DDD Episode 008

Managing Products with Data ft. Shiva Rajaraman


Shiva: I think it’s really important for designers now to have a really good grip of what you can measure analytically, and look at the logs for what you can’t do that and what tools might help you understand that? Because the best way to defend your design and get it shipped often is going to be the evidence about why this design makes a lot of sense. 

Ted: Welcome to Design Despite Disciplines, a podcast series from the Master of Design program at UC Berkeley, where we explore the important fissures and emerging territories of interdisciplinary and design practice. This is episode “Managing Products with Data”, featuring Shiva Rajaraman. We are your hosts, Ted, Shuyang and Yu.


Who are product managers? In the context of tech companies, product managers are the ones who wear several hats to take charge of a specific function of a digital product. They’ll look into the market trends, study their user’s needs and make decisions. In a broader sense. Product managers are also data-driven designers. With user data and market insights, they design and iterate product features, optimize and nourish product platforms. In almost every job description of a product manager recruitment page, data analysis is always prioritized on the requirements. However, is data analysis really the ultimate, omnipotent weapon of a product manager, but might be its trade-offs in product management? How will the future of product management unfold? 

In this episode, we’re very honored to have Shiva Rajaraman as our guest, an experienced product manager lead at Facebook who has also served in YouTube, Spotify, and WeWork. He started his career as a software engineer after graduating from UC Berkeley with a bachelor’s degree in molecular biology, and later shifted to the path of product management. Shiva will first share his unique understanding of the data-driven process and product feature iteration, and platform operation. We will also hear about his insights into the potential risks and future outlook of the data-driven product development process.


Data-Driven Iteration 

Product management is an iterative process informed by data from multiple sources. Although product iteration might sound like a phrase specific to tech companies in the information age, it actually came into being long before the birth of the Internet. Modern product management originated in 1931 with a memo written by Neil McElroy at Procter and Gamble. It started as a justification to hire more people as “ Brand man” and later became a cornerstone for the modern concept of brand management, and ultimately product management. According to the memo, McElroy gave the description of the brand man’s responsibilities: tracking the sales, managing the products, advertising, and promotions. A unique point in the memo, was that such responsibilities are accomplished with two methods: field testing and client interaction. The Brand man will learn from the data collected from testing results and the client feedback and then modify their product plans accordingly. 

Shiva: If you think about traditional CPG, back in the way, a lot of experimentation, a lot of research, et cetera, but eventually you get to a product that sits on a shelf. And then it works, right? Or it doesn’t work, right? And it’s hard to say, hold on, my soap, didn’t smell perfect. I’m going to take those away and put out a new one.

Ted: That is the primitive concept of what is called product iteration in the industry of CPG, which means consumer packaged product, but it was still a slow and laborious process back then.   Today, for digital products like mobile and web applications, the product iteration process is accelerated like never before. Product managers now are able to monitor user behavior to analyze data in an accurate and rapid way. They use “buried points”, which are the embedded codes in the interfaces to track the events triggered by users. Our featured speaker, Shiva Rajaraman, who has improved YouTube with its advertisement and search feature, shared his understanding of product iteration back then and now. 

Shiva: In our world, one of the key things is like, what are these feedback loops and how can you basically roll out a service? How can you roll out software? How can you roll out an ecosystem and keep making it better over time? The internet is largely that you can ship something and keep iterating it, or you can roll something out to one country or one subset of users and see how they react. And there’s no cost to doing that. And you know, for a lot of CPG, if you’re getting it out in the world means getting it into Walmart or Amazon or whatever your ecosystem is . And you have to make that work at scale and make it fully functional. In our world, I think you can basically still make it fully functional, but you have a chance to iterate or try six things at once. 

Ted: Computational power, cloud services and artificial intelligence are developing rapidly. With the help of that, the product iteration process is evolving into a brand new framework, no matter what form of product or business you’re managing, you cannot avoid being attracted to the trend of digitalizing the management process, a more agile way to get access to data directly from the users. Shiva shared an example of how Disney plus is attaching more importance to collecting data directly from the subscribers to improve the series iteratively.

Shiva: Increasingly you’re seeing a lot of people want to get the data about how their customers use products independent of whether it’s software or not. So for example, a Disney now has Disney plus. Disney plus goes directly to the consumer. Disney plus can now see who’s watching what. And they can optimize the types of contents they go after for that consumer base based on what they observe. This is fairly new in the world. You know, it’s not that Procter and gamble can bypass the grocery store tomorrow, nor should they. But they are eager to get more direct feedback about their products and how people are using them. So they’ve done that through research in the past, but increasingly people want direct channels so they can see that commerce activity themselves and get the data. That’s a really a big shift, which is how by going direct to consumer, I can get that data and understand it. 

