Facebook's Margaret Gould Stewart on product design at scale

Currently Director of Product Design at Facebook, Margaret oversees UX for all of the company’s ad products (mobile alone is a $2.9-billion business). She previously held senior user experience roles at YouTube and Google, and you might have caught her giving a Ted Talk or two.

I caught up with Margaret to talk about the pressure of designing products for more than one-sixth of the world’s population, where design concepts differ and overlap in consumer and business products, and how to maintain a high bar for quality while scaling a creative team.

If you like what you hear, check out more episodes. Or subscribe to the Inside Intercom Podcast on iTunes, Stitcher, or over at Soundcloud.

If you’d prefer to read Margaret’s insights, a lightly edited transcript of our conversation is below.

Des Traynor: For the sake of our audience could you briefly introduce yourself?

Margaret Gould Stewart: I have been at Facebook for a little more than three years. I’ve been in high tech and designing large scale platforms for most of my career. I previously led design at YouTube and at Google for search and consumer products.

Creating A Design Fit For Billions

Des: Recently Facebook announced it had a day with a billion users. That must intimidate the hell out of a design team. You’ve said one of the requirements of working at that scale is having a sense of audacity – you believe you can design for that many people – but also a sense of humility because you have to take yourself out of the design process. How do you balance these?

Margaret: It’s a really big challenge but it’s an exciting one. It’s why I’ve invested so much of my career in this kind of work and I think striking a balance between those two things is the key. If you’re all audaciousness then you tend to lose sight of what matters, which is the people that you’re designing for and the impact that you’re having on their lives. If you are all about humility and you don’t have some risk taking and audaciousness in your approach, you may not actually grapple and tackle the big, big challenges, questions and problems that are going to make a big impact in the world and in society.

Having that combination is critical to this work. It’s something I certainly look for in designers who become a part of my team and I also try to provide the opportunities for people to cultivate that balance over time. It can be quite an intimidating prospect, on day two of work, to start work on surfaces that potentially millions, if not billions, of people are going to interact with.

Des: Are you trying to encourage some designers to be a little bit braver and at the same time trying to encourage other designers to realize it’s not just them?

Margaret: Often it’s leveraging experiences where designers experience what it feels like to use products as everyday folks, whether business users or consumers, and take themselves out of their own experience. That’s what creates the humility and the empathy but also encourages people to think big and bold about some of the problems they can solve. It’s about respecting both aspects of the work and making sure a team has, as individuals but also as a team, a good balance of that. It’s also collected in our design process. We are a company, and many companies that are in this space are full of super smart and creative and highly motivated people, people with a lot of innovative ideas about what they could do, what they could build.

There are a lot of ways that you can gather data and bring the people you’re designing for into the design process so that it’s not just about you and your ideas. It’s also about looking outside of your own experience to enrich your perspective and understand what problems are worth solving. And what’s the most effective way to solve it on behalf of the people of the world.

Des: Is it even possible to understand a billion users? When you think of all the standard UX techniques like user journey, storyboards, personas, do any of them work at the scale you design at?

Margaret: It can be challenging because at some level you’re designing for kind of the whole human race.

Des: Right, no pressure.

Margaret: I don’t know what more audacious goal there could be. The audaciousness is not just the way in which you engage in the process but also saying, could my work be relevant to that many people? It’s kind of an audacious notion but I think we’ve seen with Google Search that it does have that impact. People are that much in need of finding information, understanding the world around them, that it’s a universal human need. With YouTube it is a universal need for people to creatively express themselves, to educate themselves and other people. That is an enduring human need, and with Facebook perhaps the most enduring human need of all, to feel connected and loved.

These are kind of audacious notions that you can use technology to address but those are three examples where there’s been a huge amount of success in creating something that seems to have almost universal relevance and value to people. Then the trick is, getting back to your question, how do you understand that enormously diverse population of people, what aspects are universal, and which actually need to take into account a particular context in which people are living and working day to day. That can differ quite dramatically. User research, obviously, is completely critical to everything we do. Data analysis helps us understand the macro trends and not skew too much toward the qualitative. This is all about balance – bringing a variety of different data sources together to get a holistic understanding of what’s going on and how to respond to it.

Getting to Know A Global Audience

Des: My experience with user research has been, well we’ll talk to like 30, or 60, or 90 people. That seems justifiable when you’re talking in thousands or even tens of thousands. What does user research look like for you?

Margaret: It dramatically depends on the context. With an existing product where there’s a ton of usage there’s going to be a particular approach to that. If we’re developing something new from scratch there’s going to be a different approach. If you have a given interface and you’re looking at a tactical level, say can people understand how this flow works and from a usability perspective get through it, the same rules apply. Seven to 10 well recruited people can actually help you understand whether or not a particular task is understandable and doable by the vast majority of people.

