Project Moneyball: An Oral History Of How Yahoo Built A Native Ad Platform In 43 Days

A tale of 18 very busy people, an ambitious goal, free trips to Hawaii, and a new $300 million business.

Project Moneyball: An Oral History Of How Yahoo Built A Native Ad Platform In 43 Days
Marissa Mayer, CEO of Yahoo! Inc. [Photo: Simon Dawson, Bloomberg, Getty Images]

When Marissa Mayer was named as president and CEO of Yahoo in July 2012, she arrived at a company that had long made most of its money from traditional display advertising: big, splashy banners plastered on web pages.


It was a business in decline. And it was in decline in part because “native advertising”–ads that take on the format of the content they accompany, such as one on a Tumblr blog with a blog post-like presentation–were on the rise. To take advantage of the native-advertising trend, Yahoo would need technology to sell and display such ads. That infrastructure had not been built under Mayer’s predecessors.


To date, one of Yahoo’s most significant accomplishments in the Mayer era is the creation of Gemini, a native-advertising platform that made $300 million in revenue in 2014, helping the company reach $1.1 billion for the amalgam of areas it calls MaVeNS: mobile, video, native, and social. But before Gemini was a big business, it was an 18-person, 45-day skunkworks project codenamed Moneyball.

As part of my research for the May 2015 Fast Company cover story on Yahoo’s mobile initiatives, I spoke with Mayer and Yahoo vice presidents Enrique Munoz Torres and Rohit Chandra about the Moneyball project.

Going Native

Past Yahoo advertising-infrastructure projects had tended to be enormous in their ambition, including a particularly overwhelming one that Chandra worked on after joining the company in 2004.


Chandra: The project codename internally was Panama, and it was essentially a rebuild of our search advertising stack. I got pulled into that relatively early, in June 2005. And over the course of the next 12 to 18 months, that effort grew in scope. It eventually became, probably, a 300- or 400-person effort, for all aspects of the advertising stack, to rebuild and be really competitive in the marketplace.

This time around, Mayer knew that Yahoo needed native ads fast, and saw a way for the company to get something up and running without devoting vast resources to the effort right away.

Mayer: We saw this trend coming up on all the different social networks. Ads showing up in streams. When I looked at it, I said, “You know what? They’re not actually different than search ads.” A lot of social networks are having to build those databases of ads from scratch. We already have our search ads. All we need is permission from various advertisers to take the various campaigns they’ve already carefully crafted, and already carefully priced and budgeted, and be able to run them in a display context.


That’s not exactly what they ended up doing. But they basically said, “Look, we can actually take this amazing set of ads. We can get millions of ads. And get them to the point where they’re personalized and targeted incredibly well to each user, and we’re sensitive to the context, basing it off of this great database of search ads, as well as attracting new advertisers to the platform. I said, “Given that we already have the ads, we’ve got to get it up and running. Figure out what they look like, build it, have a way of onboarding advertisers. But let’s see if we can get it done. Let’s see if we can get it done in six weeks.”

Chandra: Panama was a big, large project, with operating teams and steering committees. With [Moneyball], we had the luxury that we could start from scratch. That’s very much how Marissa likes to approach things. She’s not fond of big, large projects.

To help make Moneyball happen, Mayer called on Torres, a former Google colleague who had recently joined Yahoo without knowing exactly what he’d be working on.


Torres: I came to Yahoo a little over two years ago after being at Google for eight and a half years. And at Google I joined as an associate, pretty much right out of college. It was Marissa who recruited me. When I heard that she was coming from Google to Yahoo, I immediately shot her an email saying, “It’s not like I’m unhappy at Google, but if at any point you think that we should talk, let’s talk,” and eventually, she got back to me and told me about some of the exciting things that were going on at Yahoo and that’s what piqued my curiosity.

From the beginning, the notion with Moneyball was that it would involve a small group of people doing something really important in very little time.

