Depict.ai

Amazon-quality product recommendations for any online store

Depict.ai generates recommendations without relying on customer transactional data, since most e-commerce stores lack a critical mass of data. Instead, we analyze product attributes like product images & descriptions.
discussion
Would you recommend this product?
16 Reviews4.4/5
Hi all! I'm Oliver, I just turned 18 y/o and am running Depict.ai (currently in the YC S20 batch). Depict.ai offers Amazon-quality product recommendations for any online store. Above are two examples from Staples Nordics (where we’re currently live!). We generate recommendations without relying on customer transactional data, since most e-commerce stores lack a critical mass of data. Instead, we analyze product attributes like product images & descriptions. Want to try us out? We offer free A/B-test vs your current recommendations. Setup takes ~5 min. You can expect an overall revenue increase of 4-6%. If you or anyone you know would find https://depict.ai/ useful, please email oliver@depict.ai! I'm here to answer any questions in the comments.
Share
Amazing team! Love the product!
Share
@mokhtar_bacha Thank you Mokhtar!
Share
That is amazing. That means that any e-commerce that just got started can get recommendations straight of the bat, without having any customer transactional data?
Share
@eric_leijonmarck Yup, all existing recommender systems try to use the same methods as Amazon, Netflix and the other giants, the problem is that all other stores don't nearly have as much data as they do. We do it the other way around, instead of creating models that are really good at understanding the transactions, we create Machine Learning models that are really good at understanding the actual products, like a human would do.
Share
Philipp EisenLikes building things with ML
Looks super cool! I have been getting so frustrated with recommendations I get on some online shops. Out of curiosity, what is you business model @oliver_edholm ?
Share
@phileisn Good question Philipp! Our business model is simple. Since we're so confident that our product recommendations are the best on the market, we'll set up an AB-test against your current recommendations free of charge. Based on this AB-test you'll objectively see a significant revenue uptick, and we'll take a percentage cut of this as a fixed monthly recurring fee going forward.
Share
Thiago MonteiroBuilding stuff for the people.
Some years I tried to add a simple recommended products block to a Prestashop ecommerce and it was a nightmare. Most of the time it was showing repetitive items or things that didn't make any sense. Wish this existed back then. :)
Share
@borisandcrispin We know this problem all too well, all existing recommender systems want to do the "Others also bought" logic like Amazon does, this doesn't work since most products don't have sufficient of purchases. Instead of looking at the transactions behind each product, we look at the product like a human would do, and through this approach, we can generate product recommendations that produce significantly more revenue.
Share
Hunting down comments...