fbpx Brand Name

Ep17. Product Innovation, Experimentation and Evolution of UX for Emerging Tech with Ha Phan

Career development Design leadership & strategy featured-podcast Innovation Podcast Process

Ha Phan⁠ is an inventor, speaker, product and design leader. She specialises in emerging tech, user-centric product innovation and experimentation.

In this episode, we delve into her product innovation experiences at Zillow, GoPro and many others. We explore the evolution of design beyond traditional UX practices, especially in the realm of emerging technologies like AI. This conversation will spark new ideas and perspectives.

Listen to the episode

Also available on all major podcast platforms and Youtube.


Vy: [00:00:00] Mhm. Hey, what is up? Welcome back to experience design podcast. And today I have a special guest Hafan and she’s a UX leader currently at Zillow. She has been for a product leader in the past, and this episode really digs into that experience. We’re going to talk about her challenges with specific examples.

We’re going to talk about her recent role. We’re going to cover design innovation and how she has approached it in the past. In a past and her philosophy for what design and UX should evolve in a future to keep up with the AI as well as other variables, which we’re dealing day to day. And this episode as usual is going to have so many gold nuggets.

I hope you enjoy this. If so, make sure to share it with a friend. Let’s spread that word. And on that note. Enjoy this episode. It’s good to gain extra perspectives, especially from super experienced people like yourself, but how did you get to that state? Cause you’ve been in [00:01:00] UX, um, you’ve been in product management or did product management and then kind of go back now, I guess, in design engineering or UX within the technologist team, if I’m putting it correctly, but what’s the journey like for you?

Ha Phan: No, let me, let me just take you back. I’ve been doing this a long time. Probably I started my career at the early days of when the first browser was created. So like when internet Explorer was the hot browser, I was an art student. So I, I started out working for this company that built a edutainment software.

And so I, my role was the storyboard, these game screens. And then I learned interaction design that way. And then I was kind of trained by engineers. So we didn’t, we had to write these very long, what did they call them? They call them treatments. So treatments are used when you’re pitching, you know, like, [00:02:00] uh, film.

And so, uh, I realized nobody read them and the engineers that they don’t read And they’re full productions. So they have like audio animation and all that. And, um, the VP of engineering that that time, it was a very small company at the time, he taught the designers, um, how to flowchart out other functions, uh, you know, the, of the game.

And, uh, it’s, I think it’s to this day, it’s still much more complex than software because software is usually, you know, like the action is instigated by the user. So when you flowchart it out, it tends to be. More linear or top down, but in the game, a game can prompt the user to do things. So it feels like it’s, you know, like it branches out, you know, like, so I think that that gave me a lot of foundation for understanding systems.

Even to this day, when I design, I can kind of see the layers of that flowchart in my head, so I can tell right away if it’s too [00:03:00] complex. Uh, if it’s too deep down, then the navigation’s gonna be hard. There are things like that that I just internalized during those days. I also learned a lot about game theory, because there was a guy who worked for Activision.

And, uh, I learned things like, uh, when people talk about gamification in software, I just laugh because, um, I learned things like, uh, there’s like automaticity in interaction where you just have like muscle memory. And, um, that a good game is a game that you can never win. And then there’s the ratcheting up of, you know, driving engagement.

I just internalize those things and I can see systems. When I, you know, put things down on, uh, when I design things. So that was my background. Um, but, uh, you know, so I went through, I worked for every type of company, but, um, I was in UX for many years before it was called UX, before I think I was called like user interface designer.

And then I think the, the funny thing to me is [00:04:00] people think that I have this. plan for my career, like I’m strategic and I actually have zero plan. I just, uh, one of my friends said that I’ve always chased interesting problems. And I think that is very true. Uh, I also kind of stumbled into product. I didn’t intentionally do that.

So. I was at GoPro for a number of years and I was an acting product manager for Advanced Tech, which was a team that was building out like machine learning. It was the very beginning of, you know, software teams even using that word machine learning. And so I was doing these little experiments for them and I was driving some of the work.

And so, They kind of gave me the title acting product manager and then so it wasn’t 

Vy: by choice, right? It was almost necessity then. 

Ha Phan: Yeah, because I had so much knowledge about the domain and I’d done a lot of the research that was one of my hats before I left GoPro. And then I applied for this job as UX designer at Pluralsight at the [00:05:00] time there were like a e learning platform and they had a lot of videos.

So I thought, okay, I know a lot about videos. So when I interviewed as a designer, the hiring manager saw my case study. And said, do you want to apply for the product role? And also the interesting thing about that was they, at the time they hired me for the, you know, PM of AI role back in 2017. And when I came in just the first three months of just understanding the platform, I just thought there’s no way I, because I wasn’t technical enough, but I knew enough to know that it would take a lot for them to have AI, you know, so I, but I saw like a very clear problem and I also saw like, That search was like central to the platform.

Like every content platform is a search platform. Um, and also I, I felt like it would be a good, um, I’m [00:06:00] not sure instigator is the right word, but it’s a, it’s a good place to be, to run experiments, to get the house in order. To kind of push people to have standardized data. I’m still like, I’m still amazed that the leaders gave me the opportunity because I basically say, Hey, let me, let me, they’ll search.

And I was, I joined as a, a PM who recognizes a clear problem to solve, but there was no team assigned to search at that time I had to like a lobby for a team and all the engineers who joined were super green. So we all learned together and, uh, within about, so I think when I started with the team, Was four engineers.

And within four years, three or four years, we grew to about 20 plus people. Um, so I was very proud of that accomplishment. 

Vy: Yeah, it must be a lot of lessons learned from it. And that’s, that’s another thing which I super keen to pick your brain on. Um, but out of interest, what was in the case study, [00:07:00] like where you got offered, um, A PM position or like suggested, uh, to go for PM instead of UX.

