让产品开发更人性化 | Adriel Frederick(Reddit, Lyft, Facebook)
Humanizing product development | Adriel Frederick (Reddit, Lyft, Facebook)
Adriel Frederick: There are probably, I call them techno utopians who would say, feed all data to the algorithm, give it an objective, and it will do the right thing. And I was like yeah, the reason that falls down is the algorithms don’t understand long term effects often, nor do they understand how people might respond to it, nor do they understand your intent for the product, and I think it’s really important for product managers to play that role. That is our job. When you are working on algorithmic heavy products, your job is figuring out what the algorithm should be responsible for, what people are responsible for, and the framework for making decisions.
Lenny: Welcome to Lenny’s Podcast. I’m Lenny, and my goal here is to help you get better at the craft of building and growing products. Today my guest is Adriel Fredrick. Adriel is a VP product at Reddit where he focuses on incubating and scaling new products within Reddit. Before that, he was director of product at Lyft where he led the marketplace teams and the pricing teams over the course of five years, and before that, he was an early PM at Facebook where he spent four years leading the user acquisition team. Adriel is one of these incredible product leaders who’s way too under the radar because he doesn’t spend all day on Twitter and instead is executing and building great products. One of the goals of this podcast is to highlight incredible product leaders who you may not be aware of. And Adriel is a great example.
Lenny: In our chat we talk about the origins of growth hacking, how to get better as a product leader, ways to increase diversity at your company, what it was like to work on Facebook’s growth team early on, the future of AI and a lot more. It was such a joy chatting with Adriel and I am really excited to share this episode with you. With that, I bring you Adriel Frederick.
Lenny:
Lenny: Hey Ashley, head of marketing at Flatfile. How many B2B SaaS companies would you estimate need to import CSV files from their customers?
Ashley: At least 40%.
Lenny: And how many of them screw that up and what happens when they do?
Ashley: Well based on our data, about a third of people will consider switching to another company after just one bad experience during onboarding. So if your CSV importer doesn’t work right, which is super common considering customer files are chop full of unexpected data and formatting, they’ll leave.
Lenny: I am 0% surprised to hear that. I’ve consistently seen that improving onboarding is one of the highest leverage opportunities for both signup conversion and increasing long term retention. Getting people to your aha moment more quickly and reliably is so incredibly important.
Ashley: Totally. It’s incredible to see how our customers like Square, Spotify and Zoro are able to grow their businesses on top of Flatfile. It’s because flawless data onboarding acts like a catalyst to get them and their customers where they need to go faster.
Lenny: If you’d like to learn more, get started, check out Flatfile at flat file.com/lenny.
Lenny: Adriel, welcome to the podcast.
Adriel Frederick: It’s good to be here, Lenny. Thanks for having me, man.
Lenny: It’s absolutely my pleasure. I actually found out about you through a guy named Jules Walter who we both know.
Adriel Frederick: Yeah.
Lenny: He’s a PM at YouTube and I actually asked him, who should I have on this podcast that is maybe a little bit under the radar that is just amazing and immediately he suggested you. And so I’m really excited to be chatting.
Adriel Frederick: Man, that is hype praise coming from Jules. Jules is my boy. I love him. He’s such a great guy. Awesome product manager and dedicated to the craft that I just like being in his presence.
Lenny: Yeah, and we’re going to get him on this podcast at some point. He’s busy with some kind of secretive project that we can’t talk about. He’s scheduled, I could talk about Jules all day, but he actually has the 10th most popular guest post on my newsletter still. How about that?
Adriel Frederick: Wow. Yep, he’s awesome.
Lenny: Anyway, enough about Jules. So to give listeners a little bit of context on yourself, can you just give us a 55 second overview of all of the wonderful things that you’ve done in your career?
Adriel Frederick: Ooh, we’ll do real fast. So big highlight about me is I’m originally from Trinidad and Tobago, an island in the Caribbean. Came to the US for college, double E, got seduced by consulting and did that for a couple years, worked in oil and gas, electric power, heavy industries, loved that stuff, but also liked writing code on the weekends for fun. So I thought I should move into tech and I did. Worked at Intuit and helped develop their first iPhone app, which was a thing back in the day. Worked at a startup growth team at Facebook for four years working on user acquisition, which was really fun and a good kind of strong formative experience. I had. Quick stint in biotech and then worked on marketplace at Lyft. So rider pricing, realtime driver incentives, matching rider with drivers, and then a lot of the operational tools that we use to manage our marketplace. And so that’s a bit of my journey in maybe 45 seconds.
Lenny: That was great. I don’t have a timer for these, but that sounded right. So we’re going to talk about a lot of the things that you’ve learned along the way at all those places. Can you also share what you do now?
Adriel Frederick: Awesome. Yes. So I’m the Vice President of Product Management for Reddit X, which sounds like we’re out launching balloons into space, but that’s not exactly what we’re doing. We’re more of a team at Reddit that’s thinking about the evolving the modes of interaction with Reddit. So content, temporality, the audience that you’re talking to, if you think about it, Reddit is primarily about asynchronous conversations between anonymous strangers about shared interests. Sometimes other people find answers to their questions on Reddit, but we’re looking into and on the X team, evolving that to look at problems like helping people communicate fast during easier about shared interests, perhaps changing who they’re having conversations with. Maybe it’s about something other than a shared interest or maybe they have something else in common that brings them together. Maybe bringing video, audio, and other media into being a part of the product and playing with permanence and things like that.
Lenny: Whoa, I felt like you were going to go into a metaverse direction. Is there metaverse angles to this?
Adriel Frederick: Not really. I think we look at that as a potential technology, but our primary focus is a lot more on, I see modes of interaction and platforms that are a lot more at scale today.
Lenny: Got it. Is there anything coming out in the near future we should be looking forward to? I imagine you can’t talk about too much of what you’re actually working on more concretely.
Adriel Frederick: I think there’s a few things that we’ve done recently that have been fun. We have an avatar marketplace that we’ve been working on recently where creators have been able to make art, put it up for sale on Reddit and make that available for other folks to buy and use. And that’s been performing amazingly well. The underlying technology behind it is NFTs and we thought that technology was really important to use because it gives a creator a public way of acknowledging their rights to a piece of content. And so they have some form of IP protection, especially in a marketplace where you’re doing something like selling digital art, we felt like that was incredibly important. I think the technology behind NFTs has been used for some really nefarious things, but I think we’re still in the infancy of using these technologies appropriately. There’s a lot of terrible use and a lot of uses that are wastes, but I think there’s some gems in there and we’re hoping to find some of those.
Lenny: Sweet. I will avoid getting pulled into a web3 rabbit hole here, but that is very cool. Something I wasn’t planning to ask about, but I’m curious because I was just talking to some other guest about this topic is the idea of these kind of R and D ish teams at larger companies and companies that have been around for a while. I know you’re relatively new there and that’s kind of maybe a new thing, but I’m curious, is there anything you’ve learned about how to set up teams like this and investments like this, these kind of long term horizon bets R and D teams?
Adriel Frederick: Yeah, I think it’s really good coming to this from being on the other side of it. If you think about where I’ve been, I’ve been on growth and on marketplace, which is as far as you get from seeing where on the new stuff kind of team and what I’ve seen happen a lot is organ rejection, that this thing looks so different to the rest of the body and the rest of the organization that you get some form of rejection of the ideas entirely. So I think what I’ve learned is a few things.
Adriel Frederick: First is the rest of the company needs to see what you’re doing as being core and critical to the mission. It can’t seem like these guys are just playing off in a corner on something that isn’t related to what we are doing every day. Because I think that leads to some of the resentment because you can imagine any team internally is fighting for resources and they look at this group as having resources that they can’t get. They’re like, Oh we got to get rid of that because they’re not helping us do what we are here to do. So you have to be part of the core mission, otherwise you’re going to have problems culturally with that. So I think that’s one thing.
Adriel Frederick: The second is it has to be everyone’s success. So if you end up doing something on one of these R and D teams, it shouldn’t just be the R and D team that wins. Everyone should feel like they win. And that is kind of relates to that first goal I was talking about. I think the third is you have to set up the work that these teams are doing such that people don’t believe all innovation is going to happen on that team. It can like, okay, we’re just stuck with the operational stuff and they’re getting to have all the fun. Other teams are still going to innovate, but maybe we’re taking on something that other teams don’t have capacity for, that the organization needs and it’s part of the core mission. And so I think that’s been a lot of what I think about when working on and setting up these teams is to make sure that we are part of the organization and everyone wants to hug us as being, yes, you are one of us not, you kind of need to go off in your little corner and behave.
Lenny: Amazing. That is really helpful. So just to kind of recap, you want it to feel like it’s core and critical to the company. You want it to feel like it’s everyone’s success. It’s not just, oh, Adriel over there is doing great, but we’re stuck with these terrible hard problems. And then this idea of not all innovation’s going to be just coming from that. We can all innovate but they’re just working on this one specific innovation.
Adriel Frederick: Yeah.
Lenny: Awesome. Okay, great. Another question I definitely wanted to ask you. So you said you were born in Trinidad and Tobago, not something that you hear very often in tech. I’m curious your background and your journey to what you do now, how does that impact the way you lead, the way you build product, the way you just think about your career broadly?
Adriel Frederick: Yeah, it’s not something that I really think about consciously, but it affects me every day and it’s tough not to see it in retrospect. I was the first Black product manager at Facebook.
Lenny: Oh wow.
Adriel Frederick: And so it’s tough for me to not see that having some effect on what was built or how things were built or on me. So it’s pretty meaningful. But I think one of the ways to see how it effects things is actually just to understand a little bit about Trinidad. It’s kind of its own little unique animal. So Trinidad is an island in the southern Caribbean all the way up bottom next to Venezuela. It’s a really diverse place. So ethnically it’s 35% Indian, like from East India, 35% African, 25% mixed, and that last 5% is everything under the sun. European, Chinese, Arab, et cetera.
Adriel Frederick: And then for religions it’s about 60% Christian, but then that’s a lot of different forms of Christianity in that 60%, 20% Hindu, 7% Islam. The media diet is a mix of British and American TV. You have a really broad range of incomes, but then schools are a melting pot, so you don’t have as much of the kind of class and income segregation with schools that you get in most of the west. And so when you have that kind of a melting pot of ethnicities, religions, media consumption and kind of socioeconomic status in one place, you learn a lot of them because in school you’re mixing up with everyone. One of the jokes we have is Trinidad probably has the most public holidays of any country because you have to celebrate everyone’s holidays from Diwali for Hindus, Eid al-Fitr, to Christmas, I have friends who were fasting for Ramadan. I know a lot of the names of Hindu gods and I always love shocking my coworkers with my knowledge of this stuff. So that gives me this really different perspective that shows up at work.
Adriel Frederick: So I’ll give you an example. Something I’ve noticed in almost everybody I worked with in tech, as we work on mobile devices, people make an assumption that one phone number plus one device tied to one person. And growing up in Trinidad, I just knew that wasn’t true. Someone who is using a prepaid phone could have their number change all the time, so that one person could have multiple phone numbers just because they were using prepaid. You have phones for two sim cards, that was pretty common. And a phone is and definitely was a really expensive digital device. It’s a computer. So it was often shared and people couldn’t just have one for themselves.
Adriel Frederick: So when I was working on user acquisition and designing registration for Facebook, that knowledge was incorporated into the design of the product in ways that I think other companies not caught onto yet. And I know for a fact that a lot of that thinking that went into designing how you think about a phone number and a device and it’s use among one, it’s how to say pairing with an individual, has been helpful for Facebook’s growth back then and even after I left, I know that’s still been providing benefit. So that’s a simple example of how just being in that environment and soaking up information could help product design in a way that I think wouldn’t have happened if I and others like me weren’t there.
Lenny: You said you were the first Black PM at Facebook. I didn’t realize that. How many PMs were there at that point when you joined?
Adriel Frederick: Oh man. I remember we all fit into this conference room called Canada and that was probably my second week. It was probably about maybe 30 of us in there.
Lenny: Wow. Yeah. Is there anything that you learned from that experience about just how to help with diversity at a company? Did Facebook do this well? Have you seen other companies do this better? Is there something you could share there for folks that are trying to work on this?
Adriel Frederick: Man, that’s a tricky one. So there’s two parts of that, what was that like for me? I think it went quickly from being a little bit of imposter syndrome, like that day when I was sitting in that group was like, Dude, I’m one of 30 people working on Facebook. What am I doing? I’ll belong in this group. This is crazy. And then I recognized after talking to a lot of the other PMs and the engineers, it’s like no, no, no, they want me for what I know from my perspective because they’re really trying to grow this product globally. And being this guy from Trinidad working on growth with the perspectives I just mentioned was appreciated. I think I was lucky enough to be on the growth team and having leaders on that team really valued diversity. I think about some of the teams I was on and they were awesome. I joke about them sometimes.
Adriel Frederick: I remember being on a team where I was a Black Trinidadian product manager with a female Israeli engineering manager, a female Brazilian tech lead, then the rest of the engineering and design team was from all over the world. We had Russians, Chinese, some folks from Slavic countries and it made designing products fun because a lot of times when you’re building a product and you want to think and get into your head of your customer, you have to go out and talk because you don’t necessarily get them really well. Man, we didn’t need to on that team. We would just argue with each other. We would think about how our friends would use it, how our cousins would use it, and we are covering a broad swath of the world when we were arguing about how to design a product. But I think the original leadership of the growth team, I think starting with Chima but then followed up with Javi.
Adriel Frederick: [inaudible 00:16:51] that and kept bringing in that diversity of, again, ethnicities, religions, cultures from all over the world so that you could actually build a product that way. And it just makes you so efficient because an argument that might take two weeks to resolve because you have to go recruit a panel of users and talk to them and figure out what’s going on. We kind of knock out in 15 minutes just throwing it back and forth with each other. And I can’t stress how much that’s important for building products that you want people across the world to use. You got to have your teams look like the world, it just makes you so much faster. It’s not perfect. You still have to go out and talk to folks because we still have our own kind of monocultures that form that we need to get out of. But it helps a lot.
Adriel Frederick: To your second point about diversity and how to foster it, man, from the beginning of my career at McKinsey to today at Reddit, I’ve been in rooms where everyone’s asking the same questions about how to fix it and here’s what I’ve seen work. When you recognize that you get business value from it, then it all of a sudden becomes something that you look out for and you take care of. That’s it. And there’s definitely a lot more to it, but I think when it goes from frankly something people feel they need to do to be PC or for cultural reasons or because they’re getting social pressure to do it to something that you really recognize concretely, no, I get value from this and you are willing to take the other steps to have a culture at your company that utilizes it, then it becomes easy because when you bring folks in from diverse backgrounds, they retain and that’s always the number one step to growth as you will know.
Lenny: Retention.
Adriel Frederick: You have to retain them, you have to retain diverse talent. And so you have to have an environment that values it, cares about it and uses it and rewards it because it’s part of the core system of the company. Then once you have that working, it becomes a lot easier to recruit because people see you valuing it and bringing it in and wanting it. And it’s not just lip service that you’re paying. That’s been what I’ve seen to be true in all the conversations I’ve had on the topic.