In terms of the value that you’re providing and people’s likelihood to access it and understand it, there are a lot of regional differences and cultural differences that can impact people’s use of a product. Some of the most impactful research we do is going into particular countries and communities to ask e.g. what is the nature of how small businesses work in India, or Indonesia, or in Brazil? What are the challenges that people have in transacting in those environments? Or in connecting with friends, what are the social norms around connecting with friends and family that may differ dramatically from culture to culture? A lot of it is more on the early stage, almost ethnographic side of things. Then we do a ton of validation work. You have to when you have so many people relying on us to get these things right. We don’t release anything without it being tested nine ways to Sunday because that’s the reality of the responsibility of what we do.

Des: When you read design articles and design blogs, is it appallingly hard for you to take anyone else’s work seriously when you’re like, “Yep, I’m sure that works when you’ve got 15,000 users?” Is there anyone you can actually look to for insight?

Margaret: Oh no, absolutely. I think there’s so much to learn in so much of the work that’s going on across the industry whether it’s small scale or large scale. Don’t forget, we invent new products that start out with a user base of zero. So we have a lot of experience growing and scaling things quickly. We have certain advantages to having a brand that is very well known and can help jump start a project. But there’s a big difference between the work that the Instagram team did in designing and releasing something like Hyperlapse versus the team that works on refining the experience of the Facebook news feed, right? Those are dramatically different design contexts, so it’s not like all of the different projects work in the same way, even within the company of Facebook, and that’s true for other companies like us.

There’s a lot that we learn from smaller companies and projects that are working at a smaller scale. I also think it’s extremely useful for us to see the level of experimentation and innovation that goes on that doesn’t have the constraint of millions of users. There’s a certain care we need to take in introducing too much radical change in a context where there’s so much current usage. We’re always excited to see people experimenting at early stage because it’s exciting to see what people do from a blank slate. It’s a different context, but I think it’s still very valuable from a design community point of view.

Users Can Lead Product Evolution

Des: You said you’re a big believer in not just designing for people but also designing with people. What does that mean to you?

Margaret: It gets back to that humility thing. As a company, and I think most smart companies in the high tech industry operate this way, you don’t believe that all of the good ideas exist within your own company, right? A lot of the genius of the world exists outside, even though we have a ton of smart people. Just from a percentage point of view you know there’s going to be a lot of interesting ideas. That’s why our API and third-party development community is so important to us: great ideas come from outside of Facebook. I think the idea of designing with people and not just for people is recognizing that even the consumers that use our products are an active, contributing force in our design process.

When we release something and we have lots of people using it we take a lot of time to observe that usage and to learn from it. There are so many examples of people taking a product that was designed to do X and they take it in a totally different direction of Y. You can sit back and say, “Hey stop doing that. That’s not what I designed it for,” or you can say, “Ooh, wow, let’s see where this thing goes,” and maybe it’s much more valuable than the thing we originally thought. Or just a new use for a technology that has been sitting around for a long time.

A great example of this is that Facebook wasn’t originally designed or developed to become a crisis communication channel. But we observed in many different contexts, most distinctly the tsunami in Japan, people using Facebook as a critical platform to check in and make sure that there loved ones were safe and accounted for.

This wasn’t something that we designed. It was just an emergent behavior that we observed. Then we were able to say, “Well, how can we make that process easier to discover and use,” which is how the safety check product came into being. That was really a product the community designed for itself. You see that with hashtags on Twitter. That wasn’t a thought or experience that Twitter developed. They just observed it and they were smart enough to say, “Oh, you know what? This is actually dramatically improving this ability of the system, so let’s design around that.”

Designing for Business vs. Consumers

Des: You’ve designed software specifically for B2B and B2C. What are the key differences between the two?

Margaret: I think the most fundamental thing is while it’s great for business products to be fun and engaging, that is not the purpose of their use, right? People are trying to get their work done. The notions of a high quality product are pretty different in those contexts. A good example is with consumer products: the amount of time people spend using it is a pretty good metric to decide how you build something that is engaging. You look to see how much time each day people are using this product. On the business side time spent is an anti-goal.

We want people to generate more business value in less time, so we do not track time spent as a key success metric. We track it because we’re interested in understanding from a task level how to shave time, especially on the repetitive tasks, but we’re really looking for value that’s generated by that use. That’s a simple example of a metric that’s pretty different on the consumer versus business side. There are lots of those.