Like, what did you put there that, you know, it was clear that you could do product. 

Ha Phan: So when I was at GoPro, I think I joined right at the time when GoPro was just building out its software capabilities. And, uh, I think we were the first design team, the software hire. So there was like a bunch of people that were hired at a time.

And, um, so we built out for, we built out, I think the first, um, media manager, everything was brand new and we were learning how to build, you know, these cross platform experience. For example, we had to figure out how the camera as a platform, the desktop as a platform. Mobile as a platform, how those things all connect together.

And, um, I worked on a kind of a skunk work project. I was really, I really loved skunk work project and there was no real skunk work team per se, but we all, just a [00:08:00] few of us kind of came together. So I worked on a project that is a machine assisted video editing application where we, uh, built a wizard of Oz.

prototype where users could just select the highlights in a video and then it would, uh, the application would, um, create a video and sync it to music. And, um, I also worked on a, like built a strategy. I drove the initiative for the post capture experience for the 360 camera. 360 camera really didn’t exist at the time.

And, um, I basically brought it to the table because Uh, I was able to tell the story of like what the user experience would be like post capture. So there’s a lot of innovation there and I earned, I think, 11 patents coming out of GoPro. Wow. But I thought it was an accident, if that makes sense. It was like, it wasn’t me per se.

I was there at the right time kind of thing. I work with a lot of brilliant people. Yeah. 

Vy: It’s always like that. Yeah. Like even, even when you were mentioning, you mentioned about machine learning and [00:09:00] you know, that, I guess, inception of AI back in the day, I remember even working with similar challenges where, um, back in 2016 or 17, AI wasn’t even a keyword, um, nobody even wanted to use.

I remember talking to one data scientist and someone mentioned AI and we were just scoffed and said, Oh, don’t you mean advanced statistics? You know, and that’s, that was about it. But it’s kind of like we were working with decision support and machine learning and building like UX for that. And that was like, to me, it’s probably some of the best years in my career as well, because like I had to kind of figure out the toolkit, the ways of working strategy of how to work with data scientists too, because it’s not just.

Another team you join and like, everybody knows how to do stuff. It’s all experimental in a sense. 

Ha Phan: Yeah. I, I think that that’s my sweet spot. And then when people say, you know, can you teach people how to [00:10:00] do what you do? I said, I don’t know. I just go and ask questions. That case study, uh, I actually presented a case study at the IA conference in 2019.

When I was a keynote there. It’s funny because when we launched. That product, you know, like the initial beta, uh, I submitted that case study to be a speaker at AAC and they rejected me. And then, uh, and then the following year, the following year or two years later, so they asked me to be a keynote. I, I thought it was, I thought it was a fluke.

I thought it was like spam or something. I didn’t believe it because I had submitted before and was rejected. So when they, uh, sent me the invitation, I actually Google all the people again to see if it was real. And I didn’t tell anybody about it until it was close, because I was wondering like, is this real?

Because I’ve never, I’ve only been speaking at small events. And then Suddenly this big conference asked me to be a keynote, so I was kind of skeptical. [00:11:00] But yeah, I got tired of presenting that case study because so many people asked me to present that case study. When they see that, they’re like, hey, can you come and talk about that?

And I’m like, I don’t want to be a one hit wonder. It’s like repeating the same thing over 

Vy: and 

Ha Phan: over. 

Vy: No, that’s fair. Is there anywhere where like people who would be listening or watching us could find that case study? Is it? And it will capture 

Ha Phan: the thing. The reason why I don’t share it, because so it’s so hard to communicate without, um, without video.

So when I was presenting it, there was, it wasn’t just the slide that was doing the talking. It was the video because the video demonstrated how the music was thinking to the moment. So half of the presentation are in video. And then after I did it, I have this habit of adding video to everything I present.

So it’s very dynamic, but for that use case, specifically, it was so important that you have to see it live with video and see the music, you know, syncing. So, um. That was why I don’t share it because when [00:12:00] people see it, they won’t get it. They just won’t until they live that experience. 

Vy: So you went from, I guess, product there to Zillow.

It’s a unique team as well, as far as I understand from your tweets and, you know, like the information you share online, you know, the learnings and things of that nature. But like, what is that like? 

Ha Phan: The team is, is I feel like an idealized UX team. It’s called design tech and AI, because I think we’re working on the emergent tech, the future products for, um, the company.

And I think that maybe the reason why it’s coined as design tech is because our conceptualization and our, um, process is a bit, it’s a bit more closer to an R and D, uh, skunk work types team. So for example, if you. I get in trouble for sometimes when I talk about my process and I, sometimes I think I shouldn’t talk about my process, not because of this NDA thing, but more because it’s so different from other products.[00:13:00] 

Vy: That’s exactly what I want. Yeah. It’s so different 

Ha Phan: from other products. So for example, it’s like, if you think about most of the products people work on today, the enterprise, their web based, um, they have existing patterns, you know, we have design patterns, so we know the behavior behind those. Patterns, right?

Like when you design for iOS, for example, you know, you know, when there’s the sheet that slides up and there’s like inherent meaning in going deeper or abroad when you’re interacting with those patterns, but when you’re working in emergent tech, uh, I’ll give you an example. If somebody told you the go design this game, it’s a 3d game.

And, uh, but you never done it before. Uh, how would you prototype that experience and what would you do first? It’s like that, it’s like something, the technology did not exist before, there might be similarities to other things, but the pattern didn’t exist before, and then you don’t, you don’t have constraints yet, [00:14:00] or you don’t know like what the possibilities are.

And the constraint lies, if that makes sense. You’re pushing science and science push you back. So I explain that it’s like a chicken and the egg problem continuously. Like, science goes and says, what should, you know, no, design comes to science and says, hey, what are the constraints? And in science, like you tell us, maybe we could focus where it should be.

