Pins at Pinterest: How Pinterest is using Machine Learning to Cultivate Inspiration

Introduction: Pinterest

Founded in 2010 by Ben Silbermann, Evan Sharp and Paul Sciarra, Pinterest is an online social platform that allows users to express creativity and define future aspirations through “pins” on a virtual board. These pins can encompass a wide spectrum of categories from DIY projects, travel, recipes and workouts that are organized on the user’s specific category boards. “Every idea is represented by a “pin” that includes an image, a description, and a link back to the image’s source online where [the user] can learn more about the idea.” [i]

With 250 million monthly active users, 175 billion items pinned on 3 billion virtual pinboards, Pinterest is a treasure chest of data that the company is constantly combing through to provide the best tailored experience for the user through pin recommendations. [ii] With the help of machine learning, Pinterest identifies content that resembles previous users pins and recommends that content to the user. The algorithms help provide inspiration to the user that he or she may have never initially searched or pinned. For example, if I pin a picture of hotel in Tahiti, the algorithm will automatically filter my home feed with pictures of similar hotels around the world.

Machine Learning at Pinterest

“With 100+ billion human-curated ideas, Pinterest is the biggest image-rich data set ever assembled. This lets [Pinterest] do interesting things like analyze trends, understand intent and predict consumer behavior.” [iii]

Machine learning is critical to Pinterest’s core business: a curated bulletin board of inspirational content. The ability of Pinterest to identify pins that the user doesn’t even know they will like is critical for Pinterest’s long-term success. The tailored content not only allows Pinterest to increase user engagement and customer retention but also enhances Pinterest’s recommendation power. The machine learning models are only as good as the data they receive so as the user becomes more active (i.e. the user pins more) the model can better suggest related content that would be beneficial to the consumer.

“The more people Pin, the better Pinterest can get for each person, which puts [Pinterest] in a unique position to serve up inspiration as a discovery engine on an ongoing basis.” [iv]

Pinterest dedicates most of its operations to the constant iteration of their machine learning models. Pinterest’s internal group, Discovery Team, has made numerous updates to the model over the past couple years to refine the prediction capability. For example, in 2015, Pinterest’s Discovery Team launched Pinnability, an enhancement to their machine learning models that estimated the relevance score of a user interaction with a specific pin. [v] This new technology enabled Pinterest to prioritize pins with high relevance scores at the top of each user’s home screen. In addition, Pinterest acquired Kosei, a machine learning company that specialized in content discovery and recommendation algorithms to further enhance their machine learning recommendation model. [vi]

Pinterest not only utilizes machine learning for their pin recommendations but also throughout their entire organization to run a more efficient business. For example, the black ops team uses machine learning to detect spam content, the monetization team uses machine learning to develop ad performance and the growth team uses machine learning to prevent user churn. [vii]

 

To facilitate continued growth in the short term, Pinterest needs to make sure they are constantly iterating on their machine learning models to better predict and recommend beneficial suggestions to their millions of users. Pinterest needs to continue to invest large amounts of capital to engineers and strategic acquisitions that can help enhance their internal machine learning models and provide additional data to their data set. In the long term I believe Pinterest needs to identify a new technology that can be incorporated into their business to further increase user engagement. One idea is an app on a phone where users can use the phone’s camera to upload photos directly to their board or an artificial intelligence component to the online and mobile platforms.

Open Questions

Is this business scalable in the future? How can Pinterest continue to improve the machine learning models to be better predictors of human needs? With the amount of data Pinterest collects should users be concerned about personal privacy and how can the company make sure they do not cross the line?

Word Count (722)

Sources

1.       Mark Wilson, “The World’s Most Innovative Companies 2018, fastcompany.com, March 2018, https://www.fastcompany.com/company/pinterest, accessed November 2018.

 

2.       Erin Griffith, “Pinterest Is a Unicorn. It Just Doesn’t Act Like One,” nytimes.com, 2018, https://www.nytimes.com/2018/09/09/technology/pinterest-growth.html, accessed November 2018.

 

3.       Pinterest Labs, “The latest in AI & Machine Learning at Pinterest,” labs.pinterest.com, https://labs.pinterest.com/projects/, accessed November 2018.

 

4.       Yunsong Gup, “Pinnability: Machine learning in the home feed,” medium.com, March 2015, https://medium.com/@Pinterest_Engineering/pinnability-machine-learning-in-the-home-feed-64be2074bf60 , accessed November 2018.

 

5.       Ibid.

 

6.       Michael Lopp, “The future of machine learning at Pinterest,” medium.com, January 2015, https://medium.com/@Pinterest_Engineering/the-future-of-machine-learning-at-pinterest-88e6d4bf1968, accessed November 2018.

 

7.       Ibid.