It’s been interesting for me, having spent most of my career on the consumer side, to orient myself to this space and ask in a business context, “What does success mean? What does delight mean?” It’s really interesting to see how that plays out when people, their jobs and their livelihoods depend on them being successful and competent in the use of any given product.

Des: If you’re making a change in design on the B2B side, you run the risk of either breaking somebody’s work flow and just really messing up their day. On the B2C side, you risk facing the ire of an unhappy community. You worked on YouTube’s rating system and that was changed. Can you talk us through what happened there?

Margaret: Change aversion, that’s a topic for an entire discussion in and of itself. You end up becoming a bit of a victim of your own success when you build up a user base for a product that’s meeting a need. It becomes increasingly more difficult to change and evolve that product because people don’t like change. Even if it’s an empirically better design there can be a lot of resistance.

Now, certainly there’s going to be a resistance to bad change. We should always test and iterate, and have the humility to admit when we’ve made a mistake, to retract and revert as needed. Hopefully if you have a really solid testing process the instances where you make a dramatic change that is a wrong change is rare. But even when everything is right people often respond pretty poorly initially. On the consumer side it’s a question of just the sheer number of people that are having to rework habits. We develop muscle memory in our use of digital products so it’s hard when people move things around or the flow has changed, to say nothing of features being removed.

On the business side it’s even more complicated. If it’s a tool that you’re using say six to eight hours a day, that muscle memory is even more intense than it is for something that you may be using for a few minutes a day. I’ve often said people can get very efficient at using bad design, so even when you put something better in front of them it is legitimately frustrating because you have to learn how to use the new version. Being cognizant of that, making sure the value you’re introducing outweighs the cost of them having to re-learn is really important. That’s a balance that you always have to be very cognizant of. Warning people that the change is coming is really, really important so they don’t wake up one day and their world is turned around.

The YouTube rating system is a really good example of this. This was years ago, probably 2009. We were looking at the data around how people were using the rating system. At the time it was a five-star system to rate any given YouTube video. We looked at the data and it was pretty clear: A handful of people were using one star. Almost everyone else was using five stars, and nobody was using two, three and four.

It was clearly telling us, through user behavior, this is an overly complicated rating system. I do not want the cognizant load to have to think through five layers or levels of value. I either like it or I don’t, and mostly I don’t care enough to tell you that I don’t like it. I just want to tell you when I like it.

The thing is is a lot of the video creators, and even the community, had really gotten used to this five-star rating system. They’re quite attached to it, the creators in particular because I think a lot of the social equity and the value of their presence on YouTube seemed to be wrapped up in these star ratings. What we did, in advance of releasing the change to a binary up/down rating system that still exists today, we basically published the data graph that showed the usage and posed the question to the community. “Hey, we’re seeing some kind of interesting things in the data. What do you guys make of this?”

Of course all of the comments on this blog post were like, “Yeah, this is obviously a broken system. You don’t need all of this complication,” and I think people came to their own conclusions. It also resulted in one of my favorite TechCrunch headlines of all time; something akin to “YouTube Has a Five-Star Realization: It’s Rating System Doesn’t Work,” which is quite witty.

Suffice to say the data helped us understand and make a really powerful design decision that was in the interest of the community. The way we approached it showed respect to the community in terms of the pain of going through that change and allowed them to emotionally prepare themselves for it. Of course there were people who were upset, and the standard riots and death threats that result when you work on these kinds of products, but relatively speaking it went quite smoothly.

Des: It would be easy to make it a quick change and just be like, “Duh, it doesn’t work.” But you need to be humble in how you handle the transition, right?

Margaret: Absolutely. Listen, all the people using our products are giving us so much value, especially when you work on these products that are largely user generated content. You have to have enormous respect for the community because you know what, we would be nothing without them.

To just go ahead and make these unilateral changes, without observing both in a quantitative sense through data analysis, but also qualitatively, to find out what’s really driving the behavior, and then communicating it in an honest and open way about the rationality of our decisions, would basically be disrespecting them in so many ways. I think any organization or team that wants to build a fruitful, sustainable relationship with a community has to approach it that way.

Des: I really liked your point as well about even the world’s most positive, obvious, fantastic improvement is still a change and changes come with a cost. I remember we had a situation at Intercom where we, in our words, fixed something and a user said, “I know you guys think you fixed it but I liked the way it was broken.” It’s a good reminder. Their workflow is way more important than the design principles you’re trying to apply.

Margaret: That’s right because again they have jobs and they get to decide if what you produce is more valuable than what they used to have.

