Inspiration
Think a baby between Honey (the online rewards extension) and Klaviyo (the eCommerce customer behavior app), but for PHYSICAL retailers. There are two sides to this problem. Grocery pricing is being taken advantage of and is starting to change day by day based on demand (so customers need an edge), but also brands in stores lack visibility into how in-store customer behavior translates to purchases or missed opportunities. We wanted to bridge that gap using real user contributed data.
What it does
Aisle IQ is a dual-platform system:
-A consumer app for scanning products, comparing prices between stores, and most importantly, predicting WHEN and WHERE your cheapest grocery trip will be
-An AI powered machine for vendors to get analytics, competitive insights, and ML-driven pricing optimization, based on priorities such as market share or overall revenue.
Currently, there is nothing on the market that delivers this intimate level of data between a human and a product. The closest competitor for similar data collection is Doordash, which recently started asking dashers to take pictures of store shelves (indicating a new market to pioneer).
10 Years ago, Tesla was the biggest player for road/traffic/driving data that was one of their most profitable products, but today there is no company that has fully labeled data of humans interacting with physical products in a store setting. We hope to be that soon.
How we built it
We built two Next.js apps connected via Firebase for real-time data. A FastAPI service integrates a TinyTimeMixer model that we trained for price forecasting and optimization.
Challenges we ran into
No reliable real-time retail pricing APIs Business logic of how to get a important piece of data while also benefitting the donor Integrating ML predictions into a live dashboard
Accomplishments that we're proud of
End-to-end data pipeline that goes from scan to insight seamlessly Real-time analytics dashboard with meaningful metrics, like what day to plan your next grocery trip (customer side), and on the flip side (for vendors) what price to reduce your products by, or what placement will help you get more market share Working ML-based price prediction + simulation for users throughout the week
What we learned
Real-world data is messy, but INTENT is what is valuable. Knowing a customer physically walked in, debated between products X and Y, yet left with Y says a much more important story than the # of stock X has vs Y has (which is currently all the industry provides). Building something where users GAIN something from the data they contribute is a much more friendly way of collecting data. We added rewards, and ensured that every piece of data they gave had a benefit to them (receipt scan, product scan, shelf scan all give them a cheaper haul)
What's next for Aisle IQ
Collect daily data from surge pricing grocery stores Improve data quality and scale user adoption Partner with retailers for discounts, rewards, etc. Figure out how to use/collect this data for robotics
Log in or sign up for Devpost to join the conversation.