Ted: Digital products nowadays are not confined to a standalone software or a mobile app that serves a detailed function. More and more products are developing into a platform where different stakeholders would interact and engage in more complex behavior. How might product managers take care of large scale platforms and communities with data insights?


Data-Driven Platform Management

Shuyang:  A good platform, needs data to understand the user’s needs, obtaining user feedback, and that evaluates whether the product is successful. Before product design, we need to analyze user behavior through data and the analysis. It was there’s pain points. When the product gets full launch, we could use user’s path to see whether the user can’t achieve their goals. 

After the product goes online for some time. We can learn about the retention of users through data analysis for iteration. Only through data analysis could we build an online platform, listen to the voice of users and the optimize the online platform. Shiva used WeWork as an example to show how they use data to find the pain points and get solutions.

Shiva: Let’s talk about it from a capability. The key capability we hadn’t, WeWork was like an incredible capacity to design things ? So,  for a period of time , WeWork with launching a building a week .

And huge expansion, what the real capability there is that we work had the ability to look at a base building, go in there and quickly design the layout, but you could go in and rechange, you could change things on a weekend and maybe you didn’t get it right. Maybe it didn’t get enough conference rooms.

You could add one. That was one of the key advantages at WeWork cause that you’re the operator and the designer . That’s unusual in the world of physical, real estate. So by putting those two things together, could we use data that had to be privacy safe.  We don’t want to see everyone going everywhere as an individual, but  are people using this common space.

Is it that more people want to use the common space, but you know, you see effectively 20 dots there and then you see. 10 dots walking up and walking away because they couldn’t find a place to sit. Maybe you should think about either making that more dense or expanding it or whatever the solution by feet.

Because we were an operator and a designer, we could go and make those changes. Right? Like that was the key thing. And so I think one of the key things, when you look at like using data, you want to match that data or those insights you get to capabilities you might have.

Shuyang: In the process of digitalization in the real world, data-driven product design is happening and that even makes traditional jobs glow with new vitality. In WeWork, architects and the developers collaborate closely to understand the user behaviors through data analysis and the optimized space design. Thanks to the platform ,designers care more about data and technology and become a new design practice —— data driven design.

Shiva: I was at, we work, I was the CTO there. We obviously had a lot of wild stuff going on, but generally. We would hire  architects and then put them right next to software engineers and say, go figure out what you need to build because  we were opening up two buildings a week.

For two years straight that’s crazy. And the design, it was not cookie cutter. Each one had its own art and graphics and layouts . And of course you inherit the base building. So. You have constraints there too. We didn’t build our buildings. Right. We remodeled them. So just getting to that pace meant that the software that existed today was not bad.

So not for us. It didn’t have a memory for us. It didn’t collect data. So we had to put those two things together. So for those architects, though, they all basically had their first true immersion and technology and how technology works and it’s very different. So it was very bumpy for some, but all of them who’ve moved on from WeWork or still at,WeWork I think, are very thankful of that, that experience because they got to design the tools for the first time that they use. And they got to see that what software engineering and AI and ML can do for their profession, that was additive. So you put those two things together is really exciting. So I’m just a big believer in stumbling around trying to find something you love, but people first never put product above people.



Ted: Fast-paced data analysis is the tool to ensure that product iteration is carried out in an efficient manner. It also provides a unique perspective for product managers to strike a balance for product platforms. Data empowers product managers as analysts. Does rigorous data analysis guarantee sensible decisions in a product management process? What are some factors that product managers should be careful about when they’re dealing with data? 

First of, availability bias. In daily life, people tend to consider the factors that are easier to recall or measure to be of greater importance or consequences. This is called an availability bias. Remember the statistics 101 lecture about dependent variables and independent variables? Similar to designing an experiment, product managers conduct A/B testing to evaluate the outcomes of different prototypes by controlling independent variables. Some variables might be easier to measure than others, like user turning rate, daily active user, et cetera, but it is never wise to only focus on them to optimize the product. If you’re a product manager who is busy with manipulating complicated and bewildering data tables every day, beware of the availability bias. Don’t jump into the metrics that are convenient at hands that quickly. 

Shiva: Can you have unintended side effects by optimizing too precisely around one metric or even worse, do you optimize for only things that are easy to measure and not things that are hard to measure?  

So for example, if I take that use case from WeWork where I see, well, a lot of people are coming into the common space. I want to make it bigger. Well, you better be measuring also as well: do people like to be in large crowds? Maybe it’s better to keep it smaller because all of a sudden you have 50 people all congregating. Well, other things happen, noise levels, et cetera. So are you measuring the dissatisfaction of all the people sitting in the office? Right. So you have to be very careful about that.

