“This crazy human indexing machine”
IDS 411 Sec. 56
In the seven years since the first prototype was launched in March of 2010, Pinterest has grown from a tool co-founder Evan Sharp used to track images for his study of architecture, to a tool used monthly by more than 200 million users worldwide. It boasts 100 billion objects in its database; 100 terabytes of data is added daily. Beyond the craft, DIY, cooking, and wedding planning tool it was once perceived to be, it’s also a place to find creative talent, templates, business plans, and free software — really anything that can be linked. And it’s become a premier shopping site, providing endless user product endorsements. Advertising revenues for 2017 are expected to reach $500 million.
All data on Pinterest is contributed by users and is displayed using a pinboard metaphor. In early versions, users clicked ‘pin’ to save a post to a digital pinboard of their own, though this was later changed to ‘save’ with an image of a push pin.
Pinterest uses card navigation, creating a very simple home page. Cards or pins which appear are selected for the user based on the following:
- Previous user searches and use of the “explore” button
- Pins a user has saved to their personal digital pinboards
- Information from advertisers, whose pins a user has interacted with (can also be switched off by user)
The first level of navigation includes only the following:
The most important element in the Pinterest hierarchy of data is the pin or link; navigation is organized around finding and interacting with pins.
- The Pinterest home page selects and presents pins based on previous site interactions, as noted above.
- Two of the seven navigation links on the home page connect back to the home page — each of these links refreshes the pins displayed every time they are clicked.
- “Search” and “explore” present new pins to the user in either selected or popular categories.
- Notifications encourage interaction with other users and pins. These include notifications that other users have pinned an item from the user’s board, new pins that fit a user’s profile, and pins from followed users.
- The remaining two links are for managing a user’s account, profile, pinboards.
Second level navigation
Searching from the home page not only presents pins related to the search terms, it also builds a second level of related category/search-term navigation on the fly, as shown below.
This not only promotes visibility of pins, it also provides users with ideas for related search terms that can get them closer to what they’re seeking. As a tool to help refine searches, it’s both useful and intuitive. A user might search for “marketing collateral”, for example. But are “corporate identity” or “brochures” a more exact match?
Pins are typically URLs, made intuitive
Pinterest is an elegantly simple site, built on pins, which are often simply links or internet “bookmarks”. Pinterest co-founder Evan Sharp struggled to use architectural images he’d downloaded from websites in an effective way. He and co-founders Ben Silbermann and Paul Sciarra flipped the paradigm, creating instead a tool that captured links to these website objects, in a visual way that made the links (urls) accessible by adding related content. Images of website objects became easily scannable, along with the related data contained in the visual pin of the link.
The data elements involved are simple: an image from the linked object, the link (connected to both the image and a “read it” button in the pin), information about who posted it, plus their comments / thoughts if any. Then over time, information about interactions with the link by other users is also added to the pin.
Once pinned to a personal ‘pinboard’ by a user and labeled— for example some of Sharp’s links might be labeled, “modern architecture in London”, images are not only easily managed and used by Sharp/the person who posted, but are now categorized in a way that makes them useful to others. Any user interested in London or modern architecture could now access Evan’s organized view of his architectural images. As well, it made these web objects attributable because they link users to the source website, rather than downloading the images.
Pins can also be original images
Users can post original images to Pinterest, for example photos of their own, which don’t necessarily need to have a link or url. Examples might be user photos of products, which will include product information in the comments section.
Key elements of a pin:
An Ocean of Human-Indexed Content
The contents of Pinterest are a vast store of links and images, all of which were added by humans. Web crawlers index everything on the internet; Pinterest contains only items, often objects from websites, which at least one human is interested in. Evan Sharp called Pinterest, “this crazy, human indexing machine” and maintains that Pinterest is one of the four ways to find things, the others being Google search, Facebook, and Twitter.
I don’t remember exactly when we were like “Holy crap! Pins aren’t just images. They are representations of things and we can make them rich and we can make them canonical and link back to the best source and we can attribute this properly to the creator.” — Evan Sharp, Pinterest co-founder
The Impact of User-Created Content
While not explicitly part of Pinterest’s data architecture, links to popular social media have been added to make pins both more accessible, and more integrated with popular social media. Pins can be published simultaneously to Pinterest, Facebook, and Twitter by Pinterest users.
Pinterest has also enabled commercial websites to connect more easily to Pinterest. “ Pinterest-infused” means sites — typically commercial ones — offer links to enable end-users to easily add site content of interest to Pinterest. In this way, Pinterest becomes a vehicle for end users to expand access to commercial website information. This can give tremendous credibility to the information (typically product data) added; it’s user-endorsed, which helps drive purchasing decisions. In November of 2016, Hootsuite reported that 87% of Pinterest users have purchased a product because of Pinterest.
Pinterest invests heavily to keep this commercial impact somewhat honest. Their crowdsourcing technology breaks users into three categories: “spammers”, whom they work to block, “triers” whose posts are not always of the best quality, and “producers” whose posts product quality ratings, and whose content they amplify.
Statista, Number of Monthly Active Pinterest users from Sept 2015 to Sept 2017 (in millions), https://www.statista.com/statistics/463353/pinterest-global-mau/
Hootsuite Blog, November 2016, The Pinterest Statistics that Matter to your Business, https://blog.hootsuite.com/pinterest-statistics-for-business/
Pinterest Engineering on Medium, Scalable and reliable data ingestion at Pinterest, May 19, 2017, https://medium.com/@Pinterest_Engineering/scalable-and-reliable-data-ingestion-at-pinterest-b921c2ee8754
Pinterest Engineering on Medium, Do you trust the crowd? 3 ways to improve crowdsourcing at your company, https://medium.com/@Pinterest_Engineering/do-you-trust-the-crowd-3-ways-to-improve-crowdsourcing-at-your-company-6f19df729a89
Recode, Pinterest expects to make more than $500 million in revenue in 2017, https://www.recode.net/2017/3/21/14991260/pinterest-advertising-revenue-500-million-growth-ipo
Atlantic interview with Evan Sharp, co-founder, July 2014