Tender is a prototype online platform for building and training personal shopping bots to help users reduce the time, mental load, and annoyance of searching through e-commerce sites for the perfect product.

Tender helps users:

  • Seamlessly communicate their information through linked social media profiles
  • Quickly and easily setup their ShopBot assistant
  • Train their ShopBot over time with minimal daily prompts about products

How It Works

ShopBot works by scraping data from user’ social media profiles including age, personality, and likes/dislikes to build an initial model of their stated preferences compared to a database of others users. ShopBot then learns user’s specific preferences by showing a series of images of different products and asking them to like or dislike them, while also tracking engagement with each product through facial recognition. ShopBot continues to learn preferences daily, sending out a notification asking users’ to swipe right to like, or swipe left to dislike a handful of products. All of this information is run through the Tender neural network, constantly retraining and perfecting each ShopBot’s individual purchasing ability.

Looking Forward

Today, e-commerce websites attempt to match us with the perfect product by monitoring our purchase history, collecting previous products viewed, and sending customer quizzes and surveys to capture our stated preferences. Others employ a subscription-based model leverages artificial and human agents to monitor us over time, signalling our preference by the items we decide to keep or return.

Imagine If…

…your personal shopping AI, you ShopBot, could comb your social media profiles, purchase history, and even reaction to specific products across websites to help you find the perfect match, every time. Imagine if your ShopBot could know your preferences, price points, and events so well that it could ship you what you needed, when you needed it, and ensure that you loved it.

Technology Used:

For Social Media Profile Scraping

  • Facebook API
  • Apply Magic Sauce Datamining API

For Training the Sockbot

  • Affectiva API for Facial Recognition
  • TensorFlow, Keras, Jupyter for Machine Learning

User Interface

  • node.js
  • HTML/CSS/Javascript

Check out the code on Github.

Contributors: Dana Martens, Noa Kaplan, GS Shop Team Members

Tender was created in a three-day exploratory sprint as part of the IDEO CoLab Summer 2017 Fellowship in San Francisco.