Lenny: That first piece is interesting that it answers the second piece, which is the point you made about how having a large diverse global group of employees early on, especially for a company that’s trying to go global and international is so powerful, you just save all this time. You don’t have to necessarily interview people that you don’t already have.
Adriel Frederick: Yeah, there’s something that feels like the approach to not doing it that way feels colonial. It almost feels like we’re a group of people sitting down in this tower in this country, in this relatively sterile environment and don’t worry, we know exactly what you need and these are the parts of the world. It just doesn’t work well. So doesn’t feel right to me also.
Lenny: Yeah. Awesome. Thanks for sharing all that, that was really helpful. There’s another topic I definitely wanted to spend a little time on, which is this interesting trend that I noticed when I was looking at your LinkedIn and your background. You worked at Facebook, Lyft, Reddit, and interestingly, they’re all very in the news full of controversy type places. People like to tear them down and show all the reasons that they’re doing bad things to the world. And I imagine as a PM that’s just a challenging place to be and the fact that you’ve been at three different places. I imagine you’ve learned some stuff about how to operate as a product leader at companies full of chaos and fires and bad PR and things like that. So is there anything to share about what you’ve learned there?
Adriel Frederick: I think the biggest thing is that as a PM you are a leader. You have to provide a buffering or damping effect on the team, and that goes two weeks. Sometimes we’re doing stuff that everybody thought was amazing, this is the best thing we’ve ever seen, and you kind of got to bring people back down to earth and go, Look, that was cool, but we got a lot more stuff to do. We are really not there on providing the value that we want to provide to people in the real world, so slow your role and recognize that there’s a lot more to do. And then when it’s terrible and the press is telling you that you’re the worst thing to ever happen in the world, you kind of have to also go back and say, Guys, slow down. We’re not anywhere near as bad as what they think you see and know what we’re doing and they’re going to misunderstand us sometimes. And so pull your team up at this point in time and keep charging forward with the mission.
Adriel Frederick: I think some controversy is necessary and so I may be at a different point on that one. I don’t think you’re going to have any meaningful influence on the world without changing some pattern of behavior. And if you’re changing a pattern of behavior, there’s somebody who’s invested in that pattern of behavior and that’s going to create some conflict. The most fun news stories to read involve conflict, so that’s always going to make for a great story and put you in the press. For Facebook, it was traditional media and other social networks were one side of the fight and then Facebook was the other side of the fight and then it became other tech companies now and that always makes for a great story. With Lyft it was taxis and unions, and so you have to recognize that you’re always going to have some bit of a challenge.
Adriel Frederick: Now, the really hard part about dealing with this is understanding what criticism is valid and how much of it is just because the source of power is being changed. So I’ll give you an example. Let’s say with Lyft, rich medallion owners in New York, I had no sympathy for them when they were complaining about trying to ban Lyft because when I was in New York City putting my hand out to get a cab, they were tried right by my Black ass. So I’m sorry, I’m like, I do not feel that much empathy for them, but I think there were really legitimate complaints about the structure of driver pay that were coming up and that were behind I think some of the complaints and some of the big stories in the press and some of the big kind of legal action that was taken. Paying for pickup time when a driver’s on their way to pick you up or when they drive somebody far out of town and they have a dead head to come back into a place where they can work, that’s real.
Adriel Frederick: That’s a real problem that I think we got called out for that we weren’t paying enough attention to and it got us off our ass to go fix it. I don’t think we’ve, and I say we, but I’m not there, I don’t think the problem has been fully solved, but I think as a PM listening to this, you kind of have to find the truth behind it and try to find a way to work on that and not get too lost in responding to the specific criticism. And so to walk this line between kind of going, yeah, some of this controversy is just part of the game, versus like, nah, this is really valid. Dude, to figure out where that is, you got to do what is so cliche, but you got to stay close to your users. And so to give you an example of how I did that, when a lot of the complaints were happening about driving on Lyft, I drove, I would just pick up the car and I would get out and I’d go drive and I’m like, let me me go feel this for myself. Let me go see what these guys are talking about.
Adriel Frederick: But I can give you a story about Rick. I still remember this drive I did with Rick in Berkeley. So I’m at home, I just get in the car, I turn on the app, it’s time to go driving. I get up ping 15 minutes away and I’m thinking do this, I go do this right now, this guy might cancel on me, I’m not really getting paid for this, but maybe the ride is worth it. So I drive on over, I’m dodging traffic, pedestrians, drunk college kids, stop signs, I make my way over to Rick, he’s coming out Chez Panisse, and he’s about 80 years old, jumps in the car and then I pushed the button to figure out the destination and it says the ETA to the destination is two minutes. So I was like, “Hey Rick, you get this right? What’s going on?”He’s like, “Hey, I had a little bit too much to drink. I’m worried about breaking my hip, so that’s why I called a ride.” And so I went from wanting to curse Rick out for making me drive 15 minutes to come pick him up to feeling like, all right, no, no, no, there’s real value I’m providing you in driving him just two minutes. But I recognized that wasn’t embedded in the structure of paying. Rick would’ve been happy to pee for my 15 minutes to come pick him up, but we weren’t one, giving drivers compensation for that, nor were we finding a way to pass that through into pricing for Rick. It’s a much more difficult problem than it seems from that simple example. But it clued me into why drivers were complaining. So then I went, got it. I understand what we need to do. So when there were all the PR was going on about AB5 and prop 22, I was out driving and I was out sitting with the team trying to figure out how we’re going to design our product that helps pay a driver for this.
Adriel Frederick: It still keeps prices reasonable for users, doesn’t create bad incentives where you end up with riders not getting picked up when they really need a ride because I didn’t want Rick to break his hip. He still needs a price that makes him feel like it’s okay for him to take that ride, and finding a way to balance this out is actually more complex than you might think. And that’s what I stayed focused on whether prop 22 passed or not, I was ready for either side with a solution that was going to work for rider and drivers. That was the job. And so I think for PMs it was so easy to get sucked into the press and it’s like, Yo, plan the work, work the plan, go back to your job. That’s what you’re supposed to do. Solve for customers in the middle of this. And then you figure out how to communicate it well.
Lenny: What I love about that strategy is it also helps you see that it’s not everybody that is worried about something. I think of Airbnb, like all hosts are pissed off about this one feature, there’s going to be a revolt. And then to your point, you talk to some, like nobody even knows about it. Nobody cares. Everyone’s fine.
Adriel Frederick: Yeah.
Lenny: And so this, there’s so many benefits to what you’re talking about doing, which is talking to customers, not just paying attention to the loud voices.
Adriel Frederick: Absolutely. I also have empathy for reporters too. The story that with the headline, some Airbnb hosts are annoyed by the-
Lenny: Right.
Adriel Frederick: I mean, come on, this is not a great headline. I recognize that they have a job to do and sometimes they hold people accountable and sometimes they’re getting people to read a story that maybe has a bit of hyperbole in it. And so they have to do their job and I have to do mine too.
Lenny:
Lenny: Well you shared this really heartfelt story about Rick, what’s your most stressful memory of working at Lyft?
Adriel Frederick: I think the most stressful time was when I had to unwind a bad product I did and actually make a better version of it. It was really a pricing algorithm change, it was something behind the scenes that nobody would really see, but this was a fairly big initiative that we worked on. We had experts in revenue management to work like PhDs and the people who wrote the textbook on the subject helping advise us on this. We build this model, you launch it and you’re expecting this big change. And it goes poof, just does a little bit. And then we work at it and we work at it and we work at it and eventually we get it to be good and it works really well in three cities. We start rolling it out to more cities and it’s a pain in the butt to roll it out to more cities because it’s super complex.
Adriel Frederick: And eventually we get it rolled out to maybe a hundred cities and then someone says, “All right, cool, I want to change prices.” And oh, we struggled for months to implement price changes and man, the sentiment around this product was just rough for a while. And I remember being on our walk after a particularly bad week of this and trying to figure out what I was going to do about this thing. Do we stay the course? After a while, the answer was kind of simple even though it was emotionally difficult. And the answer was like, Yo, we got to rebuild it. There was no answer where we couldn’t have a product like this. We needed some ability to be able to influence prices so that we could actually run an effective marketplace. The current solution didn’t work. It wasn’t as operationally flexible as we needed it to be because we didn’t consider that requirement when we were building it and we got caught up in the kind of algorithmic complexity and sweet sauce of it.
Adriel Frederick: And so I recognized that we just needed to own up to it, tell everyone we didn’t get it right and we needed to come at it in a different approach that was actually more flexible operationally. And we did. I think the big learning, at least in that business was you have to think about operational requirements and operational control as a first order requirement. And I think when a lot of us were building product at a lot of the other consumer internet companies, you didn’t have to think about operational control. You gave the algorithms an objective, you feed them some data, you let it run, you observe it and make sure it’s doing nothing crazy and you tweak it, but you didn’t need to have day to day operational and strategic control over the product and we just needed to snap our brains into being able to put people in the loop with the algorithm.
Lenny: For folks that haven’t worked at a company with this kind of on the ground ops team, can you just unpack what that is? Like operational control, what does that actually mean in practical terms?
Adriel Frederick: Okay, so I’ll give you an example. So back in the days Lyft is in 300 cities, probably roughly across the US, and in every single one of those cities you don’t have exactly the same pricing. It’s a little bit different. And so sometimes you might need to make a change seasonally because traffic gets worse or because fuel prices were different or because there’s a new tax or because your competitor did something that you need to respond to and your algorithm cannot see this. It has zero visibility into this. And so you need a person in the loop to not only give that visibility but also to make a decision about how you respond. Because I think also, let’s say you’re in Chicago and there is a snowstorm and you need to change the way, let’s say you need to update pricing so that it handles the increases in driver pay that you need to create to get people out during a snowstorm. You don’t know exactly how you want to respond, every snow storm’s different and a person has to make that judgment call and provide the right information to the product to be able to get to utilize it. Now algorithms were handing a lot of that and they could generally respond, but to be a lot more precise and needed a person to help handle that, to make that call.
Lenny: Got it. Cool. Thanks for sharing that. So you’re making this point about when you’re at a company that has a big operations component and obviously the core central product team, you were sharing some learnings about what you’ve learned to work in that environment. So yeah, I just wanted to come back to that.
Adriel Frederick: For sure. For sure.
Lenny: So the main thing you said is just ops as a first order component when you’re designing the software. Is that the big learning?
Adriel Frederick: I think it’s not just treating ops as the kind of first order requirement. The bigger picture for me was when I look across my career is the algorithms need people to help make judgment calls. And so I saw it really, I got a heavy lesson in it at Lyft, but when I look back I recognized it was there at Facebook too. It just wasn’t in my domain. There is always a judgment call that has to be made between how often are there going to be ads versus how often are we going to show organic stories from your friends and family? How often are we going to show content that you might be interested in that’s not quite in that group? How often might be want to show you things that help you find your friends or help other people find their friends? And that is a judgment call that varies for different markets and different situations and there may be algorithms behind the scene that are making that call for every single person in real time, but there still have to be people applying some strategic judgment to that.
Adriel Frederick: And I wasn’t in the position of needing to do that at Facebook, but once I saw how much I needed to do it at Lyft and I kind of looked back at history, I saw that it was there too. But I think there are too many people who don’t see this and believe that there’s an algorithmic solution to everything. I think as a product manager and especially product managers working on systems that are heavy on machine learning or operations research and optimization, to think about where you want a person to make a decision and where you want the machine to be off to the races and to think about that as a product design problem because there actually is actually a computer interface that you have to think about there. You need information about what’s going on, let’s say at Lyft, what’s going on with my market? How long does it take for somebody to get picked up? How expensive am I versus the competition? What are my goals in this market and how am I performing today with that? Give somebody information, but also give them the tools to execute the right decisions without creating trouble. And that’s like a product design problem, that’s a first order product design problem like anything else that you have to think about. And I’m not privy to it, but I would guarantee that there are people thinking about those same kinds of problems at other companies.
Lenny: That reminds me, I was just listening to Zuck on Joe Rogan and he made this point that when you look at a post, you can add a little emoji reaction and you can have a little angry emoji reaction. And he made the call that we’re not going to use the angry emoji reaction in our algorithm in any way, we’re just going to ignore that because naturally you’d be like, okay, people are angry, that’s interesting. Let’s show that because it’s interesting to people. But he specifically wants to avoid anger and facilitating anger probably because a lot of the feedback that they’ve gotten.
Adriel Frederick: Exactly. And I think they’re probably, I’d call them techno utopians who would say feed all data to the algorithm, give it an objective and it will do the right thing. And I was like, yeah, the reason that falls down is the algorithms don’t understand long term effects often, nor do they understand how people might respond to it, nor do they understand your intent for the product. And I think it’s really important for product managers to play that role. That is our job. When you are working on algorithmic heavy products, your job is figuring out what the algorithm should be responsible for, what people are responsible for and the framework for making decisions.
Lenny: Is there an example that comes to mind where you did that or didn’t do that well or someone on your team should have? Just something to make it a little more concrete even.
Adriel Frederick: Let’s assume that you are a person working on pricing and you say like, great, I have an objective that is I would like to win market share in a region. And you left that to an algorithm to say I need you to optimize prices such that you maximize market share, but what would the algorithm do? Drop your prices to the floor. All the way to the floor, and then you don’t make any money. Okay great. So then you say, okay, what’s the next step of that? Let’s give it a constraint. Let’s set some target that we might want to have for how little profit that you might be willing to take. Okay, go do it now.
Adriel Frederick: What if the guy on the other side is doing the exact same thing? Both of you will hit your constraints and then the game will stop. Okay great. So now it then flips to, oh we have to choose where we want to win. And so I think one of the things we did that I’m particularly proud of is building products that help people see and understand that game a little bit more and decide where they want to. I think the first few pieces of that are not shockers, but that conclusion at the end where you get to, oh wait, I need to create a tool that gives people information to then decide how to play this game is actually what’s critical.
Lenny: Interesting. So kind of what I’m hearing is a lot of the work is giving humans more information versus giving machine learning algorithms more information and there’s a lot more leverage potentially there, giving humans more ways to tweak and dial.
Adriel Frederick: Let me refine that a little bit more. It’s more about giving people the information that they can use for decisions that they alone are good at and giving machines the power to amplify a person’s intent. So one of the ways I like to think about it is all software in any form including ML, is just a tool like a screwdriver and you could try to put a flat head into a Phillips and maybe it’ll work a little bit but it’s better to use a Phillips screwdriver. And we’re tool designers generally and especially in part development function, you figure out how much do I put into the tool and how much do I leave it up to the person and I give the person the ability to choose what they want to do. I give them a screwdriver, a flathead, a Phillips, a torques and you let them decide how they want to use the tool for the application at hand.