Right. And it’s just go back and forth. And, and so we do a lot of building to learn versus like in existing products, because you know, a lot, you can, you could probably jump to solution very quickly because, you know, there’s an existing product or, uh, you’re working enterprise, or it’s not like you need to ideate in three different ways because you know, the customer expects these kinds of things.

So you don’t need to do all these divergent things because you’re, you Within this contact with domain, this, these rules already. Um, so, uh, you know, it’s really hard to say it without really describing the thing, but 

Vy: [00:15:00] yeah, but my, my knee jerk kind of thoughts when you were describing it is like on one hand I was thinking, Oh, but that’s, it’s still UX.

It’s still approaching very complex challenges with some sort of framework or, you know, some sort of ability to. Diverge and converge. It’s not it’s not predefined. Okay, we’re gonna take this Double or triple diamond and and kind of apply it but it’s still It’s still ability to jump from convergence to divergence appropriately, right?

Like and at least that’s how I would approach it Uh, you know Not knowing the domain or maybe how I approached it before but to me it seems like maybe that’s where like you just kind of separate the frameworks and Take the philosophy, or I don’t know, the meta level of design, right? 

Ha Phan: And when you talk to scientists, that’s, that’s what you’re trying to get to.

You’re trying to align their mental model with yours. You know, like they’re talking a different language. Like sometimes they go beautiful [00:16:00] mind on me and I’m like, I have no idea. So then I would take them to a whiteboard and I was like, okay, explain to me, draw out for me what you mean. And so sometimes they draw out like, System things.

And I would, you know, try to re say what I think they mean in my own words. And so I just so I can understand, like, from their mental model, what is the model for what the AI model doing? What problem they’re trying to solve? And more importantly, I do think that the role of design. there is to be the behaviorist on the team.

Vy: How so? 

Ha Phan: So, uh, I’m going to go back to the game analogy. In a game, whenever the game tries to get you to do something, it throws something at you. It steers you or it, you know, it maybe gives you a cue. It actually tries to shape your behavior within the game. Or like, for example, if you take, um, Any 3D game, uh, in order for you to navigate through the game, you have to build a nav, navigational mesh so that the system know where you can go.

And so you as a designer is [00:17:00] thinking of what is the behavior that you want, uh, the user to take in the game. What should be the constraint, right? I always hear designers complain about constraints, but actually the goal for you is to define what should be the right constraint for a good experience. And actually it’s very good discipline.

You always start with little and then add more. You make the user ask for it, not. You know, just give them the world. Don’t know what really works. So that’s, that’s what I mean by, I’m always questioning behavior. The other thing is when you work, uh, with AI models, you’re always dealing with hallucinations.

Um, so the other piece of it is putting on your researcher hat and saying, where is trust eroded in the experience? Uh, and then understanding what kind, what, what does the hallucination look like? And, you know, what is the, what is the variance? And how does it impact experience? The UX is really in the AI.

Uh, it’s really not in the UI. UI is just like, you know, like a hammer or whatever. [00:18:00] You know, it’s just a little thing that you need to, to interact with the AI. But, but that’s why I think having a research mindset and then you know, Uh, really thinking about the user behavior in, you know, interacting with this, with AI, um, is really critical.

Vy: Presumably then you would, I guess, focus on very evaluative research, right? No. 

Ha Phan: So I’ll give you another example. Um, when, when I was working with, uh, data scientists, On the search team, for example, they would, you know, they would build different experiments to test models or they would purposely, you know, like try things to kind of eliminate them.

Take the example of the GoPro experience. If you were looking at the market at that time, when you ask somebody about storytelling, this is before TikTok where people don’t even have a language around storytelling. Now people talk about TikTok and reframing and all that stuff. But back in [00:19:00] those days, if you asked somebody about storytelling, they would talk about Premiere and, you know, like the tools, and they were really complicated tools.

The provocative position is that you’re saying, we’re going to take away all those controls and automate this for you, but you can’t tweak it. You’re going to get something really fast and acceptable, but not, not like, it’s not going to be like a Spielberg motion picture. Right. So it’s kind of like taking those provocative positions and saying, what is the trade off that the user is willing to make?

And what is the behavior that can be scaled? And then, you know, how quickly would they be able to replicate it? So one of my continued philosophy is that the strategy of any product is in. It’s metadata. So here’s another example. You can look at the live photos in an iPhoto, right? The live photo is a completely different thing than a regular static photo, because a live photo has a duration.

It has a key [00:20:00] moment. And from that live photo, Now all of a sudden you can construct different stories because now there’s a, it’s a clip, right, and it understands your behavior, how you capture. So the metadata of a live photo, how you describe it is different than a static photo. Another example is if you look at things like any in kind of social media, you look at, you know, Long video versus short video.

The time that it takes for you to generate a Instagram, you know, re you know, like story or real, that is like a minute long. We’ll ensure that the cycle of scale of content that’s created. Right? And then so basically you’re building, you’re strategizing, like, the scale of consumption, the scale of exploration, the kinds of content, just in time content versus lean in content.

So I think about interaction at that level, and then the metadata determines the behavior that you will have. With that thing that you’re interacting with. 

Vy: But how do you translate, [00:21:00] I guess, the users needs and their behaviors like that to me, it’s, it’s quite deeply evaluative, especially if you going to work with something innovative, which doesn’t exist.

No, but you 

Ha Phan: have to, you’re forcing them to make trade offs. So you never test one thing you test. You force them to make, to make a trade off between one or two things and you, and you design your experiment in a way where neither one is the optimal experience. Each one is a behavior and that the final solution might be a combination of something, but you understand the real motivation when you do that.

