0:00

The LinkedIn Podcast Network is sponsored by TIAA.

0:03

TIAA makes you a retirement

0:06

promise. A promise of

0:08

a guaranteed retirement paycheck for life.

0:10

Learn more at tiaa.org

0:12

backslash promises pay off.

0:16

LinkedIn News. I've

0:22

always been curious. Always looking

0:24

for patterns in society. How

0:26

people navigate and find their way. You always

0:28

need some level of conflict to initiate

0:31

that dialogue. It's never a straight line.

0:34

I'm Tomer Coyne, Chief Product Officer of

0:36

LinkedIn. And this is

0:39

Building One. If I

0:41

had known the complexities of the music

0:43

industry, I would never have joined. Because

0:45

statistically, the chances of being successful with

0:47

a music startup already back then were

0:49

close to zero. That's

0:51

Gustav Stadlstrom, Spotify's Co-President and Chief Product

0:53

Officer. He's sharing with me just how

0:55

difficult it is to build in music

0:57

and the novel solutions they found to

0:59

stay ahead of the game. We're

1:02

going to get into that and so much more. So stick around.

1:11

Today's episode is a real treat. There are

1:14

just a handful of apps I use frequently throughout

1:16

the day. LinkedIn of course is one

1:18

of them. But another special one

1:20

is Spotify. As a music

1:22

lover who used to spend my entire paycheck

1:24

as a teenager buying music albums, the

1:27

delight of having a great music experience always

1:29

in my pocket is a very special one.

1:32

That's why I'm excited to end the

1:34

first season of Building One with a

1:36

very special guest and a remarkable builder,

1:38

Gustav Stadlstrom. Gustav

1:40

first arrived at Spotify in 2009 to lead

1:43

their transition to mobile. And he's been leading

1:45

the company ever since. He

1:47

is the Chief Product Officer at Spotify, but

1:49

also the Chief Technology Officer and the Co-President.

1:52

For Gustav, the labels don't matter.

1:55

His entrepreneurial experience has meant he's long

1:57

been adept at working and innovating across

1:59

disciplines. disciplines and domains. There's so much

2:01

to learn from this episode, but I'll

2:03

highlight a few items to get your

2:05

appetite going. How patience and

2:08

tenure are important for making big

2:10

long-term bets. Why

2:12

Spotify will never, quote, move fast

2:14

and break things. This

2:16

is a good one. Why being able to

2:19

explain yourself trumps having good intuition. And

2:21

how you can scale the benefits of that. How

2:24

Spotify built and developed its expertise

2:26

in music discovery over time. Of

2:29

course, how AI will reduce the cost

2:31

of coding to the point where we

2:33

can create truly personalized app experiences. Let's

2:36

get into it. So

2:39

you have back running technology. You

2:41

started multiple companies before you joined Spotify

2:43

in 2009 when Napstar was still a

2:45

thing. How did your career turn out

2:47

differently than what you expected? As

2:50

an engineer and electrical engineering data science,

2:52

my goal was to work in a

2:54

big company. I did

2:56

my master thesis at a Swedish company

2:58

named Telia on mobile peer-to-peer

3:00

networks and stuff. I

3:03

worked for the research department and it was total

3:05

nerds. I had a lot of fun. But

3:07

then this was the tail end of the

3:09

IT crash. The boom

3:11

and the bust. Telia merged with the

3:14

Finnish carrier named Sonra. They moved all

3:16

the research to Finland and I just

3:18

couldn't get a job. So I became

3:20

an entrepreneur, but involuntarily. It

3:23

was the only thing I could do because you

3:25

couldn't get a job when you just had a fresh

3:27

degree. And I had a lot of ideas and a

3:29

lot of cool stuff I wanted to do. The

3:32

flip side of not being able to get a job

3:34

was that neither could anyone else. So there was a

3:36

lot of talent available. They had no opportunity income, so

3:38

they were prepared to work for free. So

3:41

I joined up with a bunch of friends and

3:43

started a company around mobile data messaging, which back

3:45

then was a new thing. This was before the

3:47

iPhone. You could already start to

3:49

send text messages over data instead of

3:51

over the SMS protocol. So I

3:53

ended up becoming an entrepreneur and working as

3:55

a founder and a CEO. And

3:58

that was very much unexpected. There

4:00

was never my dream growing up. I actually saw

4:02

myself as a tech geek in a big company.

4:05

And here I am, tech geek in a pretty big company,

4:07

I'm curious later. That you helped build from

4:09

almost on the ground up. It's like the

4:11

notion of entrepreneurship as a necessity for you. Yeah,

4:14

exactly. There wasn't a big thing in Sweden. Being an

4:16

entrepreneur was a thing in Silicon Valley, almost

4:18

since the 90s, but not in Sweden. I actually

4:21

didn't know what it was even. So I had

4:23

to figure out how you start a company, how

4:25

you incorporate something, how you raise money, what a

4:27

VC is, like all those things. When you came

4:29

back to Sweden, which I

4:31

think is also remarkable for people who

4:33

want to build a career in tech,

4:36

you have this tremendous breadth and depth

4:38

across multiple disciplines, whether

4:40

it's technology, or user

4:43

experience, or business models, or go

4:45

to market. And what do you think was

4:47

the most important skill or quality that you developed over

4:50

the years that actually allow you to have this depth

4:52

and breadth across? It's hard

4:54

to know what is sort of nature and nurture,

4:56

how much is an inclination to want to do

4:59

full things. Like I said, I thought

5:01

of myself more as a geek or

5:03

researcher that would actually go quite deep

5:05

and narrow on something. But maybe it

5:07

was the fact that I had to

5:09

become an entrepreneur. So I've been both

5:11

founder and CEO of different companies. And

5:13

I guess having that role, even for

5:15

a few years, this forces

5:17

you to learn everything, have that

5:19

sort of helicopter perspective and the

5:22

ultimate responsibility for everything. Maybe

5:24

that's where I learned it, or at least remove

5:26

the notion that if you educate yourself as

5:29

something, you can mistake yourself for being that.

5:31

I'm an engineer, so I could never be this, or I'm a

5:34

business person, so I could never be this. But

5:36

that's not true, that's just a construction

5:38

on top. So I guess

5:41

I always try to have more of a

5:43

sort of CEO perspective, it's about getting the

5:45

job done. I was always very interested in

5:47

business models, even though I've never officially had

5:49

the role. My view was always that good

5:52

products might leverage a technology innovation or a

5:54

UI innovation, but great products almost always incorporate

5:56

a business model change. That's when you have

5:59

big impact, is just as fascinating

6:01

as technology in itself. A couple of

6:03

great insights there from you. One, it

6:05

sounds like your tendency is to go

6:07

deep, but you had

6:09

to go broad because of

6:11

the roles you had. So then just

6:13

allowing for that depth, I think you

6:15

find out connective dots, just going across

6:18

multiple spectrums, technologies. He also mentioned being

6:20

careful about being labeled. For me, like

6:22

at heart, I love building stuff. He's

6:24

an engineer, a product person, but I

6:26

love taking the two mark. There's so

6:28

many aspects of what I enjoy in

6:30

the process. So limiting yourself to

6:32

a function doesn't feel like a great way to grow

6:34

and to learn. I totally agree. I

6:36

guess I just don't enjoy the

6:38

problem as much if I can't see the entire

6:41

problem. It's the beauty

6:43

of solving the entire problem, and that

6:46

often incorporates some technology that enables something,

6:48

but then you run straight into the

6:51

user interface challenges of this new technology or how

6:53

to use it, and then you run

6:55

straight into the business model of how you get

6:57

this into the hands of people. I mean, I've

6:59

been labeled in that sense. Sometimes people say like,

7:01

how can you be both the CPO and CTO

7:03

and so forth? But that I think

7:05

is actually a little bit false. At

7:08

least to me, it doesn't feel like extra work. I

7:10

was always interested, and it's very much passion driven. Could

7:12

not agree more. You mentioned that

7:14

great products also ring with them, a great

7:16

business model. Spotify is quite unique in that.