Adriel Frederick: And so going from that analogy to concretely with ML you say look, machine learning’s going to be amazing at optimizing for a given objective, but it’s not going to understand the constraints or strategic choices I need to make. The constraints and strategic choices that we need in the external world are always going to have to be decided by a person. You make that incredibly easy for people to do and intuitive for them to do and then you go that algorithm can then amplify their effect by making decisions on hundreds of thousands, potentially millions of individual decisions to take that person’s intent and amplify it given all the information that they can learn in that single context. So I think about it as designing an interface and make it an extension of yourself rather than a black box on its own that you just need more information to. Is that helpful?
Lenny: Yeah, it makes me think about a neural link and what Elon’s trying to do, I don’t know if this is how he thinks about it, but the [inaudible 00:39:56] guy described it as Elon’s worried that AI will take over at some point and so he wants to build a tool that connects straights our brain that can access the power of AI to kind of have a chance against just a rogue computer in the future.
Adriel Frederick: Even then you’ve got to make sure the person is still in control. I hear that thought and I go, okay, you build the interface but then who’s in control? Is the person still in control or did they become a slave to the machine and you just made a better interface to make them a slave?
Lenny: Oh shit, we’re in trouble.
Adriel Frederick: I am not yet as worried about these visions of them taking over. Thus far and maybe I haven’t fathomed what they can do, they still seem like tools that need our guidance to be useful. Even the most amazing, we’ve been seeing the image generation and I’ve seen the cutting edge like text generation stuff. They can fool you into believing that they’re like near human capability, but there is a lack of decision making and judgment that I see coming out of them. I see them as being again, extensions and useful like text generation algorithms. A lot of them can’t write a paper for you and that’s what I think people are scared of because it still requires your judgment to decide. Now when you decide what the salient topics are in something you’ve read, let’s say you’re doing a book report, you’ve decided what the topics are, it can help you write the paper faster for sure, but it can’t write the paper for you. It can’t choose the topics that your background and history and interest find useful or compelling to tease out.
Lenny: This isn’t where I was expecting our conversation to go, but I’ll add another thought here because it’s interesting, the way I think about it is there’s nothing magical about our brain and so if that’s true, why isn’t there a world where we could just completely simulate it? Sam Harris talks about this a lot that it feels like once you get close then it could just accelerate so quickly beyond human potential. It’ll start from 20% as good as a human to 40, 50, 60 and then it goes to a million times better. It can move so fast beyond us very quickly. So I think that’s where a lot of the, not that I’m afraid of this, but I feel like that’s where a lot of fear comes from. It could just dolly coming out and co-pilot just like holy shit.
Adriel Frederick: Yeah, our brains are good with linear thinking, not exponential. So I’ve heard that argument that like, yes, this is increasing exponentially and you can’t fathom it. I’m like, yes, that is definitely potentially true. Completely see that possibility and recognize that I have that cognitive defect in being able to understand it and even if it’s a remote possibility, we should be paying attention to it. So I’m all for paying attention to it given the, let’s just say the high cost of a low probability outcome is still a high cost and so it’s still worth paying attention to.
Lenny: Yep. Okay, good tangent, I wanted to chat about your learnings at Facebook. We’ve been chatting about all these other places and especially about growth, just stuff you’ve learned about growth and growth hacking and I was thinking about this interesting world that Facebook is in slash Meta where on the one hand when they started, and I’m talking about growth hacking, like Facebook did a lot of growth hacks, emailed all of Harvard, he had all these interesting dating thing happening and got a lot of controversy and it was all these interesting tactics to start Facebook, but now people use Facebook to growth hack and grow like Zenga famously, a few other places. So all that to say, I’m curious, what have you learned about growth slash growth hacking from your time at Facebook and other places?
Adriel Frederick: I think growth hacking as traditionally assigned, finding those small changes you can make to a product to give you outsized impact, that is absolutely valuable. Where I’ve seen people get lost is they assume that if you do that alone, it will work. You can grow your way into something successful if you just find those few hacks and patch them together. And there’s something about that that I find disrespectful to the people using the product. It’s like you assume that they have no intelligence and they won’t catch on to what you’re doing eventually. The old saying fool me once, fool me twice, it kind of applies. So if you don’t have a product that’s providing real fundamental value to people, you can be a one hit wonder and have a flash in the pan and growth hack your way into something that might last for a few months but people will catch on it and then it’ll disappear.
Adriel Frederick: So I think that stuff is helpful, especially early on to get your initial traction. But you got to have something people and want to continue using. And when I think back over the products we did that really moved the needle, they were all things that just focus on the marginal user and figured out to make the product easier. It’s easy to get seduced to thinking that there is a fast secret way to do it. And I’m like, no. The vast majority of it was just hard work and finding ways to solve the real problems. And what are those real problems? They were pretty damn simple but we just grinded on them for a long time it just stayed on it. One, make it easy to find the product. Number two, make it easy to get into the product. Three, stupid easy to find your friends. And then once you did that you were off the races and those were the things we were doing over and over again.
Adriel Frederick: I think another big piece of it is reminding people that there’s something interesting here and building the habit of coming back to the product and it was also part of it, but we just grinded on those few things over and over again. And some of the really big wins weren’t hacks, they were just paying attention to little details. I’ll give you an example. I remember sitting one day thinking about how to help people find their first few friends and we would do this thing where we’d have recommendations, if you could get one or two friends, you’d be off to the races and we could find you more people that were in that same friend group. So I thought about the way the people you may know algorithm worked, I get one or two friends, they would find your mutual friends and then would help find you more of those kinds of folks.
Adriel Frederick: And I was like, you know what that does is it spirals you down one friend group but it doesn’t get you all your other friends. I remember just looking at somebody using the product and recognizing that we were only taking them down this one path. So I was like, man, how do I see all your friend groups? And so we had this idea that we came up with that would do it. I’m not going to let that one out. And it was like game changer, like absolute game changer. Especially for users helping them find those first few friends in a few different friend groups, which then meant we could get you down one group and another and just continue building out that graph just by using recommendations. Because we had a great tool for seating it and that was not easy. That was not a hack. That was hard work.
Adriel Frederick: I also remember one of my favorites is something Tom Allison did, Tom Allison I think now is responsible for the Facebook App, and when he was working on the engineering manager for one of those teams, there was a change we wanted to do to one of these algorithms and it was a bitch, it wasn’t a hack and it was going to take a few months to pull off and Tom just hit it in a corner. He just let everybody know that we were really going to change the way this product worked. He had a really smart guy working on it, [inaudible 00:47:03], and they just hid off in a corner, rebuilt the product in the way it needed to be built to make it easier for us to operate it and scale it and then put it up there. I know of course they crushed it and they were incredibly modest about it, but it was not a hack. And it came from them looking at this deep problem of finding that thing that mattered and then saying, we need to make a fundamental change to make it easier to recommend friends to folks and just grinding on it.And so one of the things I recommend for people when they’re thinking about growth for their product is figure out what the core actions are and then grind on them. Think about removing them, removing friction and some of them. But just keep staying at it and as you grind on it, you’ll do little hacks. You got to figure out how to put the right text in the button and get it above the fold, create the right copy. All the things that we traditionally associate with growth marketing, you got to do those things. But to me that’s stable stakes and just doing good product communication with your user, but then you got to think about this person who can’t yet figure out your product and it’s trying to take this action and making it stupid easy for them. I got a million more examples of that one, but that’s the game. It’s not just finding some trick to [inaudible 00:48:18] a site.
Lenny: I love that. The way I think about this that I’ve heard well described is just there’s no silver bullets, just many lead bullets.
Adriel Frederick: Yes. And a few massive cannonballs every now and then. Every now and then there’s some cannonballs.
Lenny: What’s an example of a cannonball as you think about that?
Adriel Frederick: Sign up with phone numbers, which is now par for the course, that was a cannonball. Getting SMSs delivered to people all over the world. Doesn’t sound glamorous, really hard to do. That was cannonball. Good friend recommendations, another big one. There’s more, I’m not going to go into all of them. What I mean by cannonball here is that there was sometimes some really big fundamental changes you needed to make to the product to make these things work.
Lenny: Got it. So you think about that in terms of investment, not necessarily the impact-
Adriel Frederick: Investment.
Lenny: Impact plus massive investment. Cool. I have so many questions along these lines. Okay, I’m going to pick a couple. One is Facebook is famous for this kind of activation milestone of getting 10 friends or seven friends, whatever it was, like there’s some number friends you got to get and the good things will happen. Were you involved in that? Do you have any insight into how that came to be? Is that real?
Adriel Frederick: That decision came before me. I saw it. I understood the data and I worked on this problem. What I thought was brilliant about that was not the metric, it was the designing it to be understood and communicated. What I think is fabulous about it is that you are talking about it now because it’s memorable and it got people to take the right actions to start chasing the goal. There was literally nothing magic about the number or the date, but basically it was a way of saying like get people as many friends as possible as fast as possible. And if you said that generically to someone, they’d be like, Yeah, I kind of get it, but yeah, I’ll go do that.
Adriel Frederick: When you create a discreet number and a discreet time and there is a concrete goal to chase and there’s a number and a graph that everybody can look at and see, we are going to go make that thing go up, the organizational effect of that is galvanizing. So what I thought was brilliant about it is, and as I’ve heard the stories, this is all secondhand. There was a lot of debate about what the number should be, what timeframe should be, and at some point Zuck just said 10 friends, 14 days, go. And it just got people past the academic debate of like, All right, got it. As many friends as possible, the fast as possible, let’s go.
Lenny: I love that. That’s exactly how I’ve always thought about it, that it’s not the number exactly, it’s just a rallying cry that everyone can just get around and just go, doesn’t need to be this perfect number that has incredibly correlated link to retention or anything like that. It’s just like, this is good enough. It’s directionally I’ll just try to do this. Let’s just go.
Lenny: There are downsides of it. Some of them are really funny. I remember looking at a graph of retention versus number of friends and what actually dropped with 11 or 12 versus 10 because somewhere in code, somebody had done something with 10 friends is the limit to help improve retention. And it shut off at 11 or 12 and it came back up, but I was like, You know what, that’s fine. That’s completely fine. Because if we didn’t get that organizational momentum, that graph would’ve just been lower. So I could take the kink where it drops, it’s fine.
Lenny: You also mentioned this term marginal user and I thought it’d be helpful just to unpack what you mean by that.
Adriel Frederick: For me, it’s a person who is just on the cusp of taking the action you want to take. I’ll give you the concrete example. When working on registration, I would try to find a country where we had a lot of growth, but for some reason our conversion rates were terrible. So we had a lot of traffic, but conversion rates were terrible. And I was like, okay, that’s the marginal user. This is the person who is just on the cusp of coming in, wants to come in as you can see by the traffic, but we can’t get them in. So why? And when you go to the extreme and you find that person who’s the worst, and most likely it was a person on a feature phone on edge, trying to access Facebook in a country that was far from one of our data centers. And then you go like, All right, what’s wrong with this person’s experience?
Adriel Frederick: Let’s go check it out. And you’re like, Oh, you see everything that’s wrong with the product. So then it gives you your list. Okay, the language is probably wrong. We didn’t get that. Are we detecting the country properly so that we can actually get their phone number formatted properly? Probably not. Oh man, it’s far from the data center, so that connection’s slow and they’re on edge. Oh, that’s terrible. And you just see and package up all those, it gives everything that’s wrong and they can just start figuring out what to do with them. Something I caution people against though is don’t use the data alone to figure out who the marginal user is. It’ll give you a clue where they are and what might be wrong, will give you some hints. It’s not going to give you the answer. You have to go watch them to find the answer.
Adriel Frederick: Because I think in a lot of these data driven places, somebody will say, Great, just create a funnel. Figure out all this drop offs of the steps in the funnel, Look at it yourself and then figure out what might be wrong and go fix those things. But what I think happens is often there’s a problem that’s orthogonal to that funnel that you can’t see from looking at the data. And you have to go look at the person and talk to them. I remember one example we had is was watching someone sign up up for the very first time for Facebook in India and they’re about to put their name in it and they ask them, So what name are you going to put in? They’re like, Okay, my full legal name. All right, cool. Does anybody in the real world call you that? No. And I was like, Oh dude, we’re screwed.
Adriel Frederick: If you send out friend requests, it’s not going to get accepted because nobody knows who this person is, and then the reverse, if they find you, they don’t know who this is. And so it was like, Yo, you’re going to look at some problem deeper in the funnel, yo, what’s going on my accept rate? And then you’re going to go tear I that little mini funnel and then recognize that you had a problem that happened ways back. And so when you thinking about that marginal user, you got to go out and look at them, talk to them, watch them use it, try to get into their shoes yourself as much as you can, and then make the call from there from what you do. But that data isn’t going to tell you, it’s going to give you the answer. I’ll just tell you how bad.
Lenny: Wow. I love that connects back to the same advice you gave in all these other contexts of just talked to people. Don’t rely on just this aggregate data.
Adriel Frederick: Now, don’t get me wrong, I’ve built the experimentation platform at Lyft. I’m a guy who loves data and loves using it and looking at experiments. I think it’s just too easy to try to sit in your laptop, pull up a funnel or pull up some charts or look at an experiment results and think that’s going to give you the clue to what to build. It’s a compliment. It’s not the only thing. And I watch people fall into that trap of assuming, especially when you’re working at companies with lots and lots of data, you fall into the trap of thinking that you’re swimming in answers because you have all this data and just need to tease it up. Just go out and talk, you’ll find it faster.
Lenny: I really like this advice of when you’re trying to optimize things, focus on your marginal user. And there’s two parts to it that you talked about. There’s the next most likely person to sign up and then there’s the worst case and going to like them to see all the things that are wrong and that be your north star, make this person successful and make so many more people successful. Is that how you think about it?
Adriel Frederick: Yeah, I do. So marginal user I think is the fun word to think about because you think of families, think about children, the person who’s right on the cusp. But I like to go to the worst. It shows me everything that’s wrong, but the marginal user thinking helps you prioritize what thing to do next. So that person, that example marginal user I was talking about, they’re on a future phone with edge too. There’s a lot wrong that’s just going to be tough. But I might look at that experience and go, All right, let’s say somebody was perfectly equipped the best phone and a great internet connection in that country. What would still be wrong? I was like, Oh, language is still wrong and the latency to their phone is actually still pretty high to our data centers, which is why it’s taking a long time to sign up. I could still fix that. So that’s how you can see the worst case to tell you everything, but then decide what is marginal by removing a few of the barriers that you know are difficult for you to attack and then see which ones are closer to being resolved.
Lenny: Awesome. I wanted to ask one more question about experiments at Facebook back in the day. So we talked about there’s all these lead bullets or some cannonballs, maybe a silver bullet somewhere. In your experience, what percentage of experiments end up being impactful and successful?
