The evolution of search at Pinterest

By Naveen Gavini, SVP, Products

When Pinterest first launched it provided a new way for people to collect and organize ideas from around the web. What started as a place to save though, quickly became a unique service for discovery where Pinners could draw inspiration from each other, brands and creators. In 2014, four years after launched, a small group of engineers and designers asked, “what would a search engine look like if there wasn’t one right answer?”.

We realized Pinterest could be the service to help guide you to ideas when you aren’t quite yet sure what you want, and when you want to explore questions like, what to cook for dinner tonight? How to style your home? Where to go on vacation?. With so many possibilities, sometimes you don’t know what you want until you see it. This is when we kicked off our search journey in earnest.

Today, Pinterest powers global search through computer vision and recognizing billions objects through the Lens camera, image and within Pins, and a product search connecting to hundreds of millions of shoppable Product Pins.

In fact, there are now more than 5 billion searches on Pinterest every month.

As people prepare for a post-pandemic life, searches for outfits, vacations, and home renovations are at all-time highs, and searches for weddings have presumed pre-pandemic levels.

For Gen Z, the number of searches per Pinner have increased 31% YoY, while the number of searches within the audience has nearly doubled (+96%) YoY.

When it comes to shopping, product searches grew more than 20x YoY (at the end of Q1).

Search outside the box

As one of the first engineers at Pinterest, I had the opportunity to work on that first version of a search engine, called Guided Search. Guided Search allowed for further refining or narrowing a search. You could start with something generic like BBQ and end up somewhere more actionable like summer chicken BBQ recipes, or vegan DIY BBQ.

Search engines are great when you know what you’re looking for. But the internet was missing a way to explore, a way to start with a few keywords and then get help expanding upon that query to explore possibilities.

The design of Guided Search became used across the industry as a way to click through topics and find the best answer for you. The product paved the way for computer vision-powered visual search. In 2015 we began our work in object recognition within Pins and later brought the technology to camera search to make it possible to search and find recommendations for anything you see online or off.

Search at Pinterest over the years, from Guided Search to search with skin tone ranges and Lens camera to shop. High-resolution image here.

The future of search is visual

Throughout this evolution, search on Pinterest became more visual, and more useful in helping people go from inspiration to action. We brought together the worlds of text and visual search and even created technology to shop the look inside Pins and complete the look. Because people are taking the same image and putting it on different boards, we can learn the deeper semantics of an image. We can then train our systems to emulate the ways Pinners are categorizing images. For example, when you think of a sofa, it isn’t a bunch of words — it’s a mix of images and notions of what a sofa is — and we’re teaching the computer this. Every day we’re making improvements to relate images to how people think of them, through machine learning. Each time we improve our model we see engagement go up.

We’ve also launched the ability to shop within Lens and extended our visual search work to augmented reality with AR Try on for lipstick and eyeshadow.

Along the way, we made improvements to Related Pins, and it has become one of the most common ways people refine their search — by clicking on a Pin they like and then scrolling through (in itself a type of visual search).

A big focus for us has also been inclusive product development, to ensure every person feels represented on Pinterest. This has included skin tone ranges so people can filter beauty searches by their skin tone range. We’ve taken this technology to AR as well, and worked to diminish bias in AI by working with a diverse data set.

Ultimately, we want to help Pinners take action on inspiration — for example, making a purchase in the shop tab or learning a new skill through a creator’s Idea Pin. We’re continuing investments in building our in-house search framework to surface relevant Pins to the right person at the right time, as well as working to make search results as inclusive as possible, including showing diverse results by default, starting with beauty queries, and more to come!

*There are more than 5 billion searches on Pinterest each month (Searches is inclusive of text searches and visual searches, which includes zooming in on an object in a Pin to see visually similar Pins.) This stat reflects global searches over any 30-day period since April 2021.




Inventive engineers building the first visual discovery engine, 300 billion ideas and counting.

Recommended from Medium

Design the Thinking

The Show-Stealing Hair & Makeup Trends From Arise Fashion Week 2019

#20 Truths about people that we design

Architecture & unbuilt legacy

I decomposing plane as an alalogue for limiuted thinking. Aplane does one thing, and then it’s gone.

Holiday Gift Guide for Designers under $50 (2021 Edition)

In defence of focus groups

Redesign Spotify collaborative experience

A cover image for “Redesign Spotify’s Collaborative Experience”

What is intuitive design?

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Pinterest Engineering

Pinterest Engineering | Inventive engineers building the first visual discovery engine

More from Medium

One AI, streaming videos for all

Improving Virtual Card Numbers with Edge Machine Learning

woman sitting at desk typing on computer

Addressing Python Dependency Confusion at Pinterest

Data Driven (Technical) Debt Reduction