7:19

Do you want to talk about like that

7:21

combination? Because I know the early days, like

7:23

what you've done in retrospect back then was

7:25

not very clear for people. And

7:27

my first companies, they were

7:30

not great successes as business models. Some

7:32

of them had great technology, and we

7:35

sold them, but they were in between mostly

7:37

Aqua hires. So we never sort

7:39

of cracked a new business model, but I learned a

7:41

lot about business models. And also working

7:44

at Yahoo, I learned about the search ads, business

7:46

model, and so forth. But it was really at

7:48

Spotify that I got to cut my teeth in

7:50

combining business model and product. So when

7:53

I joined Spotify, it existed as a

7:55

free ad driven desktop product. Spotify was

7:57

competing with piracy, piracy was free. not

8:00

interesting, paying for music, because Spotify had to be free.

8:03

And the idea was simply it could be like radio,

8:05

but on demand. You can monetize it with ads just

8:07

like radio does. And then when I

8:09

came in, the iPhone has just come

8:12

out. This was in 2008. The app store hadn't yet come

8:14

out, but it was in the cards. So

8:16

Daniel asked me together with the team to figure

8:18

out what Spotify mobile should be. And

8:20

it wasn't just the product. It was the

8:22

proposition. It was not that hard

8:25

to develop an iOS client using some of the same code

8:27

and so forth. But the thing was, if

8:29

you were on Wi-Fi, it was pretty straightforward

8:31

to get the Spotify product to stream music.

8:33

But the problem was that this product was

8:35

not going to compete with your desktop computer

8:38

that had low latency, high

8:40

bandwidth broadband. It was

8:42

supposed to compete with your iPod that

8:44

had ubiquity, could play anywhere, even without

8:46

a connection. And so back then, the

8:48

mobile networks, even in Sweden, that was

8:51

a pioneer, was terrible. It was Edge

8:53

Networks at best, not even 3G back

8:55

then. You could start

8:57

streaming a song using sort of the same core

8:59

engine. But instead of being instant, the Spotify was

9:01

famous for as fast as the MP3 that you

9:03

would pirate from Pirate Bay, it did take 20

9:05

seconds to start, then it would stutter, and then

9:07

you'd be out of data for the entire month.

9:10

So the whole thing of Spotify being, you don't

9:12

have the files anymore. They're in the cloud. You

9:15

don't pay. It didn't work. So we had

9:17

to rethink the entire thing. And we actually came up

9:19

with something that at the time was very contrarian, which

9:21

was, forget about us telling

9:23

you that you shouldn't have the files anymore. Actually,

9:25

you should get the files. And we went and

9:28

licensed a completely new model, which was

9:30

you can offline sync 10,000 files

9:32

for up to a month using sort

9:35

of local encryption to make sure that when the

9:37

encryption key expires, you can't play your files anymore.

9:40

That was very different. It was a sync paradigm.

9:43

And we had to negotiate with the labels to figure

9:45

out what this cap should be, commencement of the technology,

9:47

that the key would expire and you couldn't export those

9:49

files, and that users wanted to pay for this. And

9:51

I thought it should cost like a few dollars. They

9:53

thought it should cost $10 a month. Clearly

9:56

it became $10 a month. And

9:58

back then, was it clear that But mobile is

10:01

going to be the device you're going to listen

10:03

to music to? We

10:05

knew already then that listening on the go, the

10:08

car in your iPod was way bigger than

10:10

listening at home. So that was already clear.

10:12

So it was clear that desktop was not

10:14

supposed to be the main product. Guestop was

10:16

just a way to start. Yeah, it always

10:18

had to be mobile, but it wasn't

10:20

clear that it even could be. We

10:22

even synced files locally to an

10:25

iPod for a while. We actually reverse

10:27

engineered the iPod protocol and

10:29

synced our files. It was crazy. But

10:32

we didn't foresee the smartphone becoming as big as

10:34

it was. We got very lucky in that

10:36

sense. But it was clear that listening was going

10:38

to be mobile. But for a while, we

10:40

thought it was going to be the iPod, and we

10:42

just didn't have access to the iPod to the same

10:44

extent. So for us, it was

10:46

fortunate that the smartphone way had happened. Incredible.

10:49

And you mentioned working with partners around

10:51

figuring out the licensing model. You were

10:53

in a very diverse

10:55

and somewhat rigid ecosystem of

10:57

partners. That can take

10:59

a toll on building the ultimate product

11:02

experience you run. I could see that

11:04

being a struggle around, I know what's the

11:06

ideal user experience, but I can't do it

11:08

because there's kind of business-oriented

11:11

restrictions on the other side. For

11:13

sure. I've always felt a little bit

11:15

jealous about my peers at Metz or Twitter. They can

11:17

just think about something, use the research, and then say,

11:19

this is the right thing to do, and then just

11:21

build it. And I would have to use the research,

11:23

figure out this is the right thing to do, and

11:26

then go and negotiate like a lowest common

11:28

denominator between three or four major publishers. And

11:31

then hope that it still was an okay

11:34

product. So very frustrating at times. Now

11:36

I think if I had known the complexities of the

11:38

music industry, I would never have joined

11:41

because statistically, the chances

11:43

of being successful with the music startup already

11:45

back then were close to zero. There are

11:48

so many failed startups. Luckily

11:50

for me, I was naive enough that I

11:52

just didn't understand, in the true sense of

11:54

ignorance being bliss. I

11:57

was so impressed with the technology. And

12:00

with Daniel and his vision, already back

12:02

then, he talked about taking

12:04

over the world and go big or

12:06

go home. And this was way back

12:09

when it seemed silly. This was a small

12:11

Swedish product. It was very cool in Sweden, but

12:13

nowhere else. So I think I got enamored with

12:15

a product and the team. So that's why I

12:17

joined. And then I think

12:19

I'm actually very patient. That's

12:23

my key strength is probably not

12:25

that I'm better at anything than anyone else, but

12:28

I am very inclined to being bored for a

12:30

very long time and not give up. I

12:32

just accept a lot of, not

12:35

necessarily pain, just boredom, like

12:37

things going slowly. You know, it's funny.