Adriel Frederick: Okay, that’s a difficult and different question. So I’d say probably 60% successful, 40% you should turn off. But within that 60%, I think there’s a hidden cost to the experiment, which is that you’re futzing around with something small. You could have used your time on something bigger and more meaningful, but you’re futzing around with a bunch of these small ones. Some of the small things were incredibly meaningful and you needed to do them. So I think this is actually, it’s almost like the same problem about, I don’t know which of my marketing is best. You have to try a bunch of stuff and then figure out what was terrible. You don’t know before you do it, before you do the experiment, what the impact is. But sometimes what I’ve seen is let’s take a bunch of them as a program and let’s say you have over the course of three months, you’re going to experiment with 10 things.
Adriel Frederick: You might have been able to push on two really big ones. And what I’ve seen is there’s a laziness, and this is broadly, this is not just at Facebook, it’s broader. There’s a laziness that can creep in where you’re just finding a lot of little things because they’re easier to come up with and they’re easier to design and think about. It’s easier to build, it’s easier to talk to your boss and say, remove the number by 0.02% and you feel good about doing those few small things. And so it creates this incremental thinking where you’re just trying to do a bunch of small things that just don’t meaningfully add up to something big.
Adriel Frederick: I think what’s healthy is having a good portfolio because basically you say like look, I’m going to have, using our analogies from before, I’m going to have some cannonballs. I’m going to work on a couple cannonballs and I’m going to have a bunch of lead bullets. And maybe it’s 80% of your energies on those big cannonballs, 20% on the lead bullets, and having a constraint like that forces you to choose the fewer experiments that are actually probably the really good ones and it’s not just a whole bunch of crap that you’re trying out.
Lenny: And is that actually how you divide up those bets broadly? Is that like a rule thumb you have or is that just numbers you’re putting out there?
Adriel Frederick: Those are just numbers I’m putting out there. It’s always going to be a gut call based on where you are. I think depending on the stage your product is at, it should be a different set, a different bite. Very early on when you’re building a product, you kind of know what the big things are. You’ve talked enough people, you have enough, just go build it. You should not be playing around with experiments. It might be a hundred percent cannonballs. Just go knock the big pieces out, don’t worry, it’ll work.
Adriel Frederick: Also, the cost of experimentation is time. So if you are experimenting on every little thing and waiting for the data to come in and then also screwing up some other part of the product because your experiment’s on 50 50, it’s just not worth it. Just bang the big things out. As you get more mature, the balance needs to switch in the portfolio. Probably not that many big cannonballs anymore, probably just one. And there’s probably a lot of the refinements that you need to work on. And by then you have this scale that the time to experiment isn’t as high and the cost of experimenting is lower. So it’s fine. It’s good to do with that way.
Lenny: Okay. One last question before we get to our very exciting lightning round. So you’ve moved from IC a while back at this point to now VP of Product at one of the most trafficked sites on the internet. And I’m curious, what skills have you grown or had to grow most as you’ve gotten more senior in your career?
Adriel Frederick: Organization design and empathy.
Lenny: Whoa, I love that.
Adriel Frederick: Oh dude, for a long time, and I think this is, I and many others had this idea that the people who are the smartest are the ones who rise. The people who are the most technically competent are the ones who rise. People who are the best individual contributors are the ones who rise. And somewhere along the way I had that idea disabused of me and I recognize the job’s different. It’s more about building a great team, creating the right incentives for the team on blocking them, guiding them, and helping them work efficiently. Those mattered way more than anything else. And I guess one of the ways I slowly recognize it is as I started going up in my career, I recognized that if I wanted to have more impact, I couldn’t do everything myself. There was just more that needed to be done. And in today’s world, you can’t do anything meaningful by yourself.
Adriel Frederick: You need a lot of people to do stuff with you. There’s nothing meaningful that gets done by any single person, even though people like to make you think that in their hustle porn that they post online. So it may be just a step back and think about what helped me be productive in environments when I was productive and how I could do that for others because then that would just naturally help me. And so there were simple things like clear goals, helping people feel safe and understand that you got their back. Making it easy to do their jobs. But my job is to make sure the processes for you doing your work and the people who you have to interact with are just buttery smooth and everything just runs easily. And so that was lesson one. It was just designing a good organization, culture, skills, people, processes, et cetera, all necessary.
Adriel Frederick: That’s one piece. The second is empathy. Where a step of that was just like you have to have that as a PM for your user, but I think it’s different to having it for a peer and another function or somebody else on one of your teams. And the hardest part of it is they say getting in somebody else’s shoes. The hardest part is taking my own shoes off. Basically going, Yo, okay, I came into this, there’s something I wanted, I wanted to get rid of that. Now just talk to this person and try to understand what’s going on with them, what they care about for their life goals and motivations, what they’re scared of, what they’re excited by, how you might be able to help them. Once I was able to get out my shoes, clear my mind, try to get into their head, then I could be like, All right, cool, let’s find a nice happy middle ground in the middle here. But that’s something that works for both of us together.
Adriel Frederick: And sometimes for me it was, Yo, what I care about, I’m good. I’m going to let you do your thing. I’ve gotten into your shoes, I need to leave you alone. Like you’re good. Other times I’m ready to push. But I think once I have the empathy, I’m then able to think about what we as an organization broadly want to achieve and try to put the two shoes on at the same time and find something that works for both of us. So what’s that common objective? And I think that’s how I try to approach almost every conversation is, especially being a guy who looks different, talks different, comes from somewhere else, first thought they might have is and unconsciously might be like this guy isn’t one of us. But then once I make it clear to them that we have the same objectives, we’re about the same thing and I want to know what’s going on with you so that I can help you achieve what you want to achieve. Dude, problems go away.
Lenny: Okay. I know I have to let you go and you have to get back to real work. So we’ve reached our very exciting lightning round, the final part of our little chat.
Adriel Frederick: Darn right.
Lenny: Basically I’m just going to ask you five quick questions, whatever comes to mind, share it and we’ll go through a pretty quick.
Adriel Frederick: All right.
Lenny: Sound good? Okay. What are two, three books that you recommend most to people?
Adriel Frederick: The Prize and probably now The New Map by Daniel Yergin. They are books about the history of oil and the geopolitics of oil. It is a fascinating way to understand the world. It’s like the best books I’ve seen to understand geopolitics and how they work and why they work. It does it through the lens of oil, which explains way more than you might think. And so this comes from the early part of my career working in energy.
Lenny: I will link that into show notes. I’ve not heard of that one before. What’s a favorite podcast of yours other than this one?
Adriel Frederick: Oh, of course, you took the easy one away. Revisionist History with Malcolm Gladwell, just gives you a different look into things. I’m also a huge car nerd, like deep into modifying and tweaking and tuning cars. So there’s this esoteric one called HP Academy that I’m into, but most of your listeners will not beat to that.
Lenny: Wow. Very out there and awesome. And I think there’s a new season of her Revisionist History coming out soon.
Adriel Frederick: Yep.
Lenny: Okay. Favorite recent movie or TV show?
Adriel Frederick: Last night I discovered Mo on Netflix.
Lenny: Mo.
Adriel Frederick: And it’s short for Mohammed. It is semi autobiographical about a Palestinian refugee living in Houston, his journey to seek asylum and live and work and date in this multicultural environment, he speaks Arabic, Spanish and English fluently. Funny as hell, but also dramatic. It is fabulous.
Lenny: Amazing. Okay. Wow, these are all very unique. I love it. In a different direction, what’s a favorite interview question that you like to ask?
Adriel Frederick: These days at work, I have to go through the standard interview questions, but when I got to play and sometimes when I feel like playing a little bit more, I’ll see something like, teach me something you don’t think I know. It’s a really good test of what you’ve heard me say a lot, empathy. I heard [inaudible 01:04:58] use it once and I kept trying it to see what it was good for and it helps you understand how good somebody is at reading you, how much knowledge they have and their ability to communicate and share knowledge. So it was like it actually could test a lot of things at once and a lot of times you learn something, it’s awesome.
Lenny: Okay, final question. Who else in the industry would you say that you most respect as a thought leader?
Adriel Frederick: Well look on the discipline, [inaudible 01:05:24], discipline of product management, definitely [inaudible 01:05:26]. I think just in terms of technology development, it’s the team behind Radiant Nuclear.
Lenny: What is that?
Adriel Frederick: While taking a break between jobs, I’m studying climate change and energy because of my background and I just basically became convinced that nuclear is a bigger answer than we’re giving a credit for. A lot of the barriers are political, not technical, but the solution they’re working on I think is a technical solution to some of the political problems we have around nuclear, which seems really interesting and I am really hoping that they pull off what they’re trying to do.
Lenny: Wow. I love how out there all these recommendations are. These are great. Adriel, I am so appreciative of you making time for this. I’m also really appreciative to Jules for connecting us. This was amazing. You’re awesome. Two last questions. Where can folks find you online if they want to reach out, learn more, and then how can listeners be useful to you?
Adriel Frederick: Awesome. Before I jump into that, thank you for having me on here. It’s just good to reflect about life and work for a little bit and hopefully share some insightful stuff with the folks who listen to your podcast. So thanks for having me. You can find me on LinkedIn, Adriel Fredrick, there might be one other. Pretty sure I’m the only one. And then how can listeners be useful to me? Number one, keep listening to this podcast because if everybody keeps listening to the insights that you are teasing out, a lot of things will work well and not necessarily me. Another thing they could do to be useful to me is find somebody that’s just different to you and talk to them for five minutes. That’s it. I think that will come back to me eventually.
Lenny: Love these and really flattered, really appreciate it. Thank you for being here.
Adriel Frederick: Thanks for having me, Lenny, take it easy.
Lenny: Thank you so much for listening. If you found this valuable, you can subscribe to the show on Apple Podcasts, Spotify, or your favorite podcast app. Also, please consider giving us a rating or leaving a review as that really helps other listeners find the podcast. You can find all past episodes or learn more about the show at lennyspodcast.com. See you in the next episode.
Glossary
| English | 中文 |
|---|---|
| AB5 | AB5(加利福尼亚州 Assembly Bill 5) |
| Adriel Frederick | 保留原文 |
| aha moment | 顿悟时刻 |
| biotech | 生物科技 |
| cannonball | 炮弹 |
| Chima | 保留原文 |
| Copilot | Copilot |
| cycles | 周期 |
| Daniel Yergin | 保留原文 |
| dead head | 空车返程 |
| Dolly | Dolly(AI 图像生成模型) |
| double E | 电气工程(double E) |
| flat head | 一字螺丝刀 |
| funnel | 漏斗 |
| growth hacking | 增长黑客 |
| imposter syndrome | 冒名顶替综合征 |
| issue tracker | 问题追踪工具 |
| Javi | 保留原文 |
| Jules Walter | 保留原文 |
| lead bullet | 铅弹 |
| Malcolm Gladwell | 保留原文 |
| marginal user | 边际用户 |
| medallion | medallion 车牌(纽约出租车运营牌照) |
| metaverse | 元宇宙 |
| Neuralink | Neuralink |
| onboarding | 新手引导 |
| orthogonal | 正交 |
| permanence | 持久性 |
| Phillips | 十字螺丝刀 |
| PM | 产品经理(PM) |
| Prop 22 | Prop 22(加利福尼亚州第 22 号提案) |
| Sam Harris | 保留原文 |
| silver bullet | 银弹 |
| Tom Allison | 保留原文 |
| torques | 梅花螺丝刀 |
| traction | 牵引力 |
| Zynga | Zynga(社交游戏公司) |
Reformatted by reformat_english.py
让产品开发更人性化 | Adriel Frederick(Reddit, Lyft, Facebook)
文字记录
Adriel Frederick: 大概有那么一群人,我称之为技术乌托邦主义者,他们会说,把所有数据喂给算法,给它一个目标,它就会做正确的事。我当时的反应是,嗯,这套说法之所以站不住脚,是因为算法通常不理解长期影响,也不理解人们可能会如何应对,更不理解你对产品的意图。我认为产品经理扮演这个角色非常重要。这就是我们的工作。当你在做重度依赖算法的产品时,你的工作就是理清算法应该负责什么、人应该负责什么,以及做决策的框架。
Lenny: 欢迎收听 Lenny’s Podcast。我是 Lenny,在这里我的目标是帮助你在打造和增长产品这门手艺上变得更好。今天的嘉宾是 Adriel Frederick。Adriel 是 Reddit 的产品副总裁,专注于在 Reddit 内部孵化和扩展新产品。在此之前,他在 Lyft 担任产品总监,在五年的时间里领导市场团队和定价团队;再之前,他是 Facebook 的早期 PM,花了四年时间领导用户获取团队。Adriel 是那些令人难以置信的产品领导者之一,但他的知名度远远不够——因为他没有整天泡在 Twitter 上,而是在踏踏实实地执行和打造出色的产品。这个播客的目标之一,就是 spotlight 那些你可能没听说过的杰出产品领导者。Adriel 就是一个很好的例子。
本期话题预告
Lenny: 在我们的对话中,我们聊到了 growth hacking(增长黑客)的起源、如何成为更优秀的产品领导者、在公司内提升多样性的方法、早期在 Facebook 增长团队工作的体验、AI 的未来以及更多内容。和 Adriel 聊天非常愉快,我非常激动能和大家分享这期节目。话不多说,请出 Adriel Frederick。
赞助商:Linear
Lenny: 本期节目由 Linear 赞助。说真的,你现在用的那个 issue tracker(问题追踪工具)并不怎么好用。为什么它似乎总是在和你作对,而不是为你服务?为什么用起来感觉像是一种负担?Linear 不一样。它速度极快,设计精美,还配备了强大的工作流,从问题追踪到管理产品路线图,精简你整个产品开发流程。Linear 专为现代软件团队的工作方式而设计。用户喜爱 Linear 的地方包括:强大的键盘快捷键、高效的 GitHub 集成、真正推动进展的 cycles(周期),以及让所有人保持同步的内置项目更新。简而言之,它就是好用。Linear 是初创公司的默认工具之选,同时也服务于 Versal、Retool 和 CashApp 等众多大型成熟企业。亲自体验一下为什么产品团队形容使用 Linear 就像魔法一般。访问 linear.app/lenny,与你的团队免费试用 Linear,升级时可享 75 折优惠。网址是 linear.app/lenny。
赞助商:Flatfile
Lenny: 嘿 Ashley,Flatfile 的市场负责人。你估计有多少 B2B SaaS 公司需要从客户那里导入 CSV 文件?
Ashley: 至少 40%。
Lenny: 其中有多少搞砸了?搞砸了会怎样?