Mayer: We said, “OK, thank you for signing up. We’re going to ruin the next 12 weeks of your life. You’re going to work weekends, you’re going to work late nights, we’re going to try really hard to get a product launched in six weeks. And you’ve got six weeks to clean it up. And then you can go back to whatever you were doing before. And if you do it, we’ll give you a nice spot bonus. And by the way, if that product ever reaches a $100 million run rate, we’ll send you all to Hawaii with your families.”


Chandra: Marissa called a few of us into a room and said all these people were pitching her proposals: “Hey, I have all these cool, new ideas, and if you give me 75 engineers, in 12 to 18 months, we can go build this.” And she said that this kind of goes against her instincts. Her instinct is that the way the best things get done is to get 10 engineers into the room and tell everybody else to get out of the way.

At the start, nobody was sure where the effort might go.

Mayer: When we started out on the project we weren’t sure. This might just be something like, “Well, that was cute, we made 3, 4, 5, 7 million dollars. And now that’s over, and everybody back to their projects.” Or we might start a real franchise that’s a key part of our future, which is ultimately what happened instead.


The project, which officially kicked off in March 2013, was dubbed Moneyball, after the 2003 Michael Lewis book and 2011 film adaptation starring Brad Pitt, both concerning Oakland A’s manager Billy Beane and his hyper-efficient, data-driven approach to managing a baseball team.

Torres: I saw the Moneyball movie on a Saturday. This sounds completely made up, but I promise that it isn’t. On Sunday, I checked my calendar and there’s some meeting called “Moneyball Intro” for the next Monday. And I show up having very little context, and that’s when I learn that I had been asked to join a team that was going to be tasked with working on mobile advertising. It was a small team, mostly engineers, a couple of product managers including myself. And we’re given the task of building a product that could monetize our homepage screen, which was a recent redesign at the time of the Yahoo homepage and to do so within 45 days. That’s where the real fun started.

Mayer: They drafted 18 people, total. And I said, “This is going to be one of the most important things we try and do this quarter. So you get to designate a piece of clothing that, when you’re wearing it, anyone at the company works for you.” So they named themselves Moneyball, which we launched as Yahoo Stream Ads and later went on to become Gemini. They got an Oakland A’s, Billy Beane-signed hat. And anyone who was wearing the Billy Beane-autographed Oakland A’s hat was in charge.


Old Habits, New Approach

As the Moneyball project began, the team had to unlearn some long-term Yahoo habits.

Chandra: It’s possible that in our previous culture, we would have said, “OK, fine. Let’s have a product manager sort it out and prioritize. And then we’ll do a little bit of everything.” That’s quite often everyone’s rational, instinctive response. But I remember in one meeting, Adam Cahan, who’s running our overall execution, was in the room. He said, “I think Marissa is asking us to do something quite different.”

We happened to be on the same floor where Marissa sits. He walked down the hall and asked her, “Hey, the team is having this debate. Can you come in?” And she came in and was very crisp: “Yes, there are all sorts of issues and considerations. Over time, we’ll tackle all of them. But please, take the mindset: build something super simple.”


We were given the charter to run like a startup, and we embraced it. I had to get rid of some old habits, but I think that’s been the best part of it.

Torres: Obviously, [Mayer’s] purview has changed and evolved over time, but I think that speaking strictly from my perspective that the way that she approaches problems has not changed in a significant way. Moneyball is an example. It started from her being very clear about which problem she wanted us to solve and setting up a set of parameters. Then we started building stuff, engaged with her on an ongoing basis, and ended up shipping something that was good.

One of the keys to Moneyball’s success was that far more than 18 people thought of themselves as members of the team.


Torres: While we were working in a war room, we were not by any stretch of the imagination a secret project that nobody knew about. At FYI, which is the weekly company-wide gathering, Marissa and myself explained to the entire employee pool what we were doing and why we were doing it.

It turned out that other Yahoo staffers supported the effort, no green baseball cap required.

