Lenny's Podcast Sep 20, 2026 1h 37m 1h 5m saved
With Peter Sellis, former Head of Product at Discord and the first product manager at Snapchat
Snap was quicker to $4 billion in advertising revenue than Twitter, Reddit or Pinterest, and started after all three. Peter Sellis, who ran product there, says three structural facts capped it anyway.
The standard explanation for Snap's monetization problem is that teenagers are hard to sell against. Sellis accepts that and adds two more: the app opens the camera rather than a feed, so there is nowhere in the core flow to put an ad, and underneath everything it is a messaging app, which nobody has yet turned into a large advertising business.
"But Snap was trying to build a multi-billion dollar ads business very quickly on a lower value audience, a very tough format for ads and a service that was more utilitarian than consumptionoriented."
Peter Sellis, former Head of Product at Discord, on Lenny's Podcast, spent seven years as the first product manager at Snapchat and then ran product at Discord through the growth push of 2024, and he did this interview between jobs, with no communications team to clear it.
The full interview is covered here so you can skip it. 97 minutes of audio, 32 minutes of reading.
Here are the 14 lessons that matter.
Key Takeaways
Collaboration has a coordination cost nobody says out loud, and it sets the team's speed to that of its slowest member
His test: you never hear an executive say stop collaborating, which he treats as a warning sign
The median product manager is bad for a mathematical reason, not a cultural one: the good ones leave for founder and executive jobs and the weak ones fight to stay
Three things capped Snap's advertising business — an audience below the age where ads can be measured, a product that opens the camera instead of a feed, and a messaging app underneath
Advertising rates step down sharply below 18, because the rules on tracking and targeting minors make the audience hard to measure
A product leader's job is to make the founder the best-curated museum in the world, and the interview question that reveals taste is what someone built and never shipped
OpenAI is speedrunning a solved problem in ads, and the open question is how much trust it trades for revenue in the auction
Growth came from the core, twice: Android performance after Snap's 2018 redesign, and gamers at Discord in 2024, a group everyone assumed was already fully penetrated
Midjourney was invisible in Discord's usage data, even at the height of the attention it drew
SpaceX is sunsetting the Falcon 9 before anything has beaten it, which he calls the clearest defense against the innovator's dilemma he has seen
In advertising, price is an output, not an input — so the way to grow revenue is to drive advertiser return today
Every great PM holds three contradictions: confidence with humility, process with tolerance for ambiguity, long-term thinking with an urgent bias for action
1. Collaboration Has A Cost
Sellis opened on how he designs teams, using a description he said he should probably workshop: like a terrorist organization. He said Discord's chief people officer was simultaneously horrified and intrigued when he explained it.
The first of his two beliefs is about what working together actually costs.
The coordination cost sets the team's speed to its slowest member
Collaboration comes with incredible coordination costs. And coordination costs essentially assure that you're moving as slow as the slowest node in the system.
Peter Sellis
His evidence that the cost goes unpriced is that nobody in authority ever argues against it. He said you will never hear an executive at any company say stop collaborating, and he treats that silence as a warning sign.
The second belief is about where decisions sit. At a fast-growing company, he said, product managers should be making decisions the founders would have made six months to two years ago.
Ideology plus clear decision rights, and you can stop over-collaborating
One is it has this like ideological culture that everyone understands why they are doing what they're doing. And then it has a very very clear organizational structure of who can be trusted with which decision.
Peter Sellis
Put those together, he said, and the team reaches the result it would have reached by collaborating, without paying for the collaboration.
His own role in that structure was translation. At both Snap and Discord he worked with founders who were still in charge, and his job was to take their vision and compress it into repeatable phrases.
The head of product is a vessel for the founder's vision
The head of product is really kind of like a vessel for their vision.
Peter Sellis
2. Managing Nikita Beer
Discord acquired Nikita Beer's startup, and Beer reported up to Sellis for about a year, with what Sellis called some phenomenal engineers alongside him. Asked what managing him was like, Sellis rejected the verb.
You do not manage him, you direct him
I think much like yeah you don't manage Nikita you only like seek to control him or like direct him.
Peter Sellis
His analogy is the Taken films, and he took it to the unflattering end of the joke.
One very particular set of skills
I think of him a lot like Liam Neeson in like the Taken movie series.
Peter Sellis
But then when you like need him to like raise the daughter or like buy some groceries or stuff like that, all he knows how to do is just like kill everybody in the grocery store or something like that.
Peter Sellis
He said he once had to message Beer on four different apps to get a one-on-one. At one point he threatened to book him through Intro, the paid one-on-one booking service Beer was using at the time, before Beer finally replied.
The general point is about whether an organization can hold people like that at all.
A product org that cannot absorb spiky talent is a failed org
I wouldn't I view it as like a failure state if you have kind of like product orgs that can't essentially handle really spiky talent.
Peter Sellis
The management model he named for it is Phil Jackson, who won with the Bulls and the Lakers and as a player before that, meshing people who did not otherwise get along.
