Pinterest’s revenue chief talks tech priorities for 2023

What’s the trend forecast for 2023? According to Pinterest, we’ll see an uptick in sci-fi fashion, mushroom decor, and—from its corporate perspective—combining machine learning with user curation to continue to fuel Pinterest’s recommendations.

Last year was tough for tech platforms, with companies from Alphabet to Amazon missing analyst expectations, enacting cost-cutting plans, and reducing headcount. For Q3, although Pinterest’s growth slowed year over year, the company’s earnings report served as a rare bright spot after it beat analyst expectations for both revenue and earnings per share.

Bill Watkins, Pinterest’s chief revenue officer, said that this year the company will be focused on user growth and engagement, shopping, and ultimately monetization. At CES 2023 in Las Vegas and in a follow-up phone conversation, we spoke with him about how tech investments will power those strategic goals.

This interview has been edited for length and clarity.

Tell me about Pinterest’s 2023 priorities.

At the enterprise level, we have three core focus areas this year: user growth and engagement, shopping, and monetization…A majority of our Pinners are telling us that they come to the platform not just to get inspired, but also to shop. We’ve got nearly half a billion users around the world, but I think just as interesting, we have hundreds of billions of pins, or objects, that have been organized into single-digit billions of collections, or boards—these are products that people are going to go buy, or things that they’re going to go do, in their real lives. At that scale, when a majority of our Pinners are telling us they’re coming to shop, it’s our responsibility to make shopping ubiquitous on the platform.

In monetization, I’m having my team focus on [a few] key areas. The first is that it’s my aim for the team to be the industry-leading, full-funnel advertising solution…and the second is advertiser value…One example of how we’re investing in first-party solutions is our conversion API. If we think about the future…and where the industry is going in this tagless and cookieless environment, we think conversion API is a good first-party solution.

For those three priorities, what are the specific tech investments you’ll make to help achieve those goals?

We will absolutely continue to reinvest in computer vision. Basically, anything that can train off of the dataset that we have and that can render better results for our Pinners, we will be investing in, heavily, into next year. Because it will help further all three of our enterprise-level investments in user growth and engagement, shopping, and monetization.

Asked in follow-up: Can you elaborate on Pinterest’s computer vision goals?

We largely use computer vision to show our users recommendations, be that for the next outfit they want to buy, the next meal they want to cook, the next place they want to travel to. We’re also using computer vision to show these highly relevant, highly engaging, and oftentimes highly shoppable recommendations. When I think about investments in computer vision, I really think about three primary sub-investment areas within that category: The first is to understand everything in an image…so we use that technology to really understand and then tag everything in a given image. Second is to help people select anything in an image online, help them find related ideas, and go buy those things…and then third, is to use the camera on your phone to search in the world around you—using your camera, if you will, as a query…and then for us, we take that, understand what it is, and then render results based on computer-vision technology…At the end of the day, we’re using it to supply the most relevant, inspiring, and shoppable recommendations to our users, because that’s what they want.

Asked in follow-up: How does Pinterest’s recommendation engine set it apart from other platforms’ algorithms?

Anybody can go out and build, buy, or borrow machine learning technology, but it’s only as good as the dataset that it trains on. When you think about our unique dataset at Pinterest, we have hundreds of billions of pins, or objects, that have either been brought from the web into our walled-garden ecosystem or repinned within Pinterest. The pins have been collected into about six billion boards, or collections. Our users have organized all of the pinned content that we have into [those] collections, and that gives us a pretty good understanding of how these objects, which are largely products, services, or experiences—things that people really want to weave into the fabric of their lives—give us insight into how they do that today. But from that dataset, it also allows us to predict what someone might be interested in, or it might allow us to predict things that are trending…That pin and board dataset that our machine learning trains off renders to all of our more than 400 million global users the most relevant and engaging content…We want to serve it in a highly relevant way, based on how the machine learning actually delivers the recommended content, and make it shoppable. That’s how we think about that. The machine learning itself is based off of the pin-and-board construct, that robust content, how our users are collecting it, how our users are behaving against all of this content, and strategically, our aim of having the content and recommendations not just be inspiring but also shoppable—a bias toward actionability.