5. Reminder Feature (Not Implemented)
   • send notifications 3 days in advance of the next renewal date

note: * beta implies implemented on the backend only

How we built it

FrontEnd - Add/edit/delete subscriptions

• Reducer Logic to Manage Stateful Front-End Variables such as JSON to turn messy data into interactive and immersive insights

• Engages User to think about the Price of their subscriptions. What are their most expensive subscriptions? What are the cheapest? Which subscriptions sound ideal/frown-creasing according to contextual information gathered by Cohere API? Which subscriptions can become much more expensive than other subscriptions?

• Key-value pair (see example below)

• Parse Bank Statement to populate subscriptions Backend

• FastAPI

• Cohere (sandbox-grounded-qa https://github.com/arjunchandra/sandbox-grounded-qa) - AI subscription classification from historical bank transaction (0/1), AI gathered pricing plan information,

• Categories classifier to classify the subscription categories (e.g. Education, Entertainment)

Challenges we ran into

• The subscription classification using Cohere ran into headwind during the development as it worked well on a small scale in the playground but not with real bank transaction data when you have 50+ transaction histories.

Accomplishments that we're proud of

• We successfully integrated Cohere's QA bot in the front-end to check for the price intel

What we learned

• We learned together about how to make well use of cohere to make AI understand what we truly want based on a bit logics behind the scene. We also realize the potential of AI is infinite and also cohere is such as strong natural language processing model: they can achieve any features as long as we have the ideas and have set everything correctly, which is definitely incredible!
• We just accomplished a project with features we have never thought before as a team! Personally, I think what happened within these approximate 2 days has really blowed my mind. Initially I didn't know anything about frontend, backend, and endpoints, how to connect them together, and I even have never heard of cohere before. But big thanks to my teammates, they explained everything to me very patiently and resolved my concerns without any complains. They also told me to not stress too much about it because everybody starts at 0, and they all learned through a step-by-step process. Really appreciated! I felt like that's the actual significance of the event: not just people getting together, but mind and heart getting together.
• Debugging large language model's performance can be hard. This is why it's important to boardening the thinking when approaching using NLP models for a new use case since they will have their own strength such as content generation, etc.

What's next for Mindful Subscriptions

• Chatbox for the user to ask more information about the subscription such as steps to unsubscribe from the subscription.
• Price checker: check if there is student discount

Link to the powerpoint: https://www.hollywoodreporter.com/business/business-news/apple-disney-netflix-price-increases-1235258312/ [1] https://www.forbes.com/sites/jiawertz/2022/07/15/the-growth-of-subscription-commerce/?sh=676539a6b572 [2]https://koreascience.kr/article/JAKO202119759275785.page [3] https://www.hollywoodreporter.com/business/business-news/apple-disney-netflix-price-increases-1235258312/

Built With

• ai
• api
• cohere
• fastapi
• localstorage
• material-ui
• react