Jackson is his leadership benchmark
I think he's like one of the most meaningful leaders from a leadership styles perspective of the last half century
Peter Sellis
And his frame for a spiky product manager having a bad year is a band with a bad record.
Product leaders are bands, and bands put out duds
I think of a lot of product leaders more like bands putting out albums and like you can like love a band but sometimes they have a dud album
Peter Sellis
man, sometimes they whiff on an album, but if I still believe in them as an artist, like I'll stick with them.
Peter Sellis
3. Most PMs Are Net Negative
The host put to him a belief he holds: that most product managers are bad and often a net negative. Sellis said the reason is arithmetic rather than character, and that he was talking about himself and his peers as much as anyone.
Cut the tail off a normal distribution and the median falls below the average
if product management is normally the skill of product management is normally distributed across a population, the ones who are actually product managers are all the way on the right hand side of the distribution. So you lop that off and it becomes a power distribution. And for any for any power distribution, the median is going to be far below the average.
Peter Sellis
Two forces then push the distribution further left. The first is pay. During the era of near-zero interest rates the job became, on his account, the best-paid role in technology for someone without an engineering background, and the definition of the job is subjective enough that anyone inside it has every reason to stay.
It became the most lucrative job in tech for non-technical people
especially during the ZERP era, it became essentially the most lucrative career in tech for nontechnical people.
Peter Sellis
The second is exit. The skills that make someone good at the job are the skills for running a company.
The good ones leave and the weak ones dig in
the great product managers are exiting the system. They're becoming founders, they're becoming executives etc. And on the other side of the distribution anybody who's in is doing anything possible to stay in.
Peter Sellis
He said he has been building a system for evaluating product managers that scores the person and the place together, to check the role is net positive at all. His evidence that the complaints come from the middle of the engineering distribution rather than the top:
Strong engineers get what they need from a PM and refuse to work with a bad one
I really don't hear great engineers complain about product managers that much because great engineers a they know how to bend product managers to their will.
Peter Sellis
He added that very weak engineers do not complain either, because a product manager is somewhere to hide. The complaints come from in between.
4. Why Snap's Ads Stalled
Asked whether Snap's limit was simply that its users are hard to monetize, Sellis said he would try to be objective about what he called a bit of his life's work, then gave three reasons, all structural.
The first is age. He said there is a step change in advertising rates between 17 and 18, and again between 18 and 21, in the United States and abroad, because of what can legally be measured.
Below 18, the ads cannot be measured, so the audience is worth less
If you're under 17, the rules around how you can track and measure and target ads are extremely constraining and justifiably so.
Peter Sellis
And so, yeah, that Snap's audience is definitively like less valuable, from that perspective versus like a Pinterest or an Instagram, or a Twitter.
Peter Sellis
The second is the opening screen. Every non-search advertising product, he said, opens onto the surface where the ads go, and Snapchat opens onto the camera.
Every display ads business opens to the screen the ads sit on, and Snap opens the camera
The second thing is almost every ads business you'll ever every ad product you'll see that isn't search based, every display ad product opens to the screen where you put the ads. There's a lot of studies around attention and just mathematically where you're going to get the impressions where you need to kind of like maximize that for advertising and Snap opens the camera.
Peter Sellis
There is no slot in the core flow
in no part of that flow, the core flow of taking a snap and sending it to your friends is there like a logical place for an advertisement.
Peter Sellis
He named camera monetization as his own biggest failure at the company, and said either he could not solve it or he could not green-light the teams that might have.
The third is what Snapchat is underneath.
Nobody has built a good ads product for a messaging app
And then at its core, it's a messaging app. And like there aren't really great ad products for messaging apps right now.
Peter Sellis
He said the Asian chat apps people cite as counter-examples do not hold up on revenue per hour of time spent. Messaging is an excellent utility to own and a poor advertising surface.
All three at once
But Snap was trying to build a multi-billion dollar ads business very quickly on a lower value audience, a very tough format for ads and a service that was more utilitarian than consumptionoriented.
Peter Sellis
Against that, he wanted the record straight on what the company did achieve. Between them the two of them listed what Snapchat is to teenagers in almost every western country: the top messaging app at 70 to 100% penetration under 21, the default map ahead of Apple's and Google's, the most popular astrology app, the most common camera.
A Berkshire Hathaway of Gen Z apps
There's a there's a Gen Zers Berkshire Hathaway that has the number one camera app, the number one messaging app, the number one maps app, the number one astrology app
Peter Sellis
And it got to $4B in ads faster than the companies that started before it
Snap started after Twitter, after Reddit, after Pinterest. It went public before all of them except for Twitter. It was quicker to four billion I think in ads revenue than all of them despite starting later.
Peter Sellis
His point about the comparison is that those companies had easy advertising surfaces. Twitter is where the most valuable people on the planet talk; Pinterest is a place where users state what they intend to buy.
On the subscription side, Sellis said Snapchat Plus has been executed well, and that his regret is timing. Subscription products on a utility build slowly, because a user has to become dependent on the product before the upgrade makes sense.
Four years earlier and it would have taken the pressure off ads
if Snap had this product out four years ago, there might have been less pressure on the ads business to scale as quickly as it does.