12:39

You were saying not many music startups

12:42

or companies lasted, and I was just going to the app store to

12:44

look at them. And right

12:47

now you have Spotify, Apple Music,

12:49

there's YouTube Music, but there's

12:51

so many lists of like music company startups that existed

12:54

for a year, a year and a half. Took

12:57

off, but then kind of died almost at the same pace they

12:59

took off. Like

13:02

when you build a product for the long game,

13:04

is it a lot of deliberation, a lot of

13:06

principal thinking? Like demystify this a little bit. It's

13:09

easy to sort of post hoc reconstruct greatness. I

13:13

don't want to do that too much. A lot of it is luck. But

13:15

on the other hand, you don't get only lucky for

13:17

15 years in a row. So there's

13:19

something systematic to it. I think

13:21

a lot of it goes back to Daniel and his

13:24

sort of tenacity. So Daniel

13:26

has this incredible skill where

13:28

he's definitely back then, like

13:30

a nobody from Sweden who

13:32

goes to meet these huge personalities

13:35

of the world's biggest music labels and somehow manages

13:37

to get them to agree to what he wants

13:39

to do. I still don't quite understand

13:41

how that happens, but

13:44

I've noticed about Daniel that he manages to get

13:46

everyone to like him. He's a really nice guy,

13:48

so I understand why they do. But

13:50

it's interesting. Even people that don't like each other, they usually

13:52

agree that they like him. They have

13:54

this uncanny ability to also think with a

13:56

lot of patience. Just get things done. that

13:58

are hard to do with a lot of

14:01

great people around him in terms of lawyers

14:03

and so forth. But I think he has

14:05

that ability to inspire patients and

14:07

he has a clear vision for what he wants

14:09

to do. So I think many

14:11

of us have sort of adopted that around him. Having

14:15

done this for a while, I do think it's actually

14:17

built a skill, which is Spotify is

14:19

very good at this licensing business,

14:22

working with these behemoths where you

14:24

negotiate contracts that are multi-year and you

14:27

have to set requirements and so forth. And

14:29

I think what's interesting about that, that is

14:31

maybe a little bit different than other companies,

14:33

is that the cost of being wrong is

14:35

quite high when that feature set is written

14:37

into a deal that expires only in four

14:39

or five years and you

14:41

have to pay huge amounts of money. So

14:44

the decisions are very consequential. So like these

14:46

quick A-B tests of everything are not really

14:48

possible to do. So you both

14:50

have to try to test your way into

14:52

the things that you're gonna bet on to the extent you

14:54

can, but you don't have the licenses to test everything. So

14:57

you have to be quite deliberate and strategic. So

15:00

we've tended to discuss a lot of strategy in

15:02

my teams about what ifs and

15:04

game theories and arios and what happens if

15:07

they do this and we do this and what is the landscape

15:09

gonna be like in two years and so forth. I think that's

15:11

built a certain kind of a skill

15:14

that I haven't necessarily seen in other companies.

15:16

I don't think it's even right in other

15:18

companies because why would you spend

15:20

all that time strategizing when you can actually test?

15:23

So that is maybe a little bit different,

15:25

but we've realized that over the years and

15:28

try to lean into it and instead of doing less of that, because

15:30

it's very painful, we're now saying, let's do more

15:32

of that because it's very painful. That's

15:34

probably the area that other companies are not gonna follow

15:37

because we're really good at it. Is

15:39

there something to building away from Silicon

15:41

Valley? Is it being in

15:43

Sweden to an extent kind of allows you

15:45

freedom to think differently, by

15:48

inherently being a different company? I'm curious if

15:50

I'm making a leap there or if there's

15:52

something there. No, I think there's definitely something

15:54

to that. There are a couple of advantages,

15:56

obviously disadvantages as well with being in Sweden.

15:58

You're further away from. where the core

16:01

of the tech culture is, and

16:03

the latest discussions and findings and so forth.

16:05

So it's required a lot of traveling, remote

16:07

relationship building, trying to stay up to date.

16:10

So there are definitely downsides with being this far

16:12

away from like the center of technology, but

16:15

there are some real benefits. One

16:17

is that pretty early on,

16:20

Spotify in Sweden was

16:22

kind of like Google in Silicon Valley when Google

16:24

was at its peak as an employer. We

16:27

were the biggest in our market. So

16:29

you had quote unquote unfair access to the best

16:31

talent. You know, we had an office

16:33

in Silicon Valley for a while, but it's not great

16:35

having your back office where your competition has their main

16:37

office. So there were just geographic

16:40

advantages to being in Sweden, but it

16:42

was also cultural because Daniel certainly brought

16:44

a lot of the Silicon Valley culture.

16:46

I brought some of that and those

16:48

findings like challenging ourselves, but also we

16:50

were far enough away that we

16:52

always figured like, maybe we should do this a little

16:54

bit our way. And we had this

16:56

Swedish culture of more consensus and so forth, which

16:58

at times has slowed us down, but at times

17:01

has also helped us tremendously.

17:03

So I think you're right that being that

17:06

far away let us develop a bit of

17:08

our own culture. And certainly I think our

17:10

self-confidence in that has changed over the years.

17:13

You know, while we try to take the

17:15

best of these other companies, we developed our

17:17

own cultural skills and that started paying off.

17:19

We realized that we wanted to be as

17:21

different as possible in a sense from other

17:23

companies. You've seen both cultures closely, like what

17:25

stands out for you around those advantages that

17:27

you've built in. It's hard to know

17:30

what part of it is the

17:32

culture versus geography, et cetera. We

17:34

just have very, very high retention.