So that’s, we have this term, this phrase in our team, it’s called prototype and you see science, you know, scientists and, and. Uh, engineering do this all the time. They prototype like it’s like a spike. They do it so they can see how it works. And they go, all that didn’t work. You throw it away and you do it as quick as possible.

You’re not trying to like get every pixel perfect and so on and so forth. So sometimes I’ve really forced very [00:22:00] narrow. Uh, point of views and really a lot of constraint just to see, you know, what the user would put up with. Right. But there’s always like, there’s always two opposites and it’s the same way when I coach designers.

I used to, when we have crits, I used to really push designers to bring a lot of concepts to the table. And then I, I kind of pushed like, Oh, this one. This one is the opposite of this one. What are the foils between, if this one is your favorite one, what is the foil to that one? So there’s many dimension of foils.

But when you do that, then you can kind of think critically about the problem you’re solving versus like, Oh, I’m in love with this design. 

Vy: Which is so typical of typical UX work in product design and service design and everything in You, you’re kind of bound to deadlines and timeframes and constraints to get it right and right as soon as you can.

I think maybe it’s part of how designs position themselves or the ideas or the [00:23:00] challenges. Cause if you talk to a typical product manager, they’re going to tell you we should make more failures or fail faster, you know, however, you’re going to position that. But I think. My observation so far has been, and, and you know, I, I managed massive design teams before and it’s, that’s where I’m coming from.

Designers don’t have that experience. I guess over those years of, of putting up something, knowing that it’s a bad bet to take in a way, if that makes sense. Or like something which could fail and then that could be a learning point. Maybe it’s a luxury. I 

Ha Phan: think it. It’s everything is, it depends. I learned a lot working with the search team because we were building something, something scratch and I had designed this vision for an intelligent search and all that, but there’s, it takes, you know, years to do a homegrown search.

I just learned to be really comfortable with failing. I think it was really hard for the designer on the team. Um, but he, you know, he rolled with it. He would, he, he’s very awesome. Uh, I’ll give you an [00:24:00] example. So initially when we built our first version of search, I knew it was going to suck because it just, you can’t build search overnight.

So I gave the team the first goal. So we had a preexisting search that was, that was, uh, that we built from a third party solution. Um, and. It was Adobe’s search to promote and it was not the right solution and it actually, uh, failed. I had an outage for like went out for eight hours or something like that.

And so, uh, at the time we’re still just, you know, like brand new search team and still building the stack and everything like that. But I gave the team the goal of, um, Uh, a parody. So, so the existing search, uh, thankfully, it was terrible. So, what I meant by parody is parody in outcome, not parody in features.

You know, as long as the exit rates were the same, as long as the retention was the same, nothing blew up, no harm done. It was fine. So, I actually stripped all [00:25:00] the features. Like, we had all these filters. I killed all of it. You only have the basic search. And that experience taught me a lot that, you know, it gave, it was like, I was a new product manager.

I did this crazy thing, you know, and if it failed terribly, if things blew up, it would not be good. But you know, nothing happened. I mean, a lot of people complain, but the numbers are still the same, but people complain nevertheless. So over time, the team just got more comfortable at releasing to get data and just ran a lot of experiments.

And I remember one of the things that, so before they launch anything, the engineers had to go to the product manager and say, you know, are we good? Can we release? And I used to say, release the cracking. It was funny just to say. 

Vy: But that’s, that’s the thing, which is going to be, um, a lot of the people listening might find this because it’s going to be majority of designers, right?

Um, they’re definitely going to find this a bit, but 

Ha Phan: There are many different ways [00:26:00] to de risk, right? There are many different ways for you to It just depends. Like I, I had enough signals to, to think like, okay, I don’t think we’re going to fail. You know, I don’t think, I don’t think it’s going to be that bad, but there are many different ways to de risk before you get there, before you, you know, you’re not, you’re not going to, you’re not going to.

Fail spectacularly, you know, so that’s the thing you don’t want to do. You don’t want to fail spectacularly. 

Vy: Maybe it’s from your tweets or maybe somewhere else. I’ve seen, um, I think it, it, it, it was something from you, but the designers are quite attached to the frameworks, to like a certain process.

Maybe that has something to do with that. Like, um, what, what is the missing link? Probably a lot of your experience comes from product management and previous roles. But what could designers do better in a way? I don’t know, as simple as letting go of old notions and frameworks or like, how can we actually get better?

Ha Phan: So I think that, I think that you can let go of frameworks if you have critical thinking, you know, like if you’ve got nothing, you better hang on to the frameworks. [00:27:00] But I think this is not. Unique to designers is also for product managers. I think it depends on the problem you’re solving. If you’re working in enterprise, like the business model matters a lot.

The maturity of the product matters a lot. Um, if you already have like a lot of customers for a product, if, If your product is mature, has a lot of maturity in it. Um, and then you have to ask yourself, like, what is your bet? Is your bet like testing a new feature? Is your bet like testing a new distribution channel?

You have to define the opportunity based on the bet, right? It’s not like you talk to a few customer and they say, I need this thing. You know, I have this gap or friction. It’s not, it’s not that simple. And I think that’s why like a lot of designers and product manager think too simplistically about that.

Like, what is the outcome they’re trying to create? Are they trying to optimize for the existing customers? Did they discover a new segment, you know, that they want to deliver for? Here’s another good example. So when I was at [00:28:00] Pluralsight the last year I was there, we were getting beaten up by the competition because the competition had invested a lot in certification content.

And they only focus on that while we’re trying to be like peanut buttering across everything, across e learning for tech. And then so I remember the VP at the time came to me and asked me like, we got to improve this thing. And I said, well, I can’t build a certification center overnight. So I went to, I just, you know, went to the analyst and I said, what’s the, where’s the, where are we getting the most traffic for which topic for which certification?

