Marissa Mayer works as director of consumer Web products at Google and has spearheaded almost every user-interface change to the Google Web site for the past four years. She also teaches computer programming at Stanford.
In her session at SXSW Interactive, Mayer shed some light on Google’s flexible, dynamic, and fairly unstructured product development process. She shared how the company fosters employee-involvement in product definition, the role of creativity, and the resulting products and services. What follows is a partial transcript of her talk.
I want to go behind the scenes at Google. How do we design something? How does that happen? I’m going to give a lot of examples. Let’s start with the Google ingredients. Our mission is to organize the world’s information and make it universally accessible and useful. This mission is very open ended, even in terms of search. There’s a lot of things we can do that falls under this mission. One example is Google News. Yes, there’s a search aspect, but it’s an extension of the mission.
We also like to do things that matter. How many people will use it? What particular problem will it address? We want to ensure that our efforts are spent in such a way that people will use the end product. We laser focus on the user. What do they need?
It’s also important to focus on smart people. Brilliant people have good ideas. Today is Einstein’s birthday, so there’s a bit of an Einstein theme to my talk. It’s important to take very smart people, put them in a creative environment, and see what happens. These are scenes from what we call the Googleplex. It’s important to have a blend of fun and home life in the office. We incorporate aspects of comfort in the office. People bring in their children for dinner. We have laundry, and some people get used to doing their laundry at work.
We also use a process that lets ideas really thrive. Ideas come from everywhere. Some people think that ideas come from the top down. Some thing they come from the bottom up. They come from everywhere. We have several active email lists. We have an email list just for ideas. We also have an all-company list called Miscellaneous. It’s odd for a company of 200 people to have a miscellaneous mailing list. Now we have 1,000 people, and sometimes the list can get out of control. We might debate whether we should have bagels, but we also discuss some really important projects. Our employees are our most critical users.
It’s also important to design for users. We look at the user perspective, not at the how can we make money perspective. If there is a commercial element, we make sure that it’s really visible. We try really hard to look at products in such a way so they’re how you want them to be laid out. This is the original Google Web site. I’m often asked why we don’t put advanced search functions on the home page. People come to Google with a specific task in mind, what they want to do. So there’s only one box. Our founder created the original page just because he didn’t know how to make a fancier page, but It worked out. It wasn’t a marketing move, it was just dumb luck.
How do we take those ideas and make them an end product? First you have to compile all of those ideas, discuss them, and prioritize them. You can only get so many things done. This is a tool at Google called Sparrow. It’s a typical Web page, but it allows in-page editing. These are basically project ideas. There are also little widgets. They pop up a form, and you can add new ideas to the list. We literally have hundreds if not thousands of project ideas that we consider and prioritize. If you make the capturing of ideas simple and low cost, a lot of people will share ideas.
We also have discussions about ideas. What second order effects will this have? We actually create a Google Top 100 that looks like this mock Billboard list. This is how we prioritize ideas. The first time we did this, it was relatively controversial. Larry Page disappeared one night and came back with a list and said this order. People said how dare you! How dare you come back and tell us what the priorities are! He said, fine, I’ll take my priorities away, and you can come up with your own lists. In the end, we all came back with basically the same list. When you start looking at projects relatively, it’s easy to come up with what’s most important. Would we rather have a unified billing system or Google in Korean? Once we have the list, we start staffing from the top. We can staff maybe 40-60 of the ideas.
Isn’t 100 projects a lot? What defines a project? Sometimes there are small projects. Google News took a team of 3-5 people. If you take a pool of engineers and put them in teams of three, you can actually do 100 projects. Small teams can self-organize quickly, they can finish projects quickly, and they can move onto new projects quickly.
Communication is key. When you have so any engineers working on so many projects, you need to have tools that help you self-organize. One tool I mentioned is Sparrow pages. We use Sparrow pages all over our intranet, and not just for project ideas. We also use them for our launch calendar. We also use them to communicate project status. In addition to the core group of people who organize these pages, people working on other projects can also easily find these pages.