17:37

I've worked there for 15 years, but my co-president Alex

17:39

Nordstrom, he has worked there for 14 years. And

17:42

many of the people who work in my team, they worked for me

17:44

for 12 years or 10 years. We have

17:46

very long tenure. That was a

17:48

clear advantage versus being in Silicon Valley. If

17:51

your horizon is one and a

17:53

half at the most three years, then

17:56

as a leader, you're just not gonna attempt to

17:58

build a product that's gonna take four years. Yeah,

18:00

makes no sense for you because you're not gonna

18:03

even be around right? So I think that tenure

18:05

meant that many people made bets that

18:07

were quite long-term You know, they

18:09

may have thought they would take three years to play out But they

18:11

took five years to play out and that

18:14

that has benefited us. I think

18:16

versus some other companies And

18:19

then we try to lean into that so there

18:21

was a tremendous pressure to use them the metaphrase

18:23

move fast and break things There

18:26

was this pressure to just take don't talk, you

18:28

know code decides arguments just move move move But

18:31

as I said that didn't really make any sense in our

18:33

world because the cost of being wrong was so high So

18:36

we try to counter that and I tried

18:38

to develop counter phrases that would annoy people

18:40

like, you know Talk is cheap So

18:43

we should do more of it because it's

18:45

much cheaper than writing code and certainly much more

18:47

cheap than shipping the wrong thing and Rolling

18:49

it back for for six months. So we try

18:51

to build into the culture these things that were

18:54

a little bit counter to

18:56

the mainstream culture Yeah Internally

18:58

we at LinkedIn deliberate a lot because we're

19:00

trying to build something which is differentiated from

19:02

the pack Like we're not building an epidemic

19:04

product building a productivity product, which is very

19:06

different We get lumped into the social network

19:09

ones. I'm like you don't get us ultimately

19:11

you come to think to check in not

19:13

to check out Exactly and that comes with

19:15

a lot of deliberation. We have like the

19:17

Socratic method of really discussing film through it

19:20

I would love to test 10x more things

19:22

But not at the expense of talking them

19:24

through and that's healthy tension to have yeah,

19:26

this kind of goes into how you think

19:28

as a leader One

19:31

of my big heroes is the physicist David

19:33

Deutsch and he talks a lot

19:35

about why our explanations are valuable So

19:38

I think that in a company People

19:40

don't necessarily need to agree with you. That's

19:43

too much to ask That's actually problem in the

19:45

Swedish consensus culture that everyone needs

19:47

to agree But I think everyone

19:49

deserves an explanation and you as a leader

19:51

need to be able to explain yourself Why you're

19:53

doing something and then people can say like well I don't I

19:56

don't agree with the premise of this and that and those assumptions,

19:58

but I understand why you do it And

20:00

I think that's fine. What I don't

20:02

like is when people say, I

20:04

can't explain this to you. It's like above

20:07

your pay grade, you're not smart enough. It

20:09

usually means that the person saying it doesn't

20:11

really understand. So I'm trying to force

20:13

people to explain themselves. And

20:16

I'm trying to force myself to explain myself. And

20:18

I think a great way of doing

20:20

that is to use models, so explanations.

20:24

Models are dangerous, in a sense, because they're not the

20:26

truth. Model is a simplification of the truth. And

20:28

if it's too low dimensional, you can miss

20:30

an important dimension. But if you have either

20:34

a fairly complicated model, or at least a

20:36

few different models with different dimensions, you can

20:38

triangulate something. And you can get to a

20:40

pretty decent prediction. And the benefit

20:43

of a model versus an opinion is

20:45

that it scales. If you explain

20:47

that model, if you write it down, if

20:49

you teach it to people, it just spreads

20:51

over the organization. Has a life of its

20:53

own. That's the benefit of explanations. They have

20:55

this fantastic power. And I want

20:57

to try to push for explanations. Now, that's dangerous,

20:59

because there is such a thing as

21:01

pattern recognition and instinct. And if you've worked for

21:03

20 years, you've seen a ton of stuff, you

21:05

may have really good pattern recognition that you can't

21:07

fully explain. And that's valuable.

21:09

It's called seniority. So I don't want to

21:12

discount it. But I want to try

21:14

to push for explanations. Because ultimately, people are

21:16

trying to learn together. It's worth trying to

21:18

say, hey, this is what I'm basing this

21:20

on. I might be wrong.

21:22

But that's what my intuition tells me. And

21:24

I've seen this through. And I've seen this

21:26

through. Maybe an overgeneralizing. But that's what I

21:28

believe needs to happen. Now,

21:30

the other side, they might disagree. They might challenge you.

21:33

But it's not like you're leaving stuff off the table. And they

21:35

don't know if they're seeing the complete picture. And

21:38

I think that just builds a better organization. An

21:40

instinct of great pattern recognition is very valuable. I

21:42

want that in my senior leaders. It's

21:44

kind of second price. It's like silver. But

21:47

first price is if you can explain why you

21:49

have that intuition. Because then people can take it.

21:51

They can run with it. They can answer 15

21:53

other questions using that model that they didn't even

21:55

ask you. It's not always possible. And they can

21:57

slow you down. But that would be my ideal

21:59

state. Exactly. Let's

22:01

shift to the Spotify experience. I

22:04

love my Spotify account, the whole

22:06

family. I think I'm jokingly

22:08

in the past that I'm saying I'm almost like an

22:10

hourly user to an extent. Those are the best. Even

22:12

at work, I would put stuff for myself so I

22:14

can focus. Hey, JU's, hourly active user. Yeah.

22:18

When it comes to music, I'm generally an hourly

22:20

user. I was

22:22

thinking a lot about this dichotomy. It feels

22:24

like there's an explicit and implicit mode, at

22:27

least for me. The explicit mode is I

22:29

have a song running in my head. I

22:31

just want to play it. I don't want

22:34

to have any friction in my way. I just want to play it

22:36

through. It's almost like satisfaction, right?

22:38

Something is playing. I need it. I want to play it through,

22:40

at least for me. Then the

22:42

implicit side is more

22:44

of the discovery, the light experience. This is where

22:46

I'm not expecting you to give it to me,

22:48

but you gave it to me. And that just

22:51

elevated my experience as a whole. And for

22:53

me with music, like literally, you can feel

22:55

elevated by the discovery of it. And

22:57

I was wondering if that construct is

22:59

correct. And if so, do you

23:02

think about building for those separately? One is just

23:04

get the fundamentals right. Tomer wants

23:06

a song. Give it to him as

23:08

quickly as possible. And the other one was when he

23:10

is open, you know, just pick

23:12

one song. I didn't pick the song after. Go for

23:15

it. Try to bring something in. Yeah,

23:17

I think it's a great observation. And it's also

23:19

like a journey over time. There was a

23:22

play listing tool. So like you could search for the tracks

23:24

and you could build your own experiences using

23:26

Playlist. It was a lot of work on you. You

23:29

had to know the entire back catalogue in the back

23:31

of your head. It's like, I want this song for

23:33

that moment. You had to keep up with all the

23:35

new releases that came every day. But if you were

23:37

good at music, you could build great experiences. And then

23:39

people started building experiences for each other. People

23:42

start listening to other playlists. So

23:44

we saw that some people that had very high

23:46

intent know what they wanted to do. Many

23:49

of them sort of playlisted in traditional genres, but

23:51

more and more they playlisted sort of almost for

23:53

use cases. It's like work

23:55

out of background or studying or dinner

23:58

table. And so we actually got a lot of. ideas

24:00

for what the different modes of music listening were

24:02

from the playlist data. And

24:04

then we kind of said, okay, it's great if

24:06

people can find that socially or happen to have

24:08

a friend that can send that list, but most

24:10

people don't have that music friend. We

24:13

realized, you know, maybe we could create that friend for

24:15

everyone. So we had editors sort of identifying

24:17

the use cases like songs to sing in the

24:19

shower or songs to sing in the

24:21

car. You can almost think of them as product managers. Like

24:23

the work the product manager is to understand the use case.

24:26

What mode is Tomerian? What is he

24:28

trying to do? Is he relaxing with friends in the

24:30

kitchen? Is he driving? What is going

24:32

on? And then they create that

24:34

use case, literally the image and the label, just

24:36

fast forward like a few thousand songs to sort

24:38

of you can think of it almost as telling

24:41

the system what this use case sounds like. And

24:44

now that the machine learning algorithms can understand, oh, that's what

24:46

songs this thing in the car is like. And

24:48

they can pick out another few hundred thousand out of the,

24:50

you know, 100 million catalog. And

24:52

then when we deliver it, we can still personalize it

24:54

to you. So users were actually doing the use cases

24:57

for each other. And they covered these different modes that

24:59

you're talking about. And as

25:01

they started play listening for others, we got more users.