Ashley: 根据我们的数据,大约三分之一的人在 onboarding(新手引导)期间只要有一次糟糕体验,就会考虑换到别的公司。所以如果你的 CSV 导入工具不好用——考虑到客户的文件里塞满了意想不到的数据和格式,这种情况非常常见——他们就会离开。
Lenny: 对此我一点都不惊讶。我一直看到,改善 onboarding 是提升注册转化率和长期留存率中杠杆最高的机会之一。让用户更快、更可靠地到达他们的 aha moment(顿悟时刻)极其重要。
Ashley: 完全同意。看到 Square、Spotify 和 Zoro 等客户能够在 Flatfile 之上发展他们的业务,真的很令人振奋。这是因为 flawless(完美无瑕的)数据 onboarding 就像一个催化剂,帮助他们和他们的客户更快地到达目的地。
Lenny: 如果你想了解更多或开始使用,请访问 flatfile.com/lenny 查看 Flatfile。
正式对话开始
Lenny: Adriel,欢迎来到播客。
Adriel Frederick: 很高兴来到这里,Lenny。谢谢你邀请我。
Lenny: 这是我的荣幸。实际上我是通过一个叫 Jules Walter 的人认识你的,我们都认识他。
Adriel Frederick: 是的。
Lenny: 他是 YouTube 的 PM。我其实问过他,这个播客应该请谁——那些可能不太为人所知但非常厉害的人——他立刻就推荐了你。所以我非常期待这次对话。
Adriel Frederick: 兄弟,Jules 能这么说真是太抬举我了。Jules 是我好兄弟。我爱他。他是个非常好的人,出色的产品经理,对这门手艺充满热忱,光是待在他身边我就觉得很舒服。
Lenny: 是的,我们以后也会请他上这个播客。他现在在忙某个不能说的秘密项目。他已经排上号了,我可以聊 Jules 聊一整天,不过他至今仍是我 newsletter 上第十最受欢迎的客座文章作者。怎么样?
Adriel Frederick: 哇。是的,他很棒。
Adriel 的职业历程
Lenny: 好了,不聊 Jules 了。为了让听众对你有所了解,能不能用 55 秒快速概述一下你职业生涯中所有精彩的经历?
Adriel Frederick: 哦,那我说快点。关于我的一个重要背景是,我来自特立尼达和多巴哥,加勒比海的一个岛国。来美国上大学,学的是 double E(电气工程),后来被咨询行业吸引,做了几年,在石油天然气、电力、重工业领域工作过,很喜欢这些东西,但周末也喜欢写代码玩。所以我觉得应该转向科技行业,然后就这么做了。在 Intuit 工作,帮助开发了他们的第一个 iPhone 应用,在那年月这可是件大事。在 Facebook 的增长团队工作了四年,做用户获取,非常有趣,也是一段非常扎实的塑造性经历。短暂涉足 biotech(生物科技),然后在 Lyft 做市场相关的工作——乘客定价、实时司机激励、乘客与司机的匹配,以及我们用来管理市场的大量运营工具。这就是我的旅程,大概 45 秒。
Lenny: 很好。我其实没有计时,但听起来差不多。接下来我们会聊到你在这些地方学到的很多东西。你还能分享一下你现在在做什么吗?
Reddit X 团队
Adriel Frederick: 好的,没问题。我是 Reddit X 的产品管理副总裁,听起来像是我们要把气球发射到太空,但其实不完全是那样。我们是 Reddit 内部的一个团队,思考的是如何演进人们与 Reddit 交互的模式。包括内容、时间性、你对话的受众。你想啊,Reddit 本质上是匿名陌生人之间围绕共同兴趣进行的异步对话。有时候人们也会在 Reddit 上找到他们问题的答案,但我们在 X 团队探索的是,把这些演进为比如帮助人们更快、更轻松地围绕共同兴趣进行交流,或许改变他们对话的对象。也许话题不再只是共同兴趣,也许他们有其他共同点将他们聚在一起。也许将视频、音频和其他媒体纳入产品的一部分,以及探索信息 permanence(持久性)之类的东西。
Lenny: 哇,我感觉你接下来要往 metaverse(元宇宙)的方向说了。这里面有元宇宙的角度吗?
Adriel Frederick: 不太有。我认为我们会把它看作一种潜在的技术,但我们的主要焦点更多放在那些目前已经大规模应用的交互模式和平台上。
Lenny: 明白了。近期有什么值得我们期待的东西吗?我猜你不太方便太多地谈论你们具体在做什么。
Adriel Frederick: 我觉得我们最近做了几件挺有意思的事。我们最近在做一个虚拟形象市场,创作者可以在上面创作艺术作品,放到 Reddit 上出售,供其他人购买和使用。它的表现非常出色。其底层技术是 NFT,我们认为这项技术非常重要,因为它为创作者提供了一种公开承认其对某项内容拥有权利的方式。这样他们就有了某种形式的知识产权保护,尤其是在一个做数字艺术交易的市场里,我们认为这一点极其重要。我认为 NFT 背后的技术确实被用于一些很不光彩的事情,但我觉得我们在正确使用这些技术方面仍处于起步阶段。有很多糟糕的使用和浪费性的使用,但我觉得其中也有一些真正的亮点,我们希望能发现其中的一些。
Lenny: 不错。我要避免被拉进 web3 的兔子洞,但这确实很酷。有一个我原本没打算问、但我很好奇的问题,因为我刚好在跟另一位嘉宾聊这个话题——就是大公司、成立已久的公司里那种偏研发性质的团队。我知道你加入 Reddit 时间不算长,这可能也是一个相对新的事物,但我很好奇,关于如何组建这样的团队、如何进行这类投资——这些着眼于长期的研究型团队,你有什么心得?
研发团队的组建心得
Adriel Frederick: 嗯,我觉得从”另一边”过来再做这件事,视角非常好。你想一下我之前的经历,我一直在增长和市场团队,这与新事物团队离得最远。而我看到经常发生的情况是——“排异反应”,这个东西跟身体其他部分、跟组织其他部门看起来太不一样了,以至于整个想法会被某种形式地完全排斥。所以我学到了几件事。
Adriel Frederick: 第一,公司其他部门需要把你们做的事情视为核心且对使命至关重要。不能让人觉得这帮人只是在一个角落里玩跟我们的日常工作无关的东西。因为我觉得这会导致一些怨气——你可以想象,内部任何一个团队都在争夺资源,他们看着这个小组拥有自己拿不到的资源,就会想:“哦,得把它干掉,因为他们没有在帮我们做我们来这里要做的事情。“所以你必须与核心使命紧密相关,否则在文化上就会出问题。我觉得这是第一点。
Adriel Frederick: 第二,它必须是所有人的成功。所以如果你们在这些研发团队中做成了什么事,不应该只是研发团队赢了,而是每个人都觉得自己赢了。这也跟我之前说的第一个目标有关。第三,你必须安排好这些团队的工作,让人们不会认为所有的创新都会发生在那个团队。不能让人觉得——好吧,我们就被卡在运营工作上了,他们才有所有的乐趣。其他团队仍然会创新,只不过我们承担的是其他团队没有精力去做、但组织需要的、且属于核心使命的一部分的工作。所以我觉得这就是我在组建和运营这些团队时思考的很多东西——确保我们是组织的一部分,每个人都想拥抱我们,认可”你们是我们中的一员”,而不是”你们需要到你们的小角落里去待着”。
研发团队原则回顾
Lenny: 太棒了,这真的很有帮助。所以简单回顾一下:你要让它感觉是公司的核心和关键所在;你要让它感觉是所有人的成功,而不是”哦,Adriel 在那边做得很好,但我们被卡在这些糟糕的难题上”;然后就是这个理念——不是所有的创新都只会从那个团队产生,我们所有人都可以创新,他们只是在专注于某一特定的创新方向。
Adriel Frederick: 对。
从特立尼达到硅谷
Lenny: 很好。好的,另一个我特别想问你的问题。你说你出生在特立尼达和多巴哥,这在科技界不太常听到。我很好奇你的背景和你走到今天的经历——这些如何影响你的领导方式、做产品的方式,以及你对职业发展更广泛的思考?
Adriel Frederick: 嗯,这不是我会有意识去想的事情,但它每天都在影响我,回过头来看很难不看到它的痕迹。我是 Facebook 的第一位黑人产品经理。
Lenny: 哇。
Adriel Frederick: 所以我很难不看到这对产品的构建方式或我自身产生了某种影响。这意义重大。但我觉得,了解这一点如何产生影响的一个方式,其实是先了解一下特立尼达。它算是它自己独特的小世界。特立尼达是加勒比海南部的一个岛屿,在最底部,紧挨着委内瑞拉。那是一个非常多元化的地方。从种族构成来看,35% 是印度裔,就是东印度裔,35% 是非洲裔,25% 是混血,剩下 5% 包罗万象——欧洲裔、华裔、阿拉伯裔等等。
Adriel Frederick: 从宗教来看,大约 60% 是基督教徒,但这 60% 里包含很多不同形式的基督教,20% 是印度教,7% 是伊斯兰教。媒体消费是英国和美国电视的混合。收入跨度非常大,但学校是一个大熔炉,所以不像西方大多数地方那样在学校中存在明显的阶级和收入隔离。当你有这样一个种族、宗教、媒体消费和社会经济地位的大熔炉,而且在学校里你和所有人混在一起,你会学到很多东西。我们有一个玩笑说,特立尼达大概是世界上公共假期最多的国家,因为你必须庆祝所有人的节日——从印度教的排灯节(Diwali)、开斋节(Eid al-Fitr)到圣诞节,我有朋友在斋月(Ramadan)期间封斋。我知道很多印度教神灵的名字,我总是喜欢用我对这些知识的了解让同事们大吃一惊。这赋予了我一种非常不同的视角,这种视角在工作中也会体现出来。
Adriel Frederick: 所以我给你举个例子。我注意到几乎所有我在科技界共事过的人,在做移动设备时,都会做一个假设:一个手机号加一台设备对应一个人。而在特立尼达长大,我早就知道这不是真的。用预付费手机的人号码可能一直在变,所以一个人可以有很多个号码,就是因为他们在用预付费。双卡手机也很常见。而一部手机——尤其在当时——是一个很昂贵的数字设备,是一台计算机。所以它经常是共享的,人们不可能每人独占一台。
Adriel Frederick: 所以当我在 Facebook 负责用户获取和设计注册流程时,这些认知被融入了产品设计之中,而我认为其他公司到现在还没有跟上。我确切地知道,当时围绕电话号码、设备以及如何将其与个体配对所做的大量思考,对 Facebook 当时的增长很有帮助,甚至在我离开之后,我知道这些仍然在发挥作用。这是一个简单的例子,说明仅仅身处那样的环境、吸收信息,就能对产品设计产生帮助——而我认为,如果我和像我这样的人不在那里,这一切是不会发生的。
Lenny: 你说你是 Facebook 第一位黑人 PM。我之前不知道这一点。你加入的时候那里有多少 PM?
Adriel Frederick: 天哪。我记得我们所有人都能塞进一间叫 Canada 的会议室,那大概是我第二周的事。当时大概有 30 个人左右吧。
Lenny: 哇。是的。关于那段经历,你有没有学到什么关于如何在公司推动多样性的经验?Facebook 这方面做得好吗?你有没有见过其他公司做得更好?有没有什么可以分享给正在推动这方面工作的人的?
Adriel Frederick: 这个问题挺难回答的。所以分两个部分来说吧。对我来说那段经历是什么样的?我觉得很快就从一点冒名顶替综合征变成了——就是那天我坐在那群人中间,心想,天哪,我是做 Facebook 的 30 个人之一。我在干什么?我真的属于这个团队吗?这太疯狂了。后来跟很多其他 PM 和工程师聊过之后,我意识到,不不不,他们是因为我的知识和我的视角才需要我的,因为他们真的想把这款产品推向全球。作为一个来自特立尼达的人,带着我刚才提到的那些视角来做增长,这种背景是被认可的。我觉得我很幸运能在增长团队,而那个团队的领导者们真的重视多样性。我回想起我待过的一些团队,真的非常棒。我有时候还会拿它们开玩笑。
Adriel Frederick: 我记得有一个团队,我是一个黑人特立尼达裔产品经理,工程经理是一位以色列女性,技术负责人是一位巴西女性,然后工程和设计团队的其他成员来自世界各地。有俄罗斯人、中国人,还有来自斯拉夫国家的人。这让产品设计变得很有趣,因为很多时候当你构建一个产品、想要走进用户头脑里去思考的时候,你得出去做用户访谈,因为你不一定能真正理解他们。天哪,我们在那个团队根本不需要。我们就互相争论。我们会想我们的朋友会怎么用它,我们的表亲会怎么用它,当我们争论如何设计一个产品时,我们其实已经覆盖了世界上很大一片区域。但我觉得增长团队最初的领导层,我觉得从 Chima 开始,然后是 Javi 接续——
Adriel Frederick: [听不清 00:16:51]——并且持续引入来自世界各地的、包括不同种族、宗教、文化的多样性,这样你才能真正以那种方式来打造产品。这让你变得非常高效,因为一个可能需要两周才能解决的争论——因为你得去招募一批用户来访谈、搞清楚怎么回事——我们互相抛来抛去,15 分钟就搞定了。我无法强调这对打造面向全球用户的产品有多重要。你的团队得像这个世界一样,这会让你快得多。这不是完美的。你仍然需要走出去和用户交流,因为我们自己也会形成各自的单一文化盲区,需要跳出来。但它确实帮了很大忙。
Adriel Frederick: 关于你的第二个问题,多样性和如何培养它,从我在麦肯锡的职业生涯开始,到今天在 Reddit,我经历过无数次这样的场合,每个人都在问同样的问题——怎么解决这个问题。以下是我看到行之有效的做法:当你认识到你能从中获得商业价值时,它突然就变成了你会主动去关注和维护的东西。就这样。当然这背后还有很多,但我认为,当它从——坦率地说——人们觉得为了政治正确、为了文化氛围或者迫于社会压力而不得不做的事,转变为你真正具体认识到”不,我能从中获得价值”,并且你愿意采取其他步骤来建立一种能够利用这种多样性的公司文化时,它就变得容易了。因为当你引入来自不同背景的人,他们会留下来,而正如你所知,这始终是增长的第一步。
Lenny: 留存。
Adriel Frederick: 你必须留住他们,你必须留住多元化的人才。所以你必须营造一个重视它、在乎它、善用它、奖励它的环境,因为它已经成为公司核心体系的一部分。一旦这一点运转起来,招聘就变得容易多了,因为人们看到你真的在重视它、引入它、渴望它。而不只是在说说而已。这就是我在所有关于这个话题的对话中所看到的真实情况。
Lenny: 你说的第一点很有意思,它其实回答了第二点——就是你提到的,早期拥有一个庞大、多元、全球化的员工群体,尤其是对一家想要走向全球和国际化的公司来说,是多么强大。你省下了所有这些时间。你不一定要去访谈那些你团队里已经有了其视角的人。
Adriel Frederick: 是的,不那样做的做法,总让我觉得有一种殖民主义的感觉。就好像一群人坐在这个国家的高塔里,在这个相对单一的环境中说,别担心,我们完全知道你们需要什么,而那些是世界各地的人们。这种方式就是行不通。所以对我来说也不太对劲。
Lenny: 是的。太棒了。谢谢你分享这些,真的很有帮助。还有一个话题我 definitely 想花点时间聊聊,就是我在看你的 LinkedIn 和你的背景时注意到的一个有趣的趋势。你先后在 Facebook、Lyft、Reddit 工作,而有趣的是,这些公司都是那种新闻不断、充满争议的地方。人们喜欢把它们推上风口浪尖,展示它们对世界做了多少坏事。我想象作为 PM,这种环境本身就很有挑战性,而你居然在三家这样的地方都待过。我想你一定学到了不少关于如何在充满混乱、不断救火、负面公关不断的公司里作为一名产品领导者开展工作的经验。你有什么可以分享的吗?