And they said, Oh, it’s the cloud. So then I looked at that and I, I had a new PM at the time. She was only like a month into the job and I said, you know, do whatever it takes, but just build me a good certification center only for the cloud content, only for that, just, just make it better than the niche content providers [00:29:00] for a certification.

Just do, you know, do what you can with just that. And so then in that process, what I really learned is, uh, we didn’t look enough at segmentation. And I also learned that the way people approach different content, their learning intent is different, but we were put peanut buttering across. So when you think about that, you got to think like when you think about, Oh, if you only look at the certification problem, you’re like, Oh, well, they can’t find it or, uh, search is terrible because it doesn’t, you know, uh, recognize that, uh, solutions architect is a certification, things like that.

You’re only like. nitpicking the little, you know, the little symptoms, but not the illness. So you have to really think, you know, think broader and it doesn’t happen overnight. You have to just kind of dig into the problem and really learn it from, from a market point of view, from opportunity point of view, and define the segmentation that you want to solve for.

Designers never think about that. They just think like, oh, there’s a friction here. I’m going to solve that. Right. [00:30:00] So I think understanding the opportunity sizing is really important. 

Vy: Something to double click on is the critical thinking, which surprisingly comes up in, well, this is, I guess, 15th or 16th episode of a podcast so far, but had quite a few very senior people on and every single one highlighted critical thinking as one of the, I guess, skill sets, um, which everybody must hone in.

But it’s so hard to kind of define, like, how do you become more critical thinker? It’s a journey of really being very good systems for thinker, but then also applying the models like, um, Socratic questioning or, you know, ability to almost zoom out and question yourself as much as. Questioning the problem or, you know, the people around you and things of that nature and when developing it slowly, I would bet that nine out of 10 listeners probably are [00:31:00] not confident in their critical thinking abilities.

Ha Phan: For me, I, I’ve repeated this so many times and people never do it. It’s really becoming really good at research and I don’t mean knowing how to interview people or I mean like when you design, you’re actually trying to get frame your experiment, you’re trying to figure out what is the proof that you know what is the proof like when scientists that in exploration, they’re trying to design the proof I use this example before and meet my PM at the time said I was right, but I got yelled at by the researcher.

So I said, it’s sort of like, you know, when, when you run those experiments with mice and then you, you block the maze and the mouse goes a different direction. So you, you’re basically building different mazes to see what the mouse would do. And then, uh, my researcher would say, users are not mice. But, but the idea is that you, It’s not that really, the [00:32:00] idea is that you’re, you’re thinking critically about what you want to happen when you design.

And then you’re also thinking critically about different dimensions like scope. Because you can’t make it all happen at once. So you’re trying to say, what am I trying to prove first? And then what am I trying to, what behavior am I trying to scale? And then in my head, behavior is data. Because when I was working on search, like, we’re gathering signals from everything.

The user is clicking on. So we have like what we call implicit data and explicitly data, right? Explicit is when the user is actually doing something and the other ones were inferring by gathering data across the platform. So that, so I’m always thinking like if the user does this, what is it telling the engine, like the AI, what is it telling, uh, what, what signal is it like building confidence for?

So I think that for me, it’s. It started out with the work I did at GoPro because I was running, you know, the research, the conceptual research for a lot of Blue Sky stuff. I was telling that [00:33:00] story. So in having to explain that to people, it strengthened how, you know, like my own internal framework about how I go about doing research.

And then working on search forced me to be really rigorous on the quant side. So I used to tell the team that every activity they have before a thing is released, Is to get to a narrow, clear hypothesis about what they’re putting out in the world, right? So whether you’re doing like workshops or qualitative research or design exploration, all those things you get clearer and clearer to a narrow hypothesis that you can test at scale.

Vy: Pick one of the quotes from which I think it’s going to maybe build up upon what we’re discussing, but it’s boring from your tweet, by the way. So in the age of AI and perhaps the beginning of immersive experiences, we are required to work even closer with engineering and [00:34:00] scientists that design will evolve.

Anyone who is solving real problems have left the ivory tower long ago. Yeah, I remember. I don’t know if you remember this, but like, I was very intrigued to hear like, what could design evolve to? Cause I think we’re hinting from what, you know, from those inefficiencies, dysfunctions, but like, I wonder what’s your kind of futurist thinking for design.

And I’ve been thinking a lot the last 

Ha Phan: few months about writing an article or something about reclaiming UX. And I’ve long felt. That design needs to be more pragmatic. We have some amazing designers at Zillow. Like, they’re super, they’re amazing. Uh, I’m not sure if you got to see the, the Vision Pro there, there was a, there was a release of the Vision Pro and then there was a Zillow experience inside of it.


Vy: no, not that specifically, but two, tell one of those 

Ha Phan: designers are. You know, he’s amazing. He works on my team. I just, you know, I got to see like the tools that he [00:35:00] used to build that experience. His, you know, he leverages a lot, obviously from, you know, where we are today, but he’s using new tools. And I think in our team design tech and AI, we, we look at the problem.

We’re like, okay, how do I bring that to life? Uh, the tool I can’t just use Figma. We’re always thinking like, how do we build this Wizard of Oz thing to communicate a provocative. When we’re building those things, we’re kind of framing a new mental model, right? Because this is like the world that the user is going to be in.

This is a different way of interacting, whatever that is. So the reason why I bring that up is that in our journey of trying to define that, I think the hardest part of that job is to align the mental model, your mental model, with scientists. Thanks. Scientists are in their own right. UX designers, they, you know, like, when you think of all these generative AI [00:36:00] products that come out, somebody has to articulate, like, how people, how users are going to interact with them, what problem they solve.