25:03

And then we realized if we started play listing for

25:05

them, we brought the friction down even more. And

25:08

eventually we went into AI and we started personalizing

25:11

and combining them with editorial. And that's a

25:13

big reason why the product scale and what's

25:15

so beautiful about music is exactly that, that

25:17

it's used in many, many different

25:19

contexts. So we try to

25:21

reflect that in the UI. When you come in, we

25:23

have a pretty dense layout and we try to cover

25:26

like the different use cases that you may have. There

25:28

is certainly your liked songs, which

25:31

a lot of people listen a lot to, for example, you

25:34

have your playlist, but then you have something like the

25:36

AI DJ, for example, which is this

25:38

drastically, just, I don't even know what I

25:40

want. Just click that button. And

25:42

people go between these use cases. Sometimes they're very

25:44

lean in and come in with an opinion. And

25:47

sometimes they have no opinion. They want to be

25:49

entertained. And that's a little bit

25:51

different from other products. I think if you open

25:53

maybe Instagram, it is always the

25:55

intent. I just want to be entertained. I don't know who's going

25:57

to be there. I don't know what it is. Just a entertain

26:00

me. With Spotify, it's a little

26:02

bit tricky because sometimes you come in and

26:04

say, entertain me. Sometimes you

26:06

come in and know exactly what you want to do.

26:08

And our UI just gets in the way. We have

26:10

lots of recommendation for other stuff that

26:12

you didn't want to do. So this is actually something

26:14

you've been struggling with, finding what the Spotify UI is,

26:17

because you are in different modes within

26:20

music. But now to make it more complicated,

26:22

you also may be in different sessions. You may

26:24

be in two different podcasts, and

26:27

in an audiobook, and also listen

26:29

to music. And so we're trying

26:31

to keep all those states for you, so you

26:33

can quickly go back. But we also can't

26:35

just guess on one of them. It doesn't work. We're not

26:37

good enough at predicting. So we need to

26:39

kind of show a mix of those. It's a challenge, actually.

26:42

We're going to take a quick break, but

26:44

don't go anywhere. When we come back, Gustav

26:46

is going to share with us how he

26:49

believes the AI will completely transform music and

26:51

how we build products. Some people say, my

26:53

Spotify and sort of as a friend, but

26:55

it was always an analogy. I think it's

26:57

entirely possible that some of these products actually

26:59

become your friend. The

27:03

LinkedIn Podcast Network is sponsored by

27:05

TIAA. In the last 100 years,

27:08

we've seen financial markets swing, new

27:10

currencies come and go, decades of

27:13

savings lost in days, all showing

27:15

that a retirement plan without a

27:17

guarantee quite simply isn't enough. So

27:20

more than a retirement plan, TIAA

27:22

makes you a retirement promise. A

27:25

promise of a guaranteed retirement

27:27

paycheck for life. A promise

27:29

that pays off. Learn more

27:31

at tiaa.org backslash promises pay

27:34

off. From LinkedIn News, I'm

27:36

Jessi Hempel, host of the

27:38

Hello Monday podcast. In

27:40

my 20s, I knew what career

27:43

success looked like. In

27:45

midlife, it's not that simple. Work

27:48

is changing, we're changing, and there's no

27:50

guidebook for how to make sense of

27:53

it. So come figure it out with

27:55

me on the Hello Monday podcast. I've

27:58

been a journalist for two decades.

28:00

writing cover stories for Business Week,

28:02

Fortune, and Wired, and now,

28:05

every Monday, I bring you

28:07

conversations with people who are thinking deeply

28:09

about work and where it fits into

28:12

our lives, like Microsoft CEO

28:14

Satya Nadella on growth mindsets.

28:16

The learn it all does

28:18

better than the know it

28:20

all. Or NYU professor Scott

28:22

Galloway on choosing a career.

28:24

I think the worst advice

28:26

you can give a kid

28:29

is follow your passion. Or

28:31

MacArthur Genius winner Angela Duckworth

28:33

on talent versus grit. Your

28:35

long-term effort and your long-term

28:37

commitment are surprisingly important. Each

28:39

episode delivers pragmatic advice for

28:41

right now. Listen to Hello

28:43

Monday with me, Jesse Hempel

28:45

on the LinkedIn Podcast Network

28:47

or wherever you get your

28:49

podcasts. All

28:54

right. I'm back with Gustav Satterstrom.

28:56

The co-president of Spotify. The AI

28:58

component is where I usually get

29:00

very giddy and excited. And there's

29:03

understanding the music side of it, just

29:05

understanding the song or even

29:08

the artist to an extent. And then there's understanding

29:10

the user. So for example, if

29:12

you know I'm in a

29:14

meeting right now, I'm at work, then

29:16

probably it's not rock

29:18

music, but it's probably more focused music.

29:21

Maybe it's me imagining, but I'm feeling like the

29:23

app is starting to get this from me. Like

29:25

I do my meditation in the morning, it shows

29:28

up first kind of thing. And so

29:30

I'm wondering, like, was there

29:32

any breakthrough features that really changed

29:35

how the matching is being done? So there

29:37

are like three phases. The first was we

29:40

got all this playlisting data. And

29:42

what happens when we group things together and give

29:44

it a name, it was just the grouping. So

29:47

we had, you know, pretty early on billions of

29:49

examples. You have the cluster. Exactly. And the way you

29:51

have unsupervised learning done by others

29:53

for you. Exactly that. So I might put together

29:55

a bunch of songs and say, this is EDM.

29:57

We might not even know it's EDM. but

30:00

it is actually the perfect training data, saying

30:02

these songs are close to each other. So

30:04

we built similarity vectors between songs and playlists,

30:07

which we got a ton of mileage out of, and

30:10

that's really how we understood the music. But it's interesting

30:12

because we didn't understand anything about what was in the

30:14

music. It was just IDs on

30:16

music. Just these numbers are close to

30:18

these numbers in a vector space based

30:21

on play listing. Incredible. So that gave us

30:23

some things that no one else had. It

30:25

gave us these really esoteric discovery experiences where

30:27

if you're into this minimalist music, it

30:29

turns out if you're millions of people, there are going to be at

30:32

least tens of thousands of playlists from minimalist music.

30:34

So we found these songs that

30:36

people are like, how could you possibly know? So

30:39

that was great. But we had another

30:41

problem because we didn't understand anything about the music.

30:44

We could also play you some heavy

30:46

metal and then Justin Bieber. Because statistically,

30:48

for one reason or another, this happened

30:51

in the data. We had

30:53

no understanding that these were different genres. There

30:55

were no genre understanding, just similarity vectors. So

30:58

we actually made this acquisition of a company called

31:00

the Econest many years ago, which

31:03

was a big bet on AI, but they

31:05

were actually doing a lot of classification and

31:07

heuristics around genres and understanding the

31:09

actual music. So it was really when we made

31:11

that acquisition, we had all the similarity data and

31:13

then we bought this company that could say that,

31:16

yes, but this is actually also Christmas music, but

31:18

this is actually also Swedish pop. So

31:20

then we bought ourselves a later on a

31:22

standard of semantic understanding. So

31:24

we had mathematical understanding of similarity. We

31:27

lacked the semantic understanding. Once we had them,

31:30

then we could start building great sessions, where we could

31:32

actually promise you that this is going

31:34

to be Swedish pop, not just these songs are

31:36

similar in a vector space. That

31:38

combination was really powerful. So I would say those

31:40

two things were important. You

31:43

can look at music discovery today. It's like it's really good.