Adriel Frederick: 我觉得最重要的一点是,作为一名 PM,你就是一个领导者。你必须对团队起到缓冲或阻尼的作用,而这是双向的。有时候我们做的事情所有人都觉得棒极了,这是他们见过最好的东西,你就得把大家拉回现实,说,听着,那确实不错,但我们还有很多事要做。我们在为现实中的人们提供价值方面,离我们的目标还差得远,所以冷静一点,认清还有很多事情要做。然后当情况很糟糕、媒体告诉你你是世界上有史以来最糟糕的东西的时候,你也得回去说,大家,冷静一下。我们远没有他们想象的那么糟,他们看到了一些东西但不了解我们的全部,他们有时候会误解我们。所以在这样的时刻,把你的团队拉起来,继续朝着使命前进。
Adriel Frederick: 我觉得某些争议是必要的,所以我在这件事上可能站在一个不同的立场。我认为,如果你不对某种行为模式做出改变,你就不可能对世界产生任何有意义的影响。而如果你改变了一种行为模式,就一定有人在该模式中有既得利益,这就必然会产生冲突。最好读的新闻故事往往都涉及冲突,所以这种事总是会成为一个好故事,让你登上新闻。对 Facebook 来说,传统媒体和其他社交网络是一方,Facebook 是另一方,后来又变成了其他科技公司,这永远都是个好故事。对 Lyft 来说,对手是出租车和工会,所以你必须认识到,你总是会面临某种挑战。
分辨有效的批评
Adriel Frederick: 而应对这些问题真正困难的地方在于,搞清楚哪些批评是合理的,又有多少仅仅是因为权力来源正在被改变。我举个例子。以 Lyft 为例,纽约那些富有的 medallion 车牌持有者,当他们抱怨并试图禁止 Lyft 的时候,我对他们毫无同情心,因为当我在纽约街头伸手拦出租车的时候,他们看到我这个黑人就直接开过去了。所以不好意思,我确实对他们没什么共情。但我认为,关于司机薪酬结构的投诉是真正合理的,这些投诉也是一些重大新闻报道和一些重大法律行动背后的原因。当司机来接你的路上要花时间,或者当他们把人送到很远的地方然后不得不空车开回能接到单的地方时,他们的收入是没有保障的,这是真实的。
Adriel Frederick: 这是一个真实的问题,我觉得我们因此被批评了——我们之前对此关注不够,而这些批评推动了我们去解决它。我觉得——我说”我们”,虽然我已经不在那里了——这个问题还没有被完全解决。但我觉得作为 PM,听到这些的时候,你需要在批评背后找到真相,想办法去解决它,而不是迷失在对具体批评的回应中。所以要在”一部分争议只是游戏的一部分”和”不,这真的是合理的”之间走好这条线,要搞清楚边界在哪里,你得做一件听起来很老套但必须做的事——贴近你的用户。给你举个例子,我是怎么做的:当关于 Lyft 司机的投诉铺天盖地的时候,我自己去开车了,我就拿起车钥匙出门接单,自己亲身去感受,去看看这些司机到底在说什么。
Rick 的故事
Adriel Frederick: 我可以给你讲一个关于 Rick 的故事。我到现在还记得在 Berkeley 搭载 Rick 的那一单。我在家,直接上车,打开 app,开始接单。我接到一个 ping,15 分钟车程之外。我就在想——要现在去吗,这个乘客可能会取消订单,我这趟其实拿不到什么钱,但也许这单行程值得。于是我就开过去了,一路躲着车流、行人、喝醉的大学生、停车标志,终于到了 Rick 那里。他从 Chez Panisse 出来,大约 80 岁,跳上车,我按下按钮查看目的地,上面显示到目的地的预计时间是两分钟。我就问他:“嘿 Rick,你看清楚了吗?怎么回事?“他说:“嘿,我喝多了点,我怕摔断胯骨,所以才叫了车。“于是我一下子从想骂 Rick 让我开了 15 分钟来接他,变成了觉得——好吧,不不不,即使只开两分钟,我为你提供的价值也是实实在在的。但我意识到,这种价值并没有体现在薪酬结构中。Rick 会很乐意为我的 15 分钟来程付费,但我们一来没有给司机提供这部分补偿,二来也没有找到一种方式把这部分成本转嫁到 Rick 的价格中。这比那个简单的例子看起来要复杂得多。但这让我明白了司机们在抱怨什么。于是我想,明白了,我知道我们需要做什么了。所以当关于 AB5 和 Prop 22 的公关风暴正在进行的时候,我在外面开车,也和团队坐在一起讨论如何设计我们的产品来为司机补偿这部分损失。
Adriel Frederick: 同时还要保持对乘客来说合理的价格,不能造成不良激励,导致真正需要乘车的乘客无人接单——因为我不想让 Rick 摔断胯骨,他仍然需要一个让他觉得可以接受的价格来叫车。找到一种平衡这一切的方式,实际上比你想象的要复杂得多。这就是我始终聚焦的事情——无论 Prop 22 是否通过,我都为两种结果准备好了对乘客和司机都有效的方案。这就是我的工作。所以我觉得对 PM 来说,太容易被媒体的喧嚣吸进去了。你要做的就是——规划好工作,按计划执行,回到你的本职岗位上。这才是你应该做的。在混乱中为客户解决问题,然后再想办法把解决方案好好传达出去。
Lenny: 我喜欢这个策略的地方在于,它还能帮你看到,并不是所有人都在担心同一件事。我想到 Airbnb,比如所有房东都对某个功能很愤怒,说要造反了。然后按你说的,你去跟一些人聊聊——根本没人知道这件事,没人在乎,大家都觉得挺好的。
Adriel Frederick: 是的。
Lenny: 所以你说的这种做法——去跟客户聊,而不仅仅是关注那些最大的声音——好处真的太多了。
Adriel Frederick: 完全同意。我也对记者有同理心。一个标题写”一些 Airbnb 房东对某功能感到不满”——
Lenny: 对。
Adriel Frederick: 说实话,这不是一个好标题。我理解他们也有自己的工作要做,有时候他们在监督企业,有时候他们在用一些夸张的手法吸引人们阅读报道。所以他们要做他们的工作,我也要做我的。
Lenny: 是的。本期节目由 Eppo 赞助播出。Eppo 是一个由前 Airbnb 员工打造的下一代 A/B 测试平台,专为现代增长团队设计。Netlify、Tenfold 和 Cameo 等公司都在使用 Eppo 来驱动他们的实验。无论你在哪里工作,运行实验都变得越来越重要,但目前没有商业工具能够与现代增长团队的技术栈很好地集成。这导致了要么浪费时间自建内部工具,要么试图通过一个笨拙的营销工具来运行实验。我在 Airbnb 的时候,最喜欢我们的实验平台的一点就是能够轻松地按设备、按国家、按用户阶段来切分结果。Eppo 不仅做到了这些,还能做得更多——快速交付结果,避免冗长的分析周期,帮助你轻松定位任何发现的问题的根因。Eppo 让你超越基本的点击率指标,转而使用你的北极星指标,比如激活、留存、订阅和支付。Eppo 支持前端、后端、邮件营销甚至机器学习客户端的测试。请访问 geteppo.com,即 get E-P-P-O.com,让你的实验速度提升 10 倍。
Lyft 最紧张的时刻
Lenny: 你刚才分享了关于 Rick 的那个非常动人的故事,那你在 Lyft 工作期间最紧张的记忆是什么?
Adriel Frederick: 我觉得最紧张的时刻,是我不得不推翻自己做的一个糟糕的产品,然后重新做一个更好的版本。其实是一个定价算法的变更,是幕后的事情,没有人会真正看到,但这是一个相当大的项目。我们请来了收入管理方面的专家——博士级别的人,还有写过这方面教材的人来给我们做顾问。我们搭建了这个模型,上线了,然后期待着一个巨大的变化。结果呢,噗的一下,只起了一点点作用。然后我们反复打磨、反复打磨、反复打磨,终于在三个城市跑得很好了。我们开始向更多城市推广,而推广到更多城市简直是头疼得要命,因为系统超级复杂。
Adriel Frederick: 最终我们推广到了大约一百个城市,然后有人说,“好的,很酷,我想改价格。“结果我们花了好几个月去实现价格调整,天哪,那段时间大家对这个产品的情绪真的很难受。我记得在一个特别糟糕的一周之后,我出去散步,想弄清楚我到底该怎么办这件事。我们要坚持下去吗?过了一段时间,答案其实很简单,虽然情感上很难接受。答案就是:我们必须重建它。不存在我们不需要这种产品的可能。我们需要有影响价格的能力,这样才能真正有效地运营一个市场。但现有的方案不行。它在运营灵活性上达不到我们的需求,因为我们在构建的时候没有考虑到这个需求,我们被算法复杂性和那些精妙的技术迷住了。
Adriel Frederick: 所以我意识到我们需要承认现实,告诉大家我们没有做对,需要换一种实际上运营更灵活的方式来重新做。我们确实这么做了。我觉得最大的教训,至少在那个业务中,就是你必须把运营需求和运营控制作为一等需求来考虑。我认为当我们在其他很多消费互联网公司做产品的时候,你不需要考虑运营控制。你给算法一个目标,喂给它一些数据,让它跑,你观察它、确保它没有做出什么疯狂的事情,然后微调,但你不需要对产品进行日常的运营和战略控制。而我们就是需要让自己的大脑切换过来,能够在算法中加入人的参与。
Lenny: 对于那些没有在这种有地面运营团队的公司工作过的人,你能展开讲讲这是什么意思吗?运营控制,实际操作中到底指的是什么?
Adriel Frederick: 好,我给你举个例子。当时 Lyft 在全美国大概三百个城市运营,每一个城市的定价都不完全一样,都有一点不同。有时候你可能需要做季节性调整,因为交通变差了,或者燃油价格不同了,或者有了新的税费,或者竞争对手做了什么你需要回应的事情,而你的算法看不到这些。它对这些完全没有可见性。所以你需要一个人参与进来,不仅要提供这些可见性,还要做出如何应对的决定。因为我觉得,比如说你在芝加哥,遇到暴风雪,你需要改变定价方式,比如你需要调整价格来提高司机收入,从而让人们在暴风雪期间愿意出车。你并不完全清楚你想怎么应对,每场暴风雪都不一样,必须有人做出那个判断,并提供正确的信息给产品来利用它。当然算法已经处理了很大一部分,它们能做出一般性的反应,但要精确得多,就需要人来帮忙处理,来做那个判断。
Lenny: 明白了。很酷,谢谢分享。你刚才提到了一个观点,就是当你在一家有大型运营部门的公司工作时,显然还有一个核心的中央产品团队,你分享了一些在这种环境下工作的经验。所以我只是想回到这个话题。
Adriel Frederick: 当然,当然。
Lenny: 你说的主要观点就是,在设计软件的时候要把运营作为一等要素来考虑。这是最大的教训吗?
Adriel Frederick: 我觉得不仅仅是把运营当作一等需求来对待。对我来说更大的启发是,当我回顾整个职业生涯时,算法需要人来帮助做出判断。我在 Lyft 上了很深的一课,但当我回头看的时候,我意识到在 Facebook 也是一样的,只是不在我的职责范围内。在广告出现的频率和我们展示来自朋友和家人的有机内容的频率之间,总是需要做一个判断:我们展示你感兴趣但不在那个圈子里的内容的频率是多少?我们展示帮助你找到朋友或帮助其他人找到朋友的内容的频率是多少?这是一个判断性的决策,在不同的市场和不同的情况下是不同的。也许幕后有算法在实时为每一个人做这个决策,但仍然需要有人对这些施加一些战略性的判断。
Adriel Frederick: 我在 Facebook 的时候没有处在需要做这件事的位置上,但当我在 Lyft 意识到我需要做多少这样的事之后,我回头看了历史,发现 Facebook 那边也是一样的。但我认为有太多人没有看到这一点,相信一切都有算法解决方案。我认为作为产品经理,尤其是从事大量机器学习或运筹优化方面工作的产品经理,要去思考哪里需要人来做决策,哪里可以让机器自主运转,并且要把这当作一个产品设计问题来对待,因为那里实际上存在一个计算机界面需要你去考虑。你需要关于当前情况的信息,比如在 Lyft,我的市场情况如何?有人多久能被接上?我比竞争对手贵多少?我在这个市场的目标是什么?今天的表现如何?你要给人提供信息,同时也要给他们工具来执行正确的决策而不制造麻烦。这就是一个产品设计问题,是一个与其他任何东西一样的一等产品设计问题,你必须去思考。我并不了解内情,但我可以保证其他公司也有人在思考同样的问题。
Lenny: 这让我想起,我刚才在听 Zuck 上 Joe Rogan 的播客,他提到了一个观点:当你看一条帖子的时候,你可以加一个小小的 emoji 反应,其中有一个生气的小 emoji。他做出了一个决定——我们在算法中不会以任何方式使用生气 emoji 反应,我们就是忽略它。因为自然而然地你会想,好吧,人们生气了,这很有意思,让我们展示它,因为人们对它感兴趣。但他明确想避免愤怒、避免助长愤怒,可能是因为他们收到了大量相关的反馈。
Adriel Frederick: 没错。我觉得他们可能是——我会称之为技术乌托邦主义者——那些说把所有数据都喂给算法、给它一个目标,它就会做出正确事情的人。我就想,好吧,这个想法之所以站不住脚,是因为算法通常不理解长期效应,也不理解人们会如何对它做出反应,更不理解你对产品的意图。我认为产品经理(PM)扮演这个角色非常重要。这就是我们的工作。当你从事算法密集型产品时,你的工作就是弄清楚算法应该负责什么、人应该负责什么,以及做决策的框架是什么。
Lenny: 你能不能想一个具体的例子,比如你做到了这一点,或者没做好,或者你团队里有人本应该做到的?随便一个例子,让这件事更具体一点就好。
Adriel Frederick: 假设你是一个做定价的人,你说,好,我的目标是在某个区域赢得市场份额。然后你把这件事交给一个算法,说我需要你优化价格,使市场份额最大化。那算法会怎么做?把价格降到地板上。一路降到地板上,然后你一分钱都不赚。好,那你说,好,下一步怎么办?给它加一个约束。设定一个我们愿意接受的利润下限目标。好,现在去做吧。
Adriel Frederick: 如果对面那个人也在做一模一样的事呢?你们俩都会触及各自的约束,然后博弈就停了。好,那接下来就变成,哦,我们必须选择我们想在哪里赢。我认为我们做的一件让我特别自豪的事情,就是构建帮助人们看清并理解这场博弈的产品,让他们决定自己想在哪里取胜。我觉得前几个部分并不令人意外,但最后那个结论——哦等等,我需要创建一个工具,给人们提供信息,让他们决定如何玩这场博弈——才是真正关键的地方。
Lenny: 有意思。所以我听到的是,大量的工作是给人类提供更多信息,而不是给机器学习算法提供更多信息,而且给人类更多调整和旋拧的空间可能杠杆更大。
Adriel Frederick: 让我再精确一下这个说法。更重要的是给人们提供他们可以用于那些只有他们才擅长做的决策的信息,同时给机器放大人的意图的能力。所以我喜欢的一种思考方式是,所有任何形式的软件,包括 ML,都只是一种工具,就像螺丝刀。你可以尝试用一字螺丝刀去拧十字螺丝,也许能凑合一点,但最好还是用十字螺丝刀。我们本质上就是工具设计师,特别是在产品开发职能中,你要弄清楚多少东西放进工具里,多少东西留给人来判断,我给人选择他们想怎么做的能力。我给他们一把一字螺丝刀、一把十字螺丝刀、一把梅花螺丝刀,然后让他们自己决定面对手头的应用场景想怎么使用这些工具。
Adriel Frederick: 所以从这个类比落到具体,在 ML 中你会说,看,机器学习在针对给定目标进行优化方面会非常出色,但它不会理解我需要做的约束条件或战略选择。我们在外部世界中需要的约束条件和战略选择,永远都必须由人来决定。你要让这件事对人来说变得极其简单和直观,然后你让那个算法去放大他们的效果——通过在成千上万、甚至上百万个单独的决策上做出选择,把那个人的意图拿过来,结合在那个单一情境中能学到的所有信息加以放大。所以我把它看作是设计一个界面,让它成为你自身的延伸,而不是一个需要你给它更多信息的黑箱。这样有帮助吗?