I always say, especially if I’m talking to people about their careers and maybe they’re looking to make a change, “You know, you always trade old problems for new ones. You just need to like your new problems more than your old problems.” That’s true in so many aspects of life but it’s true in design too, right? A new design is going to create new challenges. They just need to be better than the ones people used to work with because a new design is going to create that challenge of reworking and relearning. How much cost is associated with that versus all of the new efficiencies they may win over time?

Measuring the True Cost of Design

Des: I think you once said 280 hours went into designing the Facebook Like button. From the outside it can seem like that’s a lot of design consideration into a very small UI component. Is that the wrong way to think about these things? Is pixel square footage a stupid way to think about it?

Margaret: I’ll tell you a funny anecdote about that. I did a TED talk a year and a half ago and I used this anecdote in it. I was rehearsing the speech with another TED speaker. What they’ll often do is pair you up with another speaker so you can practice and get feedback. The person I was paired up with was an astronomer so he’s studying the history of the universe and all this kind of stuff. I go through the talk and I said, “All right, give me your feedback. Anything’s on the table. What resonates? What doesn’t?” He said, “I love the whole thing. I think it’s a great talk. The only piece that kind of fell flat for me was that story about the Like Button because I just feel like 280 hours really isn’t that long.” It’s because he’s dealing in like millennia, millions and billions of years. I was like, “Well, in a digital design context 280 hours to work on a button, that’s a long time.”

The fact is if you agonize over a single, tiny element of a really complex system, that may be a very poor use of your time. Are you really focused on the things that matter and the things that drive usage and good outcomes for whoever you’re designing for? The case of the Like Button, and things like the YouTube Player or the Google Search input box, those are elements and experiences that are seen and used so often every day that every detail of them matters. Every single pixel. Just because of the sheer exposure and usage they have in the world. They’re design problems that are very unique to a very small number of insanely, widely used products. That’s real and it’s because they’re used by millions and sometimes over a billion people. They are translated into an incredible number of languages. They have to work and degrade gracefully in a whole bunch of different kinds of browsers on different kinds of phones.

If you’re going to design in a global context for a huge swath of the world population it is complicated to make even simple details work well for everybody. That is a really interesting challenge that is tied to working on these kinds of products.

At the same time you learn to look at impact in different ways depending on the context. In the consumer space you really do have that volume, millions and potentially over a billion people, and those single elements being really important. I think what I’ve learned in the last three and a half years working on the business side of Facebook is that it’s not just about the number of people using a product but also the impact of the use of a product on people. That’s not exactly the same thing.

There are certain products that not many people in the world see but they impact the lives of many more people. Business products are some of those things. Far fewer people use Facebook advertising products than they do the consumer products, but the use of those advertising tools impact the experience of everyone who uses Facebook. That’s really important when you’re thinking about what problems are worth solving, what are the projects worth getting involved in, how do I prioritize what matters in those spaces. Having a very nuanced approach to thinking about what is “important” and what is impactful.

A friend of mine, David Cronin, who heads up GE software design, told me this story about his team and getting involved in a project to do with the design of software used in the running and management of nuclear power plants. It’s a standard design engagement for GE and I was like, “Oh my gosh! I hope you put your A team on that because that is kind of terrifying.” This is software that is probably seen by hundreds, maybe thousands of people, but the impact of the quality of that software could impact the future of the human race, not to put pressure on David Cronin.

I think as a society and an industry we’re very caught up in the numbers and the most used things and how many people are using it. Really, if we’re thinking about systems like healthcare, advertising and government, all these really complex systems, there are a lot of products that are living in the shadows that may not be used by as many people but impact our lives on a daily basis. I find that really interesting.

How to Maintain Quality at Scale

Des: You’ve been involved in growing design teams at two of the biggest software companies of all time; Google and Facebook. What helps teams thrive as they scale?

Margaret: There are a few things. One is you need to build a culture that cares about quality. I don’t think really strong design teams thrive in a context that doesn’t care about the experience of the people that you’re building for. Everyone cares about doing good work, but I think focusing on creating high quality products and having a culture that understands the very tricky balance between speed to market and quality is paramount. That’s how you attract and retain great design talent: a good pragmatic approach to that balance between speed and quality. I think a huge part of that is having very tight collaborative working relationships with the different disciplines.

One thing I think doesn’t work is creating an insular ivory tower of design. I think strong design teams work in close collaboration with product management and engineering and respect the contributions of the different disciplines, not just to product development in general but to the design process.

Then more generally, and I think this is particularly true for companies that are designing at a global scale, like Facebook, is diversity within the team. Diversity along the lines of gender, race, background, all kinds, because a lot of tech companies are looking to have a global impact but our work force does not reflect the world population.