31:45

I couldn't see where else we can go. You can look

31:47

at it and saying, wow, we were just at the early

31:49

innings because imagine understanding

31:51

the user better. So I'll give my example

31:53

of just thinking about it. I

31:55

work out in the morning. Usually my workout

31:58

starts slow and I want some like... calm

32:00

music when I do my initial stretching and

32:02

basic drills, but then it becomes very intense.

32:04

And I don't want to create a playlist

32:06

for that. I would love you to be

32:08

able to kind of read my heart rate

32:10

on my wrist and knowing the music I

32:12

want, and then you become natural. And

32:15

then I think about the spirit elevation you would get

32:17

from that, somebody who loves music. That

32:19

feels incredible. And that's just one thing. Yeah,

32:21

they actually built at one point like a

32:24

running experience that did something like that, they

32:26

measured your cadence as they would

32:28

adapt a single song. We were actually

32:30

recording many versions of the same songs with the

32:32

Tiesto and a bunch of classical music. They would

32:34

actually adapt your cadence. And it was

32:36

a beautiful experience, but it was a business model learning as well,

32:39

which was that we needed a big

32:41

supply of content for that experience to fit

32:43

everyone's taste. And because it was

32:45

so expensive to record these things in

32:47

studios, the back-end model of

32:49

it didn't scale. We just couldn't produce enough content.

32:52

But the experience was amazing. I

32:54

think now with AI, it's very straightforward how you

32:56

could just make the song change. Back then you

32:58

needed to record different versions. So I'm

33:00

very excited about looking at those things again,

33:03

now that we have the technology that makes it

33:05

affordable. When I looked at Spotify's mission

33:07

statement, it's not just music, it's the whole creative

33:09

space that you guys are looking at. What

33:12

kind of excites you when you look at it, you're like, I

33:14

don't know exactly how I do it, but I know it's possible

33:17

and we're gonna go for it. Like you, I think this is

33:19

one of the most exciting times and product because you have this

33:21

new technology of AI and

33:23

specifically maybe generative AI. And so I

33:25

think it's different things in

33:28

different categories. For music,

33:30

the goal was always to build

33:32

that perfect session as if you

33:34

had a personal DJ

33:36

that knew you as a user could

33:39

talk to you and say like, hey, Tamara, I've been up

33:41

all night looking through all the new releases and like, look

33:43

what I found for you. But it

33:45

was an analogy for a long, long time

33:47

because it was actually an editor that, you

33:49

know, play listed for 2 million people

33:51

that was personalized. But it's starting

33:54

to become true. That's the promise, you know, we

33:56

think we can actually build that perfect session with

33:58

a DJ that... understands you, understands what

34:00

happened in the world of music, have read the news

34:03

in the morning, can talk to you about your music,

34:05

tell you anecdotes about the bands you didn't know,

34:08

play, mix, all of these things. It's

34:10

now entirely possible. We're not quite there

34:12

yet, but this is my dream and

34:14

my passion. To build that perfect session,

34:17

we try to imagine this product. We

34:19

say, what if we could afford to

34:21

hire three human editors per user,

34:23

working in shift? Just

34:26

working 24 hours a day, sitting there, going

34:28

through all the new releases, all the back

34:30

catalog, thinking about Tomer, Tomer, Tomer, like they

34:32

interview you, and they just work for

34:34

you all the time. You could simulate that experience if you had a

34:36

lot of money and have experts work for

34:38

you. That's the experience we're trying to go for. And I

34:40

think it's going to be that. So

34:43

that's what I'm excited about in music. In podcasts,

34:45

it's different. I don't think it's going to be

34:47

a voice that replaces creators. I think

34:50

it's about someone helping you to find creators. There are

34:52

different problems to be solved there. There

34:54

are very exciting things about translating creators

34:56

into many languages so they can have

34:58

bigger audiences, helping them summarize, helping them

35:00

get discovered, and so forth. If

35:04

I would say it's sort of macro, the thing I

35:06

think it's possible now, back

35:08

to analogies, is that

35:11

I think products might actually become truly

35:13

personal. So we've talked about Spotify being

35:15

personal for a long time. Some people

35:17

say, my Spotify, and sort

35:19

of as a friend, but it was

35:21

always an analogy. I think it's entirely

35:23

possible that some of these products actually

35:25

become your friend. I don't think it's

35:27

very controversial even after open AI. These

35:29

things can become friends. But you just

35:31

described for me sounds remarkable. I was

35:33

thinking about the last few

35:35

decades, and both you and I came

35:38

from communication engineering where you learn

35:41

how to talk to machines. How everything

35:43

you've been taught in school was how do I talk

35:45

to a machine and get it to do what I

35:47

wanted to do. And now machines are learning how to

35:49

talk to you in a way. Like it's shifting. Exactly.

35:52

And you were talking about that kind

35:54

of person, almost like a personal trainer for you

35:56

for music. That can actually see like, am I

35:58

breathing heavily? Like, it might be a... working out,

36:00

like the voice, the tone, the facial expression, if

36:02

you can actually have a camera kind of looking

36:04

at you. There's so many things you can pick

36:06

up that you just can't pick up from text

36:09

or just when I open a session. Almost like

36:11

the app is in the background. So like imagine

36:13

a world where Spotify the app is

36:15

not the main thing. Spotify, your trainer, is

36:17

the main thing. That's exactly what I would

36:19

imagine could happen. You know, a dream scenario

36:21

would be you walked down the

36:23

street in Manhattan, you hear voice,

36:25

you know, and you're like, oh, that's Spotify, because

36:28

Spotify has turned from like this utility in

36:30

this app into actually like a

36:33

friend or a voice that you would recognize. And

36:35

that's how you think of it. That's at least a dream of

36:37

mine. We're not quite there yet. But it's, it's

36:39

very clear how that could happen now. And I

36:41

think it's very exciting, because it kind of challenges

36:43

everything. To a point, it's sort of much more

36:46

dynamic, it questions a bunch of things. It's both

36:48

scary, obviously, but also very exciting,

36:50

I think. Yeah, I have no doubt

36:52

that one of the biggest relationships you'll

36:54

have in the future is with your

36:56

AI. And in fact, there's a

36:58

word for the anxiety of being away from your

37:01

mobile device, which is called nomophobia. For

37:03

most people, the phone is always on them. It's like

37:05

there's actually an anxiety of living at home or going

37:07

to work without it, even going to the grocery

37:10

store without it. Then when I

37:12

think of how intimate AI is going to

37:14

be to you, nomophobia is nothing versus what

37:16

you're going to see in future. Yeah, for

37:18

sure. And they're obviously both very interesting and

37:21

scary things about that. But I

37:23

think what I like about the

37:25

music podcast and audiobooks

37:27

world, it's not unambiguously good, but almost

37:29

unambiguously good. At least music, most people

37:32

consider being good. And then most people

37:34

agree books are good for the world,

37:36

they'd be terrible if they disappeared. So

37:38

I feel like I'm in a space

37:41

where AI can help make humanity better,

37:43

more relaxed through music, more educated through

37:46

books, more up to speed on science

37:48

and education through podcasts and so forth.