It didn’t exist before. In science fiction, maybe, but not in real life. So, they have a certain model, and they’re innovating to see, they’re just trying things out, right? And then, you know, we come along, we collaborate, we go, No, this is, you know, we inject values like trust. How much control should the user have, you know?

Control gives people a sense of trust that they did something to it. So there’s all these like very philosophical themes that emerge when we work with scientists and engineers, we form principle because there’s nothing that exists before. So we have to form, you know, these aligned principles and then use them to make decisions.

And so that’s what I mean is like, it’s a lot of it building these products is a lot of it is a human. Uh, interact, collaboration. We’re kind of talking through like, you know, this is the problem we’re solving [00:37:00] for. These are the behaviors. Uh, this is how people think about this in this new experience. Uh, because it didn’t exist before.

And do they trust it? At what point do they don’t trust it? What does control mean? There’s many different ways, like UI is all about control, right? So, so, you know, what is, what does the control feel like a lot? There’s a lot of feel that you can’t articulate, like in the game, you know, like it feels very free in the game.

Um, but at the same time, the gift, the game gives you a lot of cues. So there are things like that. When you’re, you’re. Where you’re kind of building the story in a sense, it’s really hard to explain without like an artifact because you’re in this territory where you haven’t, you don’t have patterns yet.

You don’t have design patterns, the constraints are kind of, you’re trying to find the edge of the constraint, you have to work really closely with each other and then you have to build to, to learn because you’re like, sometimes you have a, Your intuition or you have an itch you need to scratch and [00:38:00] just go, I’m going to build this and see how it deals.

And most likely a lot of times you throw it away, but you learn something else. You’re like, Oh, I didn’t expect this to happen or that felt great. Or so I think that that’s, it’s a different territory for design because right now the patterns on the web and, and mobile are so well defined that that’s almost like graphic design, really.

You’re designing one dynamic state to another versus like the rich experience, right? 

Vy: In the last episode, uh, I, I, you know, we just published, we, we chatted with Kevin Richard, who’s a strategic designer. And I think I should have maybe cut it out, but I. You know, because of a Figma and design systems, which, which is all good, you know, it’s all optimized, uh, to quickly prototype, but it also is, is one of the reasons for bland web or, you know, like all the issues which designers also shout themselves, but I was maybe unrightfully comparing it to designers becoming forklift.

Uh, [00:39:00] drivers and stacking different boxes in different orders because everything else is already defined. And of course, some designers might take offense of that, but so be it. I feel like from what you shared, it’s like you have to try a lot of things and. Also be very open to new methods and new tools.

And like, I would even would love to hear, like when you mentioned the design tools, uh, your colleague used to create that Zillow Spatial experience. Was that in code, I guess? Like how, what were the tools? Because I presume it wasn’t just I can’t really speak to that. Maybe it started with that. There was a 

Ha Phan: special team that built that.

Um, but I, I didn’t know that he, he utilized like 3d tools. I know that I knew that, but I wasn’t intimate with that. I just attended one meeting that he was in. He kind of showed like what tools he used, but it was enough for me to understand that, um, you know, where we’re going, there are no tools yet. 

Vy: You know, if we were to put [00:40:00] our futures thinking hats, how do you think the tools are going to evolve?

And also just to preface, I think I’ve seen some good development from the likes of Figma. Yeah. Where, um, they kind of trying to bridge that 2d gap between the code and the design and us working on different abstraction layers. And then handing it over, recreating the ideas that are being expressed in different tools, different boxes, and then hand it over and re expressed.

And there’s just so many translation efforts, uh, you know, to, to simplify it. And I think in, in my head, Figma is likely just going to continue on that. Uh, no to minify the abstractions and create likely one single tool for design engineers or whoever is going to use it. But what are your ideas? I 

Ha Phan: think mostly the, what we’re doing today is conceptualizing.

So Figma right now is a conceptualization tool [00:41:00] and implementation tool. It planned the whole entire process for design pretty much. But what I was alluding to is that just by watching how my colleague was building the experience, I felt like he utilized not only different tools, but different skills, you know, like he used to work in games and everything.

So a lot of people don’t have the experience. Tools, um, pre bias you in because it’s what you can do pre bias you into a certain kind of thing you can build. Does that make sense? So if you want to stretch that, you have to, you know, get into, basically you have to get into work with engineering. That’s why design tech, you have to figure out how you’re going to conceptualize a thing that you have no tool for.

You don’t know how to, you can build it yet. Right. Looking at what things are emerging today. I do think that what’s coming out of. Apple and then the things that I know with AR and a lot of like what the new visual landscape, let’s put the next tier of visual landscape is going to be, you [00:42:00] know, more AR.

And so I’m like, well, how do we, if you had to build it today, what tool would you use? And so I think that’s the, that’s why I alluded to that, it’s already like those of us who are trying to imagine it have to, you know, borrow and rob from other, like, borrow and, and utilize tools that weren’t probably built exactly for those, you know, experiences and then just to communicate what we want.

So I think, I think that’s what I was alluding to is that that’s where it’s going. And I also think it’s, Yeah. I also think it’s weird that everybody is focusing on the LLMs and AI is so much more than LLM. So 

Vy: no, don’t get me started, you know, in my design community and everything else, they see a lot of, I guess, cause cause that’s probably easiest to pick up for a lot of designers to kind of take one of the conversational designer designs or, or apps based on conversational designs or chat, GPT wrappers, and kind of start to re skin [00:43:00] it.

make reconceptualize, but I don’t want to put to put on like a elitist UX hat just to not to discourage people, because again, design processes we discussed varies depending on the challenge, but it’s not really like deep UX. It’s very surface level. It’s very entry level challenge. Whereas, and of course, some people are going to start with that.