Torres: Normally a sales team would prefer to have more than a week to prepare. But they were like, “No, we get it. We’re excited that we’re building something new and we’re going to go make it happen.” So, yes, it was challenging and it was hard, but at the same time I cannot think of anything that Yahoo could have done to make it easier for us.


In the end, the Billy Beane’s A’s cap only mattered once, but it did its job.

Torres: The one time that we used the hat was with Marissa. We needed a decision on what we were going to be calling Moneyball externally and it had been challenging to get all the stakeholders in a room to make that decision. I think that they were preparing some kind of board meeting, I put on my silly hat, I walked into the room, knocked on the door, and said, “Hey guys, we need a name for Moneyball.”

Mayer: We stopped the meeting. We had a debate on the name. It took about five minutes. And we said “OK, call it Yahoo Stream Ads. When it gets to be more of the vision of what want it to be, with mobile search also included in the marketplace, we’ll then relaunch it as Gemini. But for now, we’ll use the placeholder name Yahoo Stream Ads.” And he was on his way.


Torres: I mean, the green hat story is kind of interesting from a storytelling point-of-view, but I think what’s most remarkable is that we didn’t really have to use them.

Gemini and Beyond

The goal had been to have a working native-ad platform in 45 days, a task that the Moneyball group accomplished in 43 days. But the deadline was a beginning, not an end.

Mayer: They spent six weeks cleaning it up, because whenever you launch a product that fast, you hack things and put things off til later. And then we said, “OK, you guys held up your part of the bargain. You got it launched, you cleaned it up. As you were. Go back to your projects.” And they were like, “Um, we don’t want to go back to our projects. We think this is gold. It’s going to make our products better. It’s going to make our ads better. It’s forward leaning. It’s the way people are going to want to get their messages in front of end users. Please don’t send us back to our projects. We want to have this continue.”

As Moneyball became Yahoo Stream Ads, and then Gemini, it grew rapidly in terms of team size, capabilities, and revenue.

Mayer: What was 18 people suddenly became more than 100 people. And Yahoo Gemini crossed their $100 million revenue run rate. They had to make $273,000 in a single day. They launched on April 29th. They crossed it on August 25th of 2013. In less than four months, they had a $273,000 day. And we sent the original 18 to Hawaii–over Christmas, so it actually wasn’t time away from the office. Which by the way was their choice, not ours.

We measured them at first, on that first $100 million, as revenue run rate. Last year, it was $300 million in revenue, which means that our exiting revenue run rate was even on the high side of that, given that they came in at a much lower revenue run rate than $300 million at the beginning of the year. And so I’m very, very proud for them, of them, and what they did. But you couldn’t look at it and say: foolproof.

Even though the Gemini effort is now vast, it’s doing its best to preserve the Moneyball mindset.

Chandra: I keep telling myself that one day the project will become big and traditional, and then maybe it will be time for someone else to run it. We actually try as much as we can to preserve a culture where it’s as flat as possible, the decision making is as streamlined as possible. In the early days, especially, when it’s a nascent project, it’s important to take risks.

Torres: I don’t think that it’s far-fetched to say that Moneyball was very successful. It was also very public within Yahoo. The notion of us quickly building something that worked well and was hopefully a positive influence over the entirety of the company and the basic tenets can be applied in places outside of advertising.

Important though Moneyball was, it’s not a template for every big, ambitious effort Yahoo will tackle from now on.

Mayer: We did a similar draft for [Yahoo] Mail last year, to build out some core features and make some key improvements. But you can’t do more than one or two drafts a year. Otherwise, everyone spends all their time drafting everybody and trying to talk them into working on their project, and nobody gets anything done. So you’ve got to have the drafts be really important and really surgical.


About the author

Harry McCracken is the global technology editor for Fast Company, based in San Francisco. In past lives, he was editor at large for Time magazine, founder and editor of Technologizer, and editor of PC World.