37:50

So I'm very excited about being one

37:52

of the few people who get to

37:54

work with applying AI to that space.

37:57

It's probably the most exciting area

37:59

since mobile phone. Because back then it was

38:01

also like, oh, everything is going to change now. Nothing

38:04

is true anymore. Have to rethink everything.

38:06

Scary, but very exciting. Yeah, you know,

38:08

Gustav, you recently went through quite a

38:10

big reorg where you centralized a lot.

38:13

I'm curious how your product shaped your organization. What

38:15

sort of changes have you made along the way?

38:18

So one of the things that get a

38:20

lot of questions on is doing all of these things in

38:22

a single app. You know, did you

38:25

ever think about maybe doing separate apps? Because

38:27

you know, a dedicated podcast app, you could have

38:29

like all the podcast features you

38:31

want, dedicated music app, dedicated audiobooks app,

38:33

you just optimize more for the user.

38:36

And this was actually a strategic decision we made.

38:38

I actually agree with that. We could

38:40

do three better apps than the single

38:42

one if you optimize only for the amount of

38:44

features you could have with uploading the

38:47

UI. But it was

38:49

a strategic decision because unlike many of

38:51

our competitors, like Apple or Google, like

38:53

pre-install themselves on all the iPhones in

38:55

the world, we don't have that kind

38:57

of distribution. So if we manage

38:59

to get ourselves installed on your phone, it

39:02

is very valuable for us to be able to double down

39:04

on that existing distribution. And you

39:06

know, we didn't come up with this. You've seen the

39:08

Chinese super apps do this, but in the West, it

39:10

was not very common. So we kind

39:12

of adopted that pattern. The second

39:14

thinking we had that was more futuristic and not

39:16

so much just for strategy, but for where we

39:18

think things are going. I think it's interesting in

39:21

the age of AI is that as a user,

39:23

what are you doing when you're switching between the

39:25

apps? I mean, it's all code.

39:27

It's all software. So why

39:29

couldn't just a software adapt the UI instead of

39:31

the user adapting the software? You

39:34

know, think about you have three different apps, you could technically

39:36

merge those three code bases and just switch the UI. So

39:39

we try to think about that. And so we chose to bet

39:42

on adaptive UI, and we chose

39:44

to bet on our existing distribution. And

39:46

that's been working very well for us. The drawback of

39:49

that back to the organization is you

39:51

obviously risk complicating the

39:53

user experience when you're trying to

39:55

do all of these things. So we

39:57

kind of had to design the org literally after.

40:00

the product. I

40:02

usually take the examples of Amazon, you know,

40:04

where they try to build small teams with

40:06

hard APIs that have no dependencies so

40:09

that everyone can run fast in parallel all the way

40:11

to the end consumer. You can

40:13

see that sometimes in the consumer experience, having,

40:15

you know, multiple search books on the same

40:18

page. Whereas in Apple, they've done it differently

40:20

because they're shipping hardware once per year. You

40:23

can't let teams run fast and do their own

40:25

thing. You have to synchronize. So

40:27

we've adopted much more of that approach. We'll

40:29

have a single consumer experience organization

40:32

that owns the application across all

40:34

interfaces, and it's instantiated in a

40:36

single person. And then

40:38

in the same vein, you know,

40:40

we're recommending audiobooks, podcasts and music to a

40:42

single user. You know, for our

40:45

competitors have three different apps, it looks like three

40:47

different users, but it's actually the same user on

40:49

the phone, just switching apps. And we're trying to

40:51

not fool ourselves. We're saying like, no, it has

40:53

to optimize globally. So when we

40:55

recommend something, we have to figure out like

40:57

Tomer right now, you know, should we show

40:59

this Beyonce song or should we maybe show

41:01

this Peter Atia podcast, or maybe this audiobook

41:03

that is into right now. And those are

41:05

different cost implications, but also different

41:07

retention profiles and so forth. So

41:10

we've had to make that a central function

41:12

as well. So we have one that we

41:15

call the PCM personalization organization under single leader

41:17

that tries to balance the user recommendations globally.

41:21

And then behind that, then we have three different

41:23

teams. So we have a music business team, a

41:25

podcast business team, and an audiobooks business. And they

41:27

kind of live as if they're different businesses, but

41:30

they go through this central point. So

41:32

we have very much shaped the organization

41:34

after the product. That

41:36

I think is crucial. Otherwise, our product wouldn't

41:38

work, we'll just collapse into complexity, if everyone

41:41

shipped their own stuff. So we

41:43

kind of said like, okay, we're gonna have

41:45

to eat all that complexity. So we built this

41:47

organization. And it's very hard work for us.

41:49

It's super painful. But hopefully, we don't

41:51

sip the pain to the user. What

41:54

gets me excited is right now we say, yeah,

41:56

you can actually remove that complexity. In fact, if

41:58

you're complex app right now, now, I think

42:01

you have a massive opportunity to reduce it to

42:03

an incredible experience, which

42:05

is simple, just having that

42:07

AI agent playing that for you. That's exactly

42:09

what I think too with AI now. Now

42:11

you can start personalizing the product. And yes,

42:14

we've personalized which songs you get and LinkedIn

42:16

has personalized which posts you see for a

42:18

long time. But the clear promise is

42:20

you could personalize more than that. You know, it could be

42:22

almost only a podcast experience for people only listen to that.

42:25

The UI could be much more dynamic. You

42:27

know, I think one of the most

42:30

interesting analogies I've heard about AI is

42:32

from this economist many years ago named

42:34

R.J. Agrawal. So he says

42:36

like, what happens when the cost of doing something goes

42:39

through the floor, right? If you have

42:41

coffee and the price of coffee goes through the floor, well,

42:44

its substitutes are going to suffer

42:46

like tea, but its compliments

42:48

are going to benefit greatly like sugar and

42:50

cream. And I think that's a really interesting

42:52

way of thinking about it, try to figure

42:54

out what the compliments are. Everyone's talking about

42:56

the substitutes. And that's scary. But

42:59

what are the compliments of like very, very

43:01

cheap computation or personalization or prediction, as he

43:03

calls it? But I think the point on

43:05

this, like the cost of computation going to

43:08

zero, you know, I think what

43:10

he didn't see back then, because the LLMs weren't here,

43:12

he said the cost of prediction going to zero. But

43:14

now that these LLMs can write code, one

43:16

way to think about it is the cost of writing code goes to

43:18

zero. What does that mean? It probably

43:20

means that it makes sense for

43:22

you to write code for things that

43:25

made no sense before. You would only write codes for

43:27

things that you would repeat many times, because

43:29

the fixed costs of getting an engineer to write code

43:31

and test it was very high. And

43:33

this is probably the reason why our UIs look like

43:36

they do, even though we don't think about it is

43:38

because the cost of writing code is so high. But

43:40

what if the cost has now gone to zero? Wouldn't

43:42

it make sense to do not

43:44

just lots of algorithms, but you could also do

43:46

custom UI. I think this is what

43:48

people underestimate all the time. It

43:51

just seems like you could write entire programs

43:53

on the fly for a very small problem

43:55

at almost no cost, including the UI. So

43:57

I would imagine that we see these getting

44:00

very dynamic. Maybe starting with search pages

44:03

where you're asking for something specific and different every

44:05

time, but maybe overall in the product.