Lenny: 是的,这让我想到 Neuralink 和 Elon 正在尝试做的事情。我不知道他是不是这样想的,但有人把它描述为——Elon 担心 AI 在某个时点会接管一切,所以他想构建一个直接连接我们大脑的工具,让人能够访问 AI 的能力,以便在未来面对一台失控的计算机时还有一线胜算。
Adriel Frederick: 即使那样,你也必须确保人仍然在掌控之中。我听到那个想法后会想,好,你建了界面,但谁在控制?是人仍在控制,还是人变成了机器的奴隶,而你只是做了一个更好的界面让他们更好地成为奴隶?
Lenny: 靠,我们完蛋了。
Adriel Frederick: 我还没有那么担心它们接管世界的那些愿景。到目前为止——也许我还没想象到它们能做到什么程度——它们仍然看起来像是需要我们的引导才能发挥作用的工具。即使是最令人惊叹的——我们一直在看那些图像生成,我也见过最前沿的文本生成技术。它们可以骗你相信它们接近人类的能力,但我从中看到的是缺乏决策和判断力。我仍然把它们看作是延伸性的、有用的——比如文本生成算法。很多算法没法帮你写论文,而我认为这正是人们害怕的——因为它仍然需要你的判断来做决定。当你决定了你读过的内容中哪些是重要主题时,假设你在写读书报告,你已经决定了主题是什么,它确实可以帮你更快地写出论文,但它没法替你写论文。它无法选择那些基于你的背景、经历和兴趣觉得有用或值得深入探讨的主题。
Lenny: 这不是我预期我们对话会走向的地方,但我想在这里再加一个想法,因为很有意思。我的思考方式是,我们的大脑没有什么神奇的,所以如果这是真的,为什么不存在一个我们可以完全模拟它的世界?Sam Harris 经常谈论这一点——感觉一旦接近了,它就可能以极快的速度超越人类的潜力,从人类的 20% 水平开始,到 40、50、60,然后变成人类的一百万倍。它可以非常快地超越我们。所以我认为很多——并不是说我害怕这个——但我感觉很多恐惧来源于此。比如 Dolly 出来了,然后 Copilot——就是那种”卧槽”的感觉。
Adriel Frederick: 是的,我们的大脑擅长线性思维,不擅长指数级思维。所以我听过那个论点,说这个增长是指数级的,你无法想象。我的反应是,是的,这确实有可能是真的。我完全能看到这种可能性,也承认我在理解这件事上存在认知缺陷,而且即使它是一个低概率事件,我们也应该关注它。所以我完全赞成关注这件事,因为——就假设低概率结果的代价仍然是很高的,所以它仍然值得关注。
在 Facebook 学到的增长经验
Lenny: 是的。好的,不错的小插曲。我想聊聊你在 Facebook 学到的东西。我们一直在聊其他各种地方,尤其是关于增长、关于增长和增长黑客你学到的东西,然后我在想 Facebook 所处的那个有趣的世界——或者说 Meta——一方面,当他们刚开始的时候,我在说增长黑客,Facebook 做了很多增长黑客的事情,给哈佛所有人发邮件,搞了各种有趣的约会功能,引发了很多争议,用了各种有趣的策略来启动 Facebook,但现在人们反过来利用 Facebook 来做增长黑客、做增长,比如 Zynga 就很有名,还有其他一些公司。所以说,我很好奇,你在 Facebook 和其他地方的经历中,关于增长或者说增长黑客学到了什么?
Adriel Frederick: 我认为传统意义上的增长黑客——找到那些对产品做的小改动来获得超乎寻常的影响——这绝对是有价值的。但我看到人们迷失的地方在于,他们以为只靠这些就够了。你以为只要找到那几个 hack,拼凑在一起,就能增长出一个成功的产品。我觉得这有一种对产品使用者的不尊重。就好像你假设他们没有智商,最终不会看穿你在做什么。老话说,骗得了一次,骗不了两次,大概就是这个道理。所以如果你没有一个真正为人们提供根本价值的产品,你可能昙花一现,靠增长黑客搞出一个也许能维持几个月的东西,但人们会看穿的,然后它就会消失。
Adriel Frederick: 所以我觉得那些手段是有帮助的,尤其是在早期获取初始牵引力的时候。但你必须有一个人们想要继续使用的东西。回想我们做过的那些真正产生了巨大效果的产品改动,它们都是在关注边际用户,想办法让产品更容易上手。人们很容易被一种想法诱惑——认为存在某种快速的秘密方法。而我的答案是,不是的。绝大多数情况就是苦干,找到办法解决真正的问题。那些真正的问题是什么?其实相当简单,但我们就是在上面死磕了很久,一直盯着不放。第一,让产品容易被找到。第二,让产品容易进入。第三,找到朋友要简单到不能再简单。一旦你做到了这些,后面就势如破竹了,这些就是我们一遍又一遍在做的事情。
Adriel Frederick: 我觉得另一个重要的部分是提醒人们这里有一些有趣的东西,培养他们不断回访产品的习惯,这也是其中的一部分,但我们就是在这几件事情上一遍又一遍地死磕。而一些真正大的胜利并不是 hack,它们只是对细节的关注。我举个例子。我记得有一天坐在那里想,怎么帮助人们找到他们的头几个朋友,我们会做一种推荐——如果你能找到一个或两个朋友,后面就势如破竹了,我们就能给你找到更多在同一个朋友圈里的人。所以我思考了一下”你可能认识的人”算法的工作方式——你得到一两个朋友,它会找到你的共同好友,然后帮你找到更多类似的人。
Adriel Frederick: 然后我想,你知道这会造成什么吗?它会把你旋进一个朋友圈里,但没法帮你找到所有其他朋友。我记得就是看着一个人在用这个产品,意识到我们只带他们走了一条路径。所以我就想,天哪,我怎么才能看到你所有的朋友圈?于是我们想出了一个办法来实现这个。我不打算把具体细节说出来。但那是一个彻底的改变,绝对的彻底改变。尤其是对用户来说,帮助他们找到不同朋友圈中的头几个朋友,这就意味着我们可以带你深入一个朋友圈,再深入另一个,就这样通过推荐不断构建你的人脉关系图。因为我们有一个非常好的工具来启动它,而这并不容易。那不是一个 hack。那是苦干。
不是 hack,而是苦干
Adriel Frederick: 我还记得我最喜欢的一个例子是 Tom Allison 做的事情。Tom Allison 现在应该负责 Facebook App 了,当时他是一个团队的工程经理,我们要对其中一个算法做一个改动,这个改动非常难搞,不是什么 hack,需要花好几个月才能完成,Tom 就带人躲到了一个角落里。他让所有人都知道我们真的要改变这个产品的工作方式。他安排了一个非常聪明的人在负责这件事,[听不清 00:47:03],他们就躲在一个角落里,按照需要的方式重新构建了这个产品,让它更容易运维和扩展,然后把它推上线。当然,他们大获成功,而且对此非常谦虚,但这不是一个 hack。它来自于他们深入观察这个深层问题,找到那个真正重要的东西,然后说,我们需要做一个根本性的改变,让人们更容易获得好友推荐,然后就在上面死磕。所以我给人们的建议之一是,当你在思考产品增长的时候,想清楚核心行为是什么,然后在上面死磕。想办法去除摩擦,去除一些障碍。就是持续不断地去做,当你在这个过程中,你会做一些小的 hack。你需要搞清楚按钮上放什么文字,把它放在首屏上方,写出合适的文案。所有我们传统上与增长营销关联的事情,你都得做。但对我来说,这只是基本功课,就是做好与用户的产品沟通,然后你还得去想那些还没搞明白你的产品的人,他们正在努力完成某个操作,你要让他们简单到不能再简单。我还有无数类似的例子,但核心就是这个。不是去找什么花招来 [听不清 00:48:18] 一个网站。
Lenny: 我喜欢这个说法。我听过的很好的总结就是——没有银弹,只有很多铅弹。
Adriel Frederick: 是的。偶尔还会有几发炮弹。时不时地会有一些炮弹。
Lenny: 你说的炮弹,能举个例子吗?
增长中的”炮弹”
Adriel Frederick: 用手机号注册,现在这已经是标准操作了,但在当时那就是一发炮弹。把短信推送到全世界的人手上。听起来没什么光鲜的,实际上非常难做。那是一发炮弹。优质的好友推荐,另一个大的。还有更多,我不打算全部展开。我所说的炮弹是指,有时候你需要对产品做一些真正大的、根本性的改变,才能让这些事情跑通。
Lenny: 明白了。所以你是从投入的角度来衡量这个的,不一定是看效果——
Adriel Frederick: 投入。
Lenny: 效果加上巨大的投入。好的。这方面我还有好多问题。好,我挑几个。一个是 Facebook 很有名的激活里程碑——获得 10 个朋友或 7 个朋友,不管具体是多少,就是你得达到某个朋友数量,然后好事就会发生。你参与了那个吗?你对它是怎么来的有什么了解吗?那个是真的吗?
Adriel Frederick: 那个决策在我之前就定了。我亲眼看到了它,理解了其中的数据,也在这个问题上做过功课。我认为其中真正精妙的不是那个指标,而是把它设计得易于理解和传达。我觉得它了不起的地方在于,你现在还在谈论它——因为它令人难忘,而且它让人们采取了正确的行动去追逐那个目标。那个数字或时间节点其实完全没有任何神奇之处,本质上它只是在说:尽快让用户获得尽可能多的好友。但如果你笼统地跟人说这句话,他们的反应会是:嗯,我大概明白了,好吧,我去做吧。
Adriel Frederick: 当你设定一个明确的数字和一个明确的时间节点,有了一个具体的追逐目标,有一个数字和一张所有人都能看到的图表——我们要让那个数字上去——这种组织效应是极具凝聚力的。所以我认为其中精妙之处在于——当然这些都是我听来的二手说法——当时关于数字应该是多少、时间范围应该是多长有大量争论,然后 Zuck 在某个时刻直接说了:10 个好友,14 天,开干。这就让大家越过了学术式的争论:好,明白了。尽可能多的好友,尽可能快,出发。
关于”10 个好友”的副作用
Lenny: 太喜欢了。这跟我的想法完全一致——重要的不是那个数字本身,它就是一个所有人都能集结起来一起冲锋的口号,不需要是某个跟留存率高度相关的完美数字。就像,够好了,方向对了,我们去干就行了。出发。
Lenny: 当然它也有缺点。有些还挺搞笑的。我记得看过一张留存率与好友数量的关系图,在 11 或 12 个好友的时候留存率实际上比 10 个时下降了,因为代码里某个人写了以 10 个好友为上限的东西来帮助提升留存率。到了 11 或 12 就断了,然后才回升。但我当时就想,没关系,完全没关系。因为如果我们没有获得那种组织动量,整条曲线都会更低。所以那个下降的折点我可以接受,没事。
什么是边际用户
Lenny: 你还提到了边际用户这个概念,我觉得把它展开讲讲会很有帮助。
Adriel Frederick: 对我来说,边际用户就是那个正处于你想要他采取的行动的临界点上的人。我给你举个具体的例子。在做注册相关工作时,我会去找一个我们有大量增长、但转化率却因为某种原因很糟糕的国家。也就是说,有大量流量,但转化率很差。我就想,好吧,这就是边际用户。这个人正处于要进来的临界点上——从流量可以看出他想进来——但我们没法让他进来。为什么?当你走向极端,找到那个处境最差的人,最可能的情况是:一个人在用功能机,连着 EDGE 网络,在一个离我们数据中心很远的国家尝试访问 Facebook。然后你就会想:好吧,这个人的体验有什么问题?
Adriel Frederick: 让我们去看看。然后你就会说,哦,你能看到产品所有的问题。于是它给了你一个清单。好,语言可能不对,我们没做对。我们有没有正确检测到所在国家,从而能正确地格式化他们的手机号?大概没有。天哪,离数据中心很远,所以连接很慢,而且还是 EDGE 网络。太糟糕了。你把所有这些问题打包起来,就得到了所有出问题的地方,然后大家就可以开始想办法怎么处理它们。不过我要提醒大家一点:不要只用数据来判断谁是边际用户。数据会给你线索,告诉你他们大概在哪里、可能出了什么问题,会给你一些提示。但它不会给你答案。你必须去观察他们才能找到答案。
不要只依赖数据
Adriel Frederick: 因为我觉得在很多数据驱动的地方,有人会说:太好了,建一个漏斗吧。搞清楚漏斗每个步骤的所有流失点,自己看看,然后判断哪里可能有问题,去修复它们。但我觉得实际情况往往是,有一个跟那个漏斗完全正交的问题,你看数据是看不出来的。你得去观察用户,跟他们聊。我记得有一个例子,我们在印度看一个人第一次注册 Facebook,他们正准备输入名字,我们就问他:你打算填什么名字?他说:我的法定全名。好,酷。现实世界中有人这么叫你吗?没有。我当时就想,天哪,我们完蛋了。
Adriel Frederick: 如果你发出好友请求,不会有人接受的,因为没人认识这个人。反过来也一样,如果别人找到你,也不知道你是谁。所以就像——你会去查漏斗更深处的问题,比如我的接受率怎么了?然后你会去拆解那个小漏斗,最后才发现问题的根源其实在很早之前就发生了。所以当你思考边际用户的时候,你必须走出去观察他们,跟他们聊,看他们怎么用产品,尽可能设身处地去理解他们的处境,然后再做出判断。数据不会告诉你答案,它只会告诉你情况有多糟。
Lenny: 哇。我喜欢这个——它跟你之前在所有其他场景中给出的建议是一样的:去跟人聊,不要只依赖聚合数据。
Adriel Frederick: 别误会,我在 Lyft 搭建了实验平台。我是个热爱数据的人,喜欢用数据、看实验。我只是觉得,坐在笔记本电脑前,拉出一个漏斗或一些图表,看看实验结果,就以为能从中找到该做什么的线索——这太容易了。数据是补充,不是全部。我看到很多人掉进这个陷阱,尤其是在那些拥有大量数据的公司里工作的时候,你会陷入一种幻觉,觉得自己游弋在答案之中,因为你有这么多数据,只需要把它们挖掘出来就行了。走出去跟人聊聊,你会更快找到答案。
Lenny: 我很喜欢这个建议——当你试图优化东西的时候,聚焦你的边际用户。你提到的这个思路包含两个部分:一个是下一个最有可能注册的人,另一个是最差的情况——走向他们,看到所有出问题的地方,把这当作你的北极星,让这个人成功,就能让更多更多的人成功。你是这么想的吗?
