AI Case Study
Pinterest enhances image search by not only searching for matching items but also related items such as suggesting recipes for grocery items doubling the click-through rates
Pinterest trained its search engine using its 100 Billion image repository to identify common search teams associated with the item. This feature is available in app and can be used to search any item by pointing the camera at it or using images.
Industry
Technology
Internet Services Consumer
Project Overview
Pinterest used its 100 Billion image repository to train its image search platform to identify the object and predict the terms users will search. If the user likes a particular piece of clothing, it will not only search for the item or similar items, it will also look other products that complement it. Similarly if the image is that of a food item it will look for recipes, sales etc. The visual search is predicting user intentions and searching accordingly. It also uses this data to recommend boards to users.
Reported Results
2X clickthrough-rate to e-commerce sites
2 - 6X engagement for products on Pinterest
Technology
Function
R And D
Product Development
Background
55% of Pinterest users prefer to shop from the site rather than going through a different source
Benefits
Data
100 Billion image repository to train the algorithm
200 million monthly active users