44:07

There's a counter side to that. Maybe you don't

44:09

want your app to change UI every time you

44:11

open it because it could be pretty frustrating to

44:14

remember how you get to your library, but there's

44:16

something interesting there. I tried that by the way.

44:18

There's a lot about just muscle memory and knowing

44:20

where to go without doing so, but I tested

44:22

it and it didn't work out

44:24

and failed miserably. But I think

44:26

the hypothesis changes right now because my assumption

44:29

was the app was the thing you came

44:31

for. So the navigation has to be stable

44:33

because you need to know where to go

44:35

without thinking about it. I think that's a

44:37

good insight. If you're navigating, you

44:39

want muscle memory and you want predictability.

44:42

But if someone is presenting to you, you

44:44

should tailor the message to what you're presenting and

44:47

it's easy to get over your skis. There's lots

44:49

of overpromise and under delivery happening as well right

44:51

now. What

44:53

a great interview. I'm excited to jump into

44:55

my key takeaway with you as always. There

44:58

were so many in this conversation, so I'll

45:00

share my top ones. First, nature

45:02

versus nurture. Gustav's tendency,

45:05

his desire has always been to

45:07

go deep. It's a

45:09

great quality, but it's just not what

45:11

fate had in mind for him when

45:13

he entered the job market as a

45:15

young professional during a bust. Circumstances required

45:18

that he take on an entrepreneurial role,

45:20

making his own opportunities. And

45:22

that meant teaching himself a broad range

45:24

of skills on the go. That willingness

45:26

to explore helped him along his career

45:28

and he's been able to pair that

45:30

with his natural desire to go deep.

45:33

That combination made him

45:35

a remarkable builder who is able to build

45:37

differentiated and unique products. Second,

45:39

similarly, Gustav, wearer of many

45:42

hats, resists feeding himself into

45:44

labels. Just because you're

45:46

an engineer doesn't mean you can't

45:48

extend yourself and innovate with business

45:50

models, licensing deals, or go-to-market efforts.

45:53

This highly resonates with me,

45:55

allowing yourself to be labeled limits your

45:57

thinking and prevent you from growing. Regardless

46:00

of how your job is defined,

46:02

venturing into adjacent domains, from engineering

46:04

to product to marketing, will make

46:06

you a superb builder. Third,

46:09

in a world where mobility between jobs,

46:11

especially in tech, happens at a frantic

46:13

pace, it's helpful to

46:15

remember that it's hard to play

46:17

the long game without building depth

46:19

and expertise. Whether it's because

46:22

of cultural, geographical, or other reasons, Spotify

46:24

has been able to take advantage of

46:26

its employees' long tenure. That's

46:28

helped them see their bets through, with

46:30

patience and stability. Especially

46:33

if a three-year bet turns out to actually

46:35

take five years to realize. Fourth,

46:38

move fast and break things has been a philosophy

46:40

that has been adopted in tech for quite a

46:42

while. The goal was

46:45

meant to encourage experimentation and speed, even

46:47

at the expense of damaging the experience in the short

46:50

term. With all of

46:52

the contractual and licensing obligations Spotify is

46:54

locked into, Gustav knows that

46:56

the consequences of a misstep could be harsh.

46:59

Breaking things could mean breaking the company.

47:01

He also notes that talk is

47:03

cheap, and that's why Spotify has

47:05

fostered a culture of debate and

47:07

deliberation, gaming out consequential decisions. It

47:09

might seem slower at first, but

47:11

done well, the final outcome would

47:14

probably be better, and maybe even

47:16

faster. Fifth, somewhat

47:18

related. Gustav is a big

47:20

believer in the value of explanations. He'll

47:22

even err on the side of over-expanding himself

47:24

because it forces him to lay out his

47:27

rationale on the table. That

47:29

way, even if your co-workers disagree with your decision,

47:32

they know what it's based on and can

47:34

either inject a useful counter-argument or at least

47:36

get aligned with you. After all,

47:38

as Gustav says, everyone deserves an explanation, and

47:40

you as the leader have to explain yourself.

47:42

For me, this goes back to principled thinking,

47:45

but it's not enough to base your thinking

47:47

on principles. You need to share those principles

47:49

with your team. Sixth,

47:52

these explanations are not just instructive

47:54

and helpful for debate, they're also

47:56

helpful to create broader understanding across

47:58

the company. out

48:00

all the dimensions of an explanation, write

48:03

it down and teach it to your colleagues. It

48:05

will spread easily throughout the organization and

48:08

people can apply your reasoning to current

48:10

and future work easily. Lastly,

48:13

depending on the context in which your

48:15

product operates, you may need to lean

48:17

in to distribution advantage over creating a

48:19

design advantage. Let me explain that. When

48:22

Spotify got into podcasts, there was a strong

48:24

argument to be made that they could have

48:26

made a different app with a simplified user

48:29

interface optimized for a podcast experience, since

48:31

most people do not listen to music and podcasts in

48:33

the same way. But Spotify

48:36

realized it wasn't competing with the

48:38

quality of different podcast players. It

48:40

was actually competing with Apple's own podcast

48:43

app, which had a distribution advantage of

48:45

coming pre-installed on over 2 billion devices.

48:47

So by creating a separate app, Spotify

48:50

would have had to reignite again the

48:52

distribution reach it already built for its

48:54

main music app. Do

48:56

you have an idea of how that might influence your

48:58

thinking as a product builder? Let me know

49:01

on LinkedIn. I'm Tomer Cohen. Thank

49:03

you for listening. I learned a lot from this conversation and

49:05

I hope you did as well. If

49:09

you liked this episode, don't forget to

49:11

rate and review us on Apple Podcasts.

49:13

It'll help people discover the show. We're

49:16

hard at work on bringing you Season 2, but

49:18

if you can't wait till then, then tune

49:20

in next week, where we'll have a bit

49:23

more wisdom from Gustav as we conclude Season

49:26

I try to understand competitors and what they're good at.

49:29

Sometimes to borrow from them, I think you should borrow with

49:31

pride. Building One is a

49:34

LinkedIn editorial production. Our host is

49:36

Tomer Cohen, LinkedIn's chief product officer.

49:39

This episode was produced by Max Miller.

49:41

Our associate producers are Lolia Briggs and

49:43

Rachel Karp. This episode was mixed by

49:45

John Partham and engineered by Asaf Gedron.

49:48

At LinkedIn, Sarah Storm is senior producer

49:50

and Enrique Montalvo is our executive producer.

49:52

Dave Pond is head of news production.

49:55

Courtney Koop is head of original programming.

49:57

And Dan Roth is the editor-in-chief of

49:59

LinkedIn. Thanks to Alicia

50:01

Mann, Haley Saltzman, Mary Wilson, Sarah

50:03

Scully, Ayanna Deldridge, Cayman Rojas, Michaela

50:06

Greer, Kyle Ranton Walsh, and Maya

50:08

Pope-Chappelle. If you know of a

50:10

product leader we can all learn

50:12

from, send us a line at

50:14

pitch at linkedin.com