So that’s why I’m kind of very, um, here, but I feel like maybe a lot of that is also bound to the web. Space, web apps, you know, figmification of the things like basically of what has been happening for the last so many years, really now with us obsessing with design systems and everything else, but we really just created ourselves these specific Lego blocks.

And now we’re thinking of how to still use those Lego blocks using the very basic. I don’t want to label LLMs as basic, but it’s a start. It’s basically, you know, you can do so much more with a machine [00:44:00] learning and creating custom models from scratch. Basically, I guess follow up question to the actual tools and everything else is really, how do you foresee.

And, and maybe if you could speak a bit about like the stance of Figma too, but that dependency because it impacts a lot of things about like design engineering as a role or some like the buzz of it, like there has been uptick in IC roles as well. But do you feel like maybe it’s a signal that the. The design and engineering or goal on marriage?

Ha Phan: I, I don’t really believe this is what it’s gonna go, but say, say for example, uh, you know, I build a, an application and I’m focused on the LLMs, and you know, the, like, if you think about search, you can say, okay, the LLMs can answer all the questions. I don’t need to see search results, you know? You can see that maybe I don’t need a lot of design.

I can just, you know, I can just, uh, have a design engineer do a little bit of design. So it depends on the product where, uh, or enough patterns exist. I don’t need design anymore. [00:45:00] I’m just going to evolve this thing. So in those instances where I feel like the domain is clear, the startup can kind of run by itself with minimal design effort.

They can. Borrow a design pattern from somewhere, you know, and then engineers can, you know, bridge that gap, you know, some good front end engineer. I can see that happening, but it depends on the product. I also see an opportunity for UX to be really strong because, uh, I, I went on, I was on, I’m on TikTok a lot and I follow Nate Jones.

He’s my hero. He didn’t say this, but in his comment section, a lot of product managers follow him. And some of the product managers said, Oh, you know, that. Data scientists are defining everything. I don’t know what my role is. And so I thought, so this is technology led product, right? Because the technology is defining the future.

It’s an existential question. If you don’t define the future, then you’ll be done. And so then what is the role of product? However, if you [00:46:00] have a really strong UX partner, the US partner can define for you, you know, what is the experience with AI? What is the idea of trust in AI? What is the interaction with AI?

What is it that users need? To see that they can’t see because before, when you didn’t have AI, you have to show the options and you know how AI reached output, but now you can’t, maybe you should show, you know, there’s all these wrinkles and layers in terms of how you build trust and how you, the values that you want to espouse in your product, like how much information does, does AI have?

You know, that’s the user need, even if you know, it’s kind of like when you’re doing a math problem and you look in the back of the book and you go, Oh, I got the answer. But then, you know, maybe that part of the experience is that you want to lead the user to the answer, whatever. But there’s many wrinkles like that that creates a more human experience.

So I do think that. Good UX designer can drive, uh, AI products because it’s so, and also it’s human, right? Because [00:47:00] then you can ask yourself, like, what is the voice of the AI? What is the role of the AI? You know, all that stuff. So, so I think that there’s real focus for UX designer, but they, they kind of have to be the jack of all trades.

They have to be like philosopher, you know, they have to be in ethics. They have to, Understand system thinking, then you have to understand like the role of a, of human in the loop. Uh, they understand like data infrastructure. So all of those things, I don’t think that there’s a lot of people in the market, in the U.

S. market. 

Vy: I agree with you 100 percent of that. When I, when you were just describing, I was like, I wanted to immediately say that people would need to almost like pull themselves up, choose themselves. To, to lead and to take that position too, because it’s too easy to take the backseat, especially when technology is driving that.

And I’ve brought up this point so many previous episodes on this podcast, but there has been. Massive uptick in product [00:48:00] management in AI roles or very technical product management roles, at least in Europe. Um, I don’t know what us let’s say, but it’s kind of like, to me, a signal that there’s going to be a lot of PMs trying to manage the tech centric feature, but there hasn’t been an uptick in UX series, but on the flip side, like, so what?

Someone is going to have to do it. And unless you really want to just skin, get another chat bot or add another thing, it’s like, this is a call for like a strong, uh, design leadership. Almost like, I guess, you know, to step up and kind of drive it. And however, as, as you said, I guess you have to go through this chasm of learning.

And. Developing new skills as well. Like if you don’t, then the future is very dire. I ran this live stream a while back and, um, and kind of tried to project exactly like which of the skills in UX are going to be automated first. And it was the tasks. It was, it was all the skills which are [00:49:00] tied into design systems, prototyping, quick production.

And then on the latter end, it was about research strategy. Architecting things, very human centric qualities. I think someone still needs to develop on top of that, the appreciation for AI and, you know, working with data scientists and things of that nature. 

Ha Phan: I don’t know if the schools teach that today, like in design.

Vy: Well, Dan Saffer, Dan defined the UX disciplines back in the day. Like so many years ago, um, he’s actually a professor teaching people in design courses, specifically designed for AI right now. So to me, it was like, okay, at least one place, um, in, in, you know, CMU is doing that. I, yeah, I think, I think there is lack of maybe information.

We maybe need to share a bit more with the communities of how to do this, um, or learning or as you do as well. Like I think your Twitter feed is a goldmine for that. Yeah. Cause I’m, I’m learning and I’m like, oh yeah, this makes sense. This [00:50:00] is kind of like, you know, it’s, um, on, on that note, uh, sorry to pivot slightly, but, um.

One of the things which always, I guess, is challenging for everyone, especially in UX, for UXers is, um, the proof of value. It’s a fresh concept, even for PMs. But for UXers, as much as I encountered this very foreign, UXers rarely have to proof of, you know, proof the value for the users or like the actual value fits and things of that nature.

What, what is your thinking and approach? 