Ted: Another note for data-driven product managers is the balance between quantitative and qualitative data. The trend of data-driven iteration could be misleading because it is possible for people with technical backgrounds to show their partiality towards accurate figures and vivid charts, but meanwhile ignoring the aspect of qualitative data such as user voice and evaluation. Product management should be applying scientific methods, but they are not equivalent to scientific experiments. Reaching out to the users and trying to open yourself to their complaints like a designer or even a psychologist is as much important as dealing with data spreadsheets or creating regression models.

Shiva: Amazon always had that kind of saying, if you have data, you know, that’s logged and instrumented, and then you get an anecdote that violates that data, trust the anecdote. For example, you take a look at the situation with space and a lot of people are using the common space and your metrics look good, but you’re getting a lot of people complaining about the noise . Or people going to customer service and saying, “Hey, I mean, there’s all these people who are sitting here, and they start hanging out and, doing whatever they do at 4:00 PM. I need to work till 6:00 PM. This is no longer a good space for me. ” Pay attention to that, even though it might be one point, you’ve got to really look into it, balance that with what you’re observing, because you just might be measuring the wrong things.


Future Outlook

Yu: While there are still some risks involved in data-driven iteration, product managers need to be careful to think holistically when using data to improve their products. But we have to marvel at how technology has so dramatically reshaped the role of product managers and the way they work. We can find that the days of on-time delivery being a product manager’s first priority are long gone. Today, evolving responsibilities plus an exponential rate of digital innovation mean the job is changing at lightning speed which makes us think about how the role of the product manager will be shaped in the future.

Shiva: There’s a couple things I would say about that. The first is just when you’re thinking about yourself as a product person, increasingly your role is to think about like the operational elements alongside of the product. And that’s why I think you’ve seen a lot of product individuals turn into general managers, or CEOs is that in order to make a great experience, now it’s not having the fastest website that loads . It’s effect. That’s a commodity. Now what is not a commodity is what happens after you make a purchase? What is the service like?  Do you have like unique abilities to delight your community? Have you invested in the community side? Because people don’t want to just buy something, they want to be part of a brand in their culture.  

One thing that’s really happening is that product people are not strictly about the product, they’re more much turning much more into owners of the experience, which means they have to be good at delivering a set of support services or post-purchase services or cultivating a community around their product. And many product managers were not trained in that they had functions do they participated with, but it was always a little bit of an afterthought to be honest, increasingly that’s really important to get right.

Yu: Yes we can be sure that in the future products and their ecosystems are becoming more complex. Life cycles are also becoming more complex, with expectations of new features, frequent improvements, and upgrades after purchase. At the same time, the value of the surrounding ecosystem is growing: modern products are increasingly just one element in an ecosystem of related services and businesses.

This has led to a shift in responsibilities from business development and marketing to product managers. New responsibilities for product managers include overseeing the application programming interface [API] as a product, identifying and owning key partnerships, managing the developer ecosystem, and more.

Product managers are no longer simply the glue between different departments,and only focused on execution. The product manager of the future increasingly the mini-CEO of the product. They wear many hats, using a broad knowledge base to make trade-off decisions.

Shiva: The second aspect of this, which is that the bar is just raising, going up higher. So you can’t really launch today with a half-baked thing and hope to get it to perfect. The bar for entry means this thing really needs to solve a job to be done very clearly.

You can use, Google cloud or Amazon web services and have full scale globally. If you want pretty quickly. You have a lot of tools to build effectively an awesome web presence or an app presence like web flow and things like that. You can do a lot of things with , low code versus, heavy engineering at this point.

So really, really the bar for design is going up because you need to design an experience, a community, et cetera.   End to end product design is really important. Compromise the design bar just to get it out quickly. That probably will not work anymore.

Yu: Indeed, we will find out the technical barriers are now gradually reduced and the product barriers are gradually reinforced. The emergence of low code platform, such as Crowdbotics, can be used to build business application without writing any code. Even a product manager with no coding background can design products and application using these local platforms.

Today, the technical barriers to build a product platform has gradually been broken and everyone can quickly build a website on the app. So a good idea of valuable starting points is the key competitiveness. We can see there is a growing trend that more and more similar products are being created because the online products can be copied at lower costs. In this case, product barriers help us to build a wall between our products and other copycats. This wall ensures that our idea will be protected from theft and imitation, and we will continue to be valuable after launch. Therefore, we will find that building product barriers will be more important than building technical barriers in the future, which will require a higher threshold for product managers and designers.

Shuyang: You have just listened to “Episode 8: Managing Products with Data“ from the Design Despise Disciplines Podcast series. Over the course of the spring, 2021 semester, eight teams of MDes students researched, interviewed, presented and the produced episodes featuring invited speakers from the colloquium. We’d like to thank Shiva Rajaraman for sharing your time and insights with our classes. To learn more about Debate in Design and the Berkeley Master of Design program, visit design.berkeley.edu.