A lot of white tech workers in the Bay Area are very smart and well intentioned folks but the lives they have led and the context they bring to their work is limited in that sense. Hiring the most diverse team possible and making sure that we leverage that diversity of point of view and life experience is critical to us in terms of doing the work that we want to do and having the impact we want. In that context design really thrives because if our goal is to change people’s lives in a positive way we need to be designing in a human-centered way. That diversity is what allows us to really design for real people all around the world instead of just for ourselves.

Des: If you’re a design team of two and you grow to 10, then 50, then 100, should you be persistently asking the question, “does the makeup of our product team match the makeup of our user base?”

Margaret: You have to hire the best talent possible, and I think what is critical is to try to build a pipeline of talent at the top of the hiring funnel that is as diverse as possible and then maintain incredibly high standards.

I don’t think having quota systems is a good idea. It’s really important that at the top of the funnel you say, “Let’s get the most diverse candidate pool we possibly can.”

If there’s bias built into your sourcing techniques figure that out. If you have a mostly male pipeline at the top – there are a lot of great women designers out there – maybe the way we’re finding and discovering talent has some unconscious bias built into it? Now, when it comes to diversity in other respects it can be tougher to solve those problems because the problem is very long term. Design schools are not statistically turning out a lot of Black and Latino designers.

In terms of achieving racial diversity you actually have to get into the middle schools and the high schools and help develop curriculum to get kids excited about these areas, careers and opportunities. To encourage design education and engineering education so that kids in those communities and in those groups see that as an exciting opportunity. If you’re just looking at the design schools you’re going to be fighting over the same vary small percentage of talent and it’s just not a long-term solution.

Des: You’re also limited to people who can afford to, and live in a place that allows them to go to those colleges and universities.

Margaret: Absolutely. We have a really interesting program called Facebook University that looks to bring folks who represent these communities into the team but who may not have had the opportunity to have the kind of training that would make them qualified for standard internships or full time jobs. We’re training very bright women, college students, those who may not have had the opportunity or even necessarily the inclination initially to be like, “Oh, let me study software engineering.” We let them spend the summer learning how to code in a really safe and welcoming environment. That is a pretty extraordinary opportunity to change the trajectory of somebody’s career in a positive way. I think that there are things you need to do that are more interventional, depending on the community and the context.

The Role of Beauty in Business Design

Des: You’ve written about aesthetics in business products. Does the attractiveness of a product matter in a business context?

Margaret: This is a really interesting question. Julie Zhuo, one of my colleagues and an extraordinary writer on design who publishes a lot on Medium, has talked about this in the past. There’s kind of a pecking order of needs in design and designing good experiences. Beauty is great on top of something that’s really usable, serves a real purpose and solves a real problem. Beauty on top of a product that is hard to use and doesn’t address a real need is not that useful.

The reality is consumer design has come a long, long way in terms of addressing real human needs and actually becoming more and more usable over the years. Delight and beauty on top of that strong foundation of usability and usefulness is so powerful. It’s what everyone wants and increasingly in the consumer context it’s what everyone expects. Business software, generally speaking, is still trying to grapple with that middle section of just being usable.

Of course people want beauty and delight in their work content, especially because they experience it on a reasonably frequent basis in their personal lives using consumer products. The pent up frustration you can feel is mounting in a world where they’re like, “Wait a minute. I use these consumer applications and these beautiful Apple products and then I go to work and these products are awful.”

People are smart. They’re figuring out that the business tools and the productivity tools have a long way to go. You’re never going to trade off beauty before you get the usability, but I can tell you right now it’s not an either-or situation. You’ve got to be solving the right problems and make sure, like in our work on the advertising tools of Facebook, ultimately what matters is the creation of business value. We’ve got to solve the right problems for businesses, the ones that really matter. We’ve got to do it in a way that people can access and use. Then we have to figure out how to make them feel great while they’re using it.

Now, what beauty and delight mean in a business context may look and feel a little different than in the consumer context, because you’re getting work done. The aesthetics of that are probably going to always be a little bit different than what you might see in a purely consumer context, but that notion of craft and quality and attention to detail is absolutely critical. It just needs to be done on a foundation of usefulness and usability. Does that make sense?

Des: It makes absolute sense and I think that’s a wonderful place where we can sign off. Thank you so much for your time, Margaret. I really appreciate it and I know our listeners will enjoy this.

Margaret: It’s been a pleasure. Thanks for the invitation to chat.

For more conversations on design, product management, marketing and other startup topics, check out all the episodes of The Inside Intercom Podcast here.

Small business in Rajasthan, India: Department of Foreign Affairs and Trade on Flickr

Nuclear power plant: Tobin on Flickr