Ha Phan: You know, when you’re adding value. You just know, you know, I mean that subjectively, I know when you work at a job and you’re adding value, you know, when you’re not, when you’re like, not sure you’re probably not adding value. In the past, when I was a director of product, we had performance evaluations or performance planning, I guess is what they call it.

And, um, basically we have, we define like the business goals that, that the individual contributor, the designer would [00:51:00] contribute to. So, uh, it could be that you’re adding something to the design system. It could be a design organization, operational goal, or it could be a product goal. So whatever your, the impact you’re making, uh, always maps up to that.

Then the person would kind of say, what did they. Define as good in what they define as great. It’s like an ongoing conversation, but at least it’s a pragmatic. It’s like that you’re working towards this. You’re aligned to, you know, add value to this business priority. Uh, and the goal should be written based on the design, kind of like a.

Design point of view, for example, it could be that you define a direction that map that informs the road map. Uh, it could be that, uh, you launch how many experiments, uh, towards that goal, something like that. But these are really tangible things so that you know that you’re mapping to that value. These are like, these are operational things, you know, ways of [00:52:00] like proving design value.

Intrinsically, I feel like for a while, when I was at GoPro, I was working on a product that wasn’t a project that was not on a roadmap. And that was one of the most, uh, impactful project of my career. And it was, you knew it was talked about. So you knew when you’re creating impact. So I don’t know. That’s, I still think it’s a hard thing to prove.

Or to show, but ultimately, uh, you can try to scope your projects yourself and just ask yourself the question of like, so one thing I try to do is I try to ask the PM, I said, Hey, how would you define this task with design? Because I’m trying to scope it. And then, um, and what, what is your, what do you think the outcome is, uh, you know, to show that this has value or, uh, what is the, you know, the defined deliverable and it’s very, you know, you can’t do that right away, but it [00:53:00] helps you to kind of shape that, right.

It helps. It’s an ongoing conversation, but 

Vy: no, 

Ha Phan: you can design forever. You can go on forever. 

Vy: It’s probably, uh, it helps. PMs think about it too, because, you know, sometimes it’s, it might be thought of in isolation, but not shared as a team. Like if, if you are doing that, or if you’re asking that bold question, naturally someone else might have thought about it, but in different angles, let’s say.

And I guess we discussed a lot of the things about UX and how UX might evolve and a lot of different skills and things of that nature. But if there would be like one. Piece of advice. What would you like to see in, I guess, UX industry today? 

Ha Phan: Man, that’s a really hard question. I would like designers to be better researchers.

And I don’t mean, like I said it again, I don’t mean that you need to, you’re writing the script [00:54:00] to interview users, but you can articulate the proof or the hypothesis behind each concept that you, uh, design that’s the beginning of everything. If you can’t do that, then everything else doesn’t matter. You have no intention behind your work.

Vy: I think it’s just maybe the separation in roles, which probably doesn’t aid us is, you know, between product design and, um, user research as to specific skill sets has also been a bit problematic in that sense. 

Ha Phan: Yeah, but I don’t think that’s an excuse. It helps you to. It helps you to be more strategic. So say for example, uh, in your, every, a lot of companies have research.

Researchers are great partners. So you have researchers and you have, uh, data analysts or business analysts. So in the past, what. What I’ve done on my team is the designers would bring many concepts to the table based on some, sometimes even poorly [00:55:00] written bets, right? They bring a lot of ideas to the table.

And sometimes we, we, I require them to write the hypothesis, even if it’s poorly written. They need to write a hypothesis for each of their concepts. They need to Articulate to the best of their ability, what metric they’re going to move. And it’s likely wrong, but it’s fine because then we can kind of dig into and say, okay, this is traction.

Why do you think this metric would change? Well, all that stuff. And then what happens after, after we debate all that, and the PM’s in the room too, then what happens is they talk it through with the, we didn’t have researchers at the time. So I was the researcher. Um, but. We, we talked to the data analyst at the time, and he would help us hone the hypothesis and the metric and engineering, and then we fill out what they call a experiment template.

So we can say, this is what success means and this is what failure means. And then I would really, really challenge designers on, on failure. I said, so if this thing fails, does that mean that you’re gonna kill the future? And I’ve actually [00:56:00] killed features before and it hurt when you have to kill features.

And so we talked about what failure means, right? Basically, you shouldn’t abdicate research thinking to researcher because they are not intimately, you know, they’re not intimate with the problem. You’re living with a problem. You’re living with all the minutiae requirements. You should start with something and then researchers come in and help you to add rigor to your process.

But I guarantee it, if you go through multiple cycles of that, you immediately become better and you immediately, like, not fall in love with, you know, your, uh, concepts, you start to think about, you know, the problem in a much broader point of view, you start thinking about outcomes, segmentation, and you, you start thinking more like a scientist than a designer.

So I. Wish all designers would do that because that was what started it for me 

Vy: because that’s you know back in the day It was one ux designer You did it all. Um, I feel like maybe that’s where [00:57:00] you know would correct a lot of the mistakes But again, uh the time will show i’m really appreciative of your time It’s been amazing to hear from you and where could we guide people would twitter be the best place?

Where would you want uh people to find more about you? 

Ha Phan: They can go on twitter. I’m not You I’m not trying to build a brand or anything. I’m just, like I said, I’m on there because it’s mental diarrhea. I’m just like spilling out, you know, uh, the outflow of my brain every day. But when people contact me, whether through LinkedIn or Twitter, I, most of the time I answer almost every question.

Vy: Amazing. Awesome. Thank you so much. Huh? 

Ha Phan: Yeah, it was really nice meeting you.


Experience designed newsletter

Twice a month I send out 5 new things you should know about: strategy, design and tech innovation, and always at least one out of the box thing. Additionally, a roundup of featured vaexperience content you might have missed.

Free, no catch, unsubscribe anytime.