Adriel Frederick: 是的,我是这么想的。所以我觉得”边际用户”是一个很有意思的词,因为你会联想到家庭的比喻,想到孩子,想到那个正处于临界点上的人。但我喜欢去看最差的情况。它会向我展示所有出错的地方,而边际用户的思维方式则帮你优先决定下一步该做什么。所以我之前举例的那个边际用户,他们用的是未来的手机,用的是 Edge 浏览器,很多东西都有问题,而且很难解决。但我可能会看那个体验,然后说,好吧,假设一个人装备完美——最好的手机,在那个国家有很好的网络连接——什么还会出问题?我就发现,哦,语言还是不对,而且他们手机到我们数据中心的延迟实际上还是挺高的,所以注册流程很慢。这些问题我仍然可以修复。所以这就是你可以通过最差的情况看到所有问题,然后通过去掉一些你知道自己很难攻克的障碍来判断什么是边际问题,再看哪些问题更接近被解决。
Lenny: 太好了。我还想再问一个关于当年 Facebook 做实验的问题。我们之前谈到有各种各样的铅弹,或许还有一些炮弹,也许还有银弹。根据你的经验,有多大比例的实验最终是有影响力、成功的?
Adriel Frederick: 好吧,这是一个很难回答也与众不同的问题。我会说大约 60% 是成功的,40% 你应该关掉。但在这 60% 里,我认为实验有一个隐性成本,那就是你在一些小事情上折腾。你本可以把时间花在更大、更有意义的事情上,但你却在一堆小事上折腾。有些小事确实极其重要,你必须去做。所以我觉得这实际上,几乎就像是”我不知道我的营销里哪个最好”这个同样的问题。你必须尝试一堆东西,然后搞清楚哪些是糟糕的。在做之前,在做实验之前,你并不知道影响是什么。但有时候我看到的情况是,把一堆实验作为一个项目来推进,比如你在三个月的时间里要实验 10 件事。
Adriel Frederick: 你也许本可以推进两个真正大的项目。而我观察到的是一种惰性,这是一种很普遍的现象,不仅仅是在 Facebook,更广泛地存在。一种惰性会悄悄蔓延进来,你只是不断找到很多小事情,因为它们更容易想出来,更容易设计和思考。更容易构建,更容易跟老板汇报说某个数字降低了 0.02%,做了这些小事情你会感觉良好。所以它催生了这种渐进式思维,你只是试图做一堆小事情,但它们加起来并不能有意义地汇聚成什么大的成果。
Adriel Frederick: 我认为健康的做法是拥有一个好的组合,因为基本上你会说,听着,用我们之前的比喻,我要有一些炮弹。我会推进几个炮弹,同时也要有一堆铅弹。也许 80% 的精力放在那些大炮弹上,20% 放在铅弹上,有这样一个约束条件,会逼你选择更少的实验,而这些实验往往才是真正好的,而不是一堆你在胡乱尝试的垃圾。
Lenny: 那你实际上是这样分配这些赌注的吗?这是你的一个经验法则,还是只是随口说的一个数字?
Adriel Frederick: 那些只是我随口说的数字。这始终要靠直觉判断,取决于你所处的阶段。我认为根据产品所处的阶段,比例应该是不同的,配比不同。在非常早期构建产品的时候,你基本知道那些大的事情是什么。你已经跟足够多的人聊过了,你有足够的信息了,直接去做就好。你不应该在实验上东折腾西折腾。可能 100% 都是炮弹。直接把大块头搞定,别担心,行得通的。
Adriel Frederick: 而且,实验的成本是时间。所以如果你对每一个小事都做实验,等数据回来,同时你的实验 50/50 分流还把产品的其他部分搞乱了——这不值得。直接把大事搞定就好。随着产品越来越成熟,组合中的平衡需要调整。大概没有那么多大炮弹了,可能只有一个。而你需要做的更多是打磨和优化。到那个时候你已经有了足够的规模,实验所需的时间没那么长了,实验的成本也降低了。所以这样做没问题,按这个方式来做是好的。
Lenny: 好的。最后一个问题,然后我们进入非常精彩的闪电问答环节。你从 IC 转到现在已经有一段时间了,现在是互联网上流量最大的网站之一的产品副总裁。我很好奇,随着你在职业生涯中越来越资深,你成长最多或不得不成长最多的技能是什么?
Adriel Frederick: 组织设计和共情能力。
Lenny: 哇,我喜欢这个。
Adriel Frederick: 天哪,很长一段时间里,我认为这也是我和很多人曾经有的一个观念——最聪明的人才会晋升,技术能力最强的人才会晋升,最好的个人贡献者才会晋升。在某个阶段,我这个想法被纠正了,我认识到这份工作是不一样的。它更多是关于打造一支优秀的团队,为团队创造正确的激励机制,为他们扫除障碍,引导他们,帮助他们高效地工作。这些比什么都重要。我慢慢认识到这一点的方式之一是,随着我职业生涯往上走,我意识到如果我想有更大的影响力,我不可能什么都自己做。需要做的事情太多了。在当今世界,你不可能独自做出任何有意义的事。
Adriel Frederick: 你需要很多人和你一起做事。没有任何有意义的事情是由任何一个人独自完成的,尽管人们喜欢在他们网上发布的奋斗叙事里让你以为可以。所以也许就是退一步想想,在我曾经高效的环境中,什么帮助我保持了生产力,以及我怎样把同样的事情为别人做到,因为那样自然也会帮助到我。所以有一些简单的事情,比如清晰的目标,帮助人们感到安全,让他们知道你为他们撑腰。让他们更容易做好自己的工作。而我的工作是确保你做工作的流程、以及你需要打交道的人,都像黄油一样顺滑,一切运转顺畅。所以这就是第一课——设计一个好的组织:文化、技能、人员、流程等等,都是必要的。
Adriel Frederick: 这是第一点。第二点是共情能力。其中一个层面是,作为产品经理(PM),你必须对用户有共情,但我认为对同事、对另一个职能的人、或者对你团队中的其他人有共情是不同的。而最困难的部分是,人们说要站在别人的角度想问题,但最难的部分是脱下自己的鞋子。 basically 就是说,好吧,我带着某种预期进来,我有自己想要的东西——先把那些放下。然后就跟这个人交谈,试着理解他们的情况,他们在乎什么,他们的人生目标和动机是什么,他们害怕什么,什么让他们兴奋,你可能怎样帮助他们。一旦我能脱下自己的鞋,清空自己的思绪,试着进入他们的头脑,那我就能说,好的,让我们在这里找到一个双方都能接受的中间地带。而且是那种对我们双方都有效的方案。
Adriel Frederick: 有时候对我来说就是这样,嘿,我在意的那个点,我没问题。我会让你去做你的事。我已经换到了你的鞋子里,我就不该再打扰你了。你没问题。另一些时候,我准备好推进了。但我认为一旦我有了共情,我就能思考我们作为一个组织整体上想实现什么,试着同时穿上两只鞋,找到一个对我们双方都有效的方案。那个共同目标是什么?我觉得这就是我试图对待几乎所有对话的方式,特别是作为一个看起来不一样、说话不一样、来自别处的人,他们脑海里可能第一个念头,也许是无意识地,会觉得这家伙跟我们不是一类人。但一旦我向他们表明我们有相同的目标,我们在乎同样的事情,而且我想了解你的情况,这样我才能帮你实现你想要实现的目标。兄弟,问题就消失了。
闪电问答
Lenny: 好的。我知道我得放你走了,你得回去干正事了。所以我们来到了非常令人兴奋的闪电问答环节,我们这段小聊天的最后一部分。
Adriel Frederick: 没错。
Lenny: 基本上我就是会快速问你五个问题,想到什么说什么,然后我们很快就过一遍。
Adriel Frederick: 好。
Lenny: 行吧?好。你最常推荐给别人的两三本书是什么?
Adriel Frederick: 《The Prize》,现在可能还有 Daniel Yergin 的《The New Map》。它们是关于石油历史和石油地缘政治的书。这是理解世界的一种极其迷人的方式。这是我看过的理解地缘政治及其运作方式和原因最好的书。它通过石油这个视角来讲述,解释的东西远比你想象的多。这来自我职业生涯早期在能源领域工作的经历。
Lenny: 我会把它放到节目备注里。这本我之前没听说过。除了这个节目之外,你最喜欢的播客是什么?
Adriel Frederick: 哦,当然了,你把最简单的答案拿走了。Malcolm Gladwell 的 Revisionist History,能让你从不同角度看问题。我也是个超级车迷,深入到改装、微调和调校汽车的那种。所以有一个比较冷门的叫 HP Academy,我很喜欢,但你的大多数听众应该不会感兴趣。
Lenny: 哇,非常独特,太棒了。我记得 Revisionist History 好像很快会出新一季。
Adriel Frederick: 对。
Lenny: 好。最近最喜欢的电影或电视剧?
Adriel Frederick: 昨晚我在 Netflix 上发现了 Mo。
Lenny: Mo。
Adriel Frederick: 是 Mohammed 的简称。它是半自传体的,讲的是一个住在休斯顿的巴勒斯坦难民,他寻求庇护以及在这个多元文化环境中生活、工作和约会的经历,他流利地说阿拉伯语、西班牙语和英语。搞笑得要命,但也有戏剧张力。非常精彩。
Lenny: 太棒了。好的。哇,这些都非常独特。我很喜欢。换个方向,你最喜欢问的一道面试题是什么?
Adriel Frederick: 现在在工作中,我得走标准的面试题流程,但当我可以自由发挥的时候,或者有时候想多玩一下的时候,我会问类似这样的:教我一个你认为我不知道的东西。这很好地测试了我反复提到的那个东西——共情。我以前听 [听不清 01:04:58] 用过一次,然后我就不断尝试,看看它到底好在哪里。它能帮你了解一个人有多善于读懂你、他们的知识储备如何、以及他们沟通和分享知识的能力。所以它实际上能同时测试很多东西,而且很多时候你还能学到东西,太棒了。
Lenny: 好,最后一个问题。在业内,你最尊敬的思想领袖是谁?
Adriel Frederick: 从学科角度来看,[听不清 01:05:24],产品管理这个学科的话,肯定是 [听不清 01:05:26]。如果是从技术发展的角度来看,我会说是 Radiant Nuclear 背后的团队。
Lenny: 那是什么?
Adriel Frederick: 在两份工作之间的间隙,我在研究气候变化和能源,因为我的背景就是这个,然后我基本上确信核能是一个比我们目前认可的大得多的答案。很多障碍是政治上的,不是技术上的,但他们正在做的方案我认为是针对核能面临的一些政治问题提出的技术解决方案,这看起来非常有趣,我真的希望他们能做成他们想做的事。
Lenny: 哇。我喜欢这些推荐都这么与众不同。太棒了。Adriel,非常感谢你抽出时间。我也非常感谢 Jules 帮我们牵线。这次太棒了。你太厉害了。最后两个问题。大家想联系你、了解更多的话,在哪里可以找到你?然后听众怎么能帮到你?
Adriel Frederick: 好的。在我回答之前,谢谢你邀请我来这里。能够稍微反思一下生活和工作真的很好,希望也能跟你播客的听众分享一些有启发的东西。所以谢谢你的邀请。你可以在 LinkedIn 上找到我,Adriel Fredrick,可能还有一个同名的。但我很确定我是唯一的一个。然后听众怎么能帮到我?第一,继续收听这个播客,因为如果每个人都继续收听你所挖掘出的这些洞察,很多事情都会运转得更好,不一定只是对我。另一件他们能帮到我的事是,找一个跟你不一样的人,跟他们聊五分钟。就这样。我觉得这最终会回馈到我身上。
Lenny: 我很喜欢这两个答案,真的很受宠若惊,非常感谢。谢谢你来。
Adriel Frederick: 谢谢你邀请我,Lenny,保重。
Lenny: 非常感谢你的收听。如果你觉得这期节目有价值,可以在 Apple Podcasts、Spotify 或你最喜欢的播客应用上订阅本节目。另外,也请考虑给我们评分或留下评论,因为这真的能帮助其他听众找到这个播客。你可以在 lennyspodcast.com 找到所有往期节目或了解更多关于本节目的信息。下期见。
术语表
| 原文 | 中文 |
|---|---|
| AB5 | AB5(加利福尼亚州 Assembly Bill 5) |
| Adriel Frederick | 保留原文 |
| aha moment | 顿悟时刻 |
| biotech | 生物科技 |
| cannonball | 炮弹 |
| Chima | 保留原文 |
| Copilot | Copilot |
| cycles | 周期 |
| Daniel Yergin | 保留原文 |
| dead head | 空车返程 |
| Dolly | Dolly(AI 图像生成模型) |
| double E | 电气工程(double E) |
| flat head | 一字螺丝刀 |
| funnel | 漏斗 |
| growth hacking | 增长黑客 |
| imposter syndrome | 冒名顶替综合征 |
| issue tracker | 问题追踪工具 |
| Javi | 保留原文 |
| Jules Walter | 保留原文 |
| lead bullet | 铅弹 |
| Malcolm Gladwell | 保留原文 |
| marginal user | 边际用户 |
| medallion | medallion 车牌(纽约出租车运营牌照) |
| metaverse | 元宇宙 |
| Neuralink | Neuralink |
| onboarding | 新手引导 |
| orthogonal | 正交 |
| permanence | 持久性 |
| Phillips | 十字螺丝刀 |
| PM | 产品经理(PM) |
| Prop 22 | Prop 22(加利福尼亚州第 22 号提案) |
| Sam Harris | 保留原文 |
| silver bullet | 银弹 |
| Tom Allison | 保留原文 |
| torques | 梅花螺丝刀 |
| traction | 牵引力 |
| Zynga | Zynga(社交游戏公司) |
此文档由 AI 分片翻译(translate_long_document)