We also do something called Snippets. Those are brief weekly reports from different teams. They give an update on what they’re working on, and it’s a way to help us self-organize and find efforts that are correlated. And with the purchase of Blogger this year, we’ve found another way to communicate. People like Evan Williams have personal blogs. We also have blogs for teams and blogs for projects. Blogs have been a really interesting addition to our use of Sparrow pages.
Once we do all this work with ideas and communication, it’s important to test, experiment, and iterate. We tend to test a lot, both in terms of QA but also in terms of usability. We do usability tests that not only find usability problems with the Web site but also elicit user opinions on the Web site. In January 2000, we went to Stanford University, and we had students use the computer in pairs. They talked to each other more than they talked to us. It was the first time anyone had gone to see how the Web site was being used outside of the company. We gave people a trivia question and asked people to search Google for answers. We asked people what country won the most gold medals in the Olympics. People called up the page, 45 seconds went by, and we weren’t really sure if we should intervene. Response was unilateral: “I’m waiting for the rest of it.” People weren’t used to a blank white page. They were waiting for the flash and the images to load. That was a usability problem, as well as a perception problem. So we decided to add the copyright notice at the bottom of the page. That’s not there for legal reasons. It’s there to punctuate the end of the page. That’s all there is. This is it. Please start searching now.
We also learned some other interesting things. People had comments like, “Is this some guy’s home page?” Another guy became concerned with the legitimacy of our company and asked how many people worked there. At the end of the study, I said 83, and her follow-up question was whether we were from the psychology department. People didn’t understand the cache page.
We also like to experiment. We like to release early and often. Once we’re relatively happy with something inside, there’s no reason to wait on launching something. The easiest thing to do is put it out there to the public, listen to the feedback, and see what people think is important. This is Google Labs, and it’s where we put out things that are in alpha. We can immediately see what the response is to a particular idea. It’s a great tool for us to get early feedback on projects.
What’s come out of Google Labs? The Google Desk Bar. That’s the thing we have at the top here that fits into your browser. The Google Tool Bar sits on your desktop as a search app all the time. They’ve actually rendered the IE engine so you don’t even need to launch IE. You can just do a search, and a little window pops up with your search results. We also released search by location, and it’s not something Google’s very good at right now. Here you can see things plotted on a map. What types of radiuses do they want? What kind of queries do we get? If we aren’t good at doing something, we don’t get good queries. One thing we can do is release search by location early to get feedback on features — but also good test cases. News Alerts also launched off Google Labs. You can enter your email, the search you’d like to do, and get everything that’s published about a specific topic, and get it emailed directly to you.
We call all of these projects Googlettes. Basically, they’re baby Googles, startups within the startups. We really like the concept because it helps us keep that innovative startup feel and launch projects quickly. Another example is Orkut, which is named after one of our core engineers. It’s our social network project. We have something we call 20% time. Orkut’s social coordinator for Google. He’s always rallying us around causes and taking us clubbing. But because of his social network interests, Orkut dedicates his side time, his 20% time, to work on whatever he wants to work on. Orkut started working on building this social networking site. He sent a mail out to the Miscellaneous list, and within hours, we had 1,000 employees signed up. This really sparked people’s imagination, so we decided to launch it publicly. We’re now at more than 150,000 members. We’ve also seen some interesting demographics. We see a nice cross section across the entire population. It’s a testimony to how quickly we can innovate.
One question is: Does the Web really change everything? It’s a pretty radical change compared to R&D cycles of brick and mortar companies. I joined a company with 10 of my friends, and now I travel all over the world. The Web allows for much faster distribution and feedback. It does help innovation happen faster. That said, there’s a certain amount of hard work and stuff that it really doesn’t accomplish. I would like to be able to talk to the search engine and call it. We still have some way to go.
My final thought is that it’s important to stay true to your mission. We need to stay true to search but also find other things that relate to it.