Peter Sellis
He also described how literal Instagram's copying was in the early years.
Instagram copied the features that were not working either
there was early on I remember a snap it was like so funny to see how aggressively Instagram was copying that they were actually copying like some of our worst features like ones that weren't working just because I think they just like assumed everything that we were touching was turning to gold
Peter Sellis
He gave the other side of it too: on Stories, the copy is now the better product.
Instagram's version of Stories is now better than the original
when it comes to stories which was obviously extremely effective from a copying perspective you'd be hard-pressed to say anything but the version of it on Instagram is now much better.
Peter Sellis
5. Saying No Is The Skill
Evan Spiegel's own answer to being copied for fifteen years is that software is no longer a moat, which is why he is spending his time on hardware. Asked about it, Sellis said he thinks Spiegel has been right about almost everything, usually early, and that investors have not given him room to be right.
Then he named what he learned from watching him decide.
He watched more good ideas get killed at Snap than most startups ever launch
I think I saw Evan personally say no to better ideas than I've seen almost any other consumer startup launch.
Peter Sellis
That became his own model of what a product leader does for a founder.
Make the founder the richest museum curator in the world
my goal as like a product leader for founders like at these high growth companies was always to like turn them into almost like the best museum, the richest museum curators in the world where like whatever is in the product at any one time is just a small slice of the overall collection.
Peter Sellis
The curator who hangs everything has not curated
If a museum has all of their collection on the walls, then the curator hasn't done anything.
Peter Sellis
He turned that into a hiring question, and it is a specific one: not what someone tested and killed on the data, but what they held in their hands, finished, and decided the world was not ready for.
Ask a candidate what they built and never shipped
one way you can kind of like get at somebody's taste is when you're interviewing them is just ask them about some of the things they built that they never shipped.
Peter Sellis
Restraint is one of the few taste muscles you can actually train
exercising that kind of muscle of like saying no of restraint is probably one of the easiest like one of the few like muscles of taste you can exercise
Peter Sellis
The host connected that to AI, pointing out that AI is very bad at removing things from a design and always adds. Sellis agreed it will be a hard skill to replicate, and said the broader effect of lowering the threshold for creation is that more of the work will have to earn human attention.
6. Ads Are The Internet's Fit
Asked whether ads are a real business for ChatGPT and the other AI products, Sellis put himself on one side of a long-running argument, naming the mobile-advertising analyst Eric Seufert as the position he shares: that advertising, not subscription, is the internet's product-market fit.
Subscription products are overrated against ads, for reasons he says are mathematical
And I still think that from a monetization standpoint there's a number of reasons again mathematical for why that is true and why I think like subscription products and things like that are kind of overrated versus ads from a consumer side of things.
Peter Sellis
His read on OpenAI is that it is rebuilding, quickly, something that has already been figured out, and that Meta's team between 2010 and 2020 deserves most of the credit for figuring it out.
Last mover advantage, and OpenAI is using it
When I look at what OpenAI is doing from an advertising perspective, they are they are speedrunning all of the like ads infra that it is exactly what you would want and do at Discord. I call this like last mover advantage.
Peter Sellis
The modern ad auction is a solved problem
what drives a modern ads marketplace and auction is essentially like a solved problem and OpenAI is speedrunning all of the components of that.
Peter Sellis
The innovation he credits Meta with is not that the highest bid wins. It is that the value of the ad to the user is priced into the bid, so reaching someone who does not want to be reached costs the advertiser more.
The user's own interest is an input to the auction
not just that the highest bidder wins but that the organic value of an advertisement and an advertiser to that user should be factored into their bid such that you essentially have to pay more as an advertiser to reach people who may not want to be reached
Peter Sellis
That also lets organic content and paid content compete in one auction with money as the common unit. Carried into a chat product, he said, the thing being traded off is not how much someone enjoys the ad. It is trust.
In chat, the variable is trust
in chat it probably feels like this could be replaced or augmented with essentially trust.
Peter Sellis
His example of the trade going wrong is Amazon search, which he was careful to say is not threatening a product he rates highly.
Sixteen ads before the first result
it is getting a little crazy when you search for something and you have 16 ads before you get to the first thing
Peter Sellis
The whole question is where they set that dial
where the push is going to come to shove is how they instantiate it mathematically in their auction mechanism to maximize trust over the long term.
Peter Sellis
He expects trust to correlate with retention, and possibly to cause it, and with what advertisers will pay. Two things are still open: the format an ad takes inside a chat, which may not even be visual if the interface is voice, and whether removing ads from the top subscription tier was right. He said he was disappointed to see the most valuable users taken out of the pool that early, while understanding why companies do it. He called ad formats at OpenAI one of the most interesting jobs in product and design right now, and a societal-level job in advertising.
7. Ride Your Best People
The first of four counterintuitive positions he had prepared: he spends his time working his strongest people to the limit rather than raising the weakest.
It runs against instinct, deliberately
if you see somebody drowning, you like you want to save them. And, like I did the opposite, absolutely opposite.
Peter Sellis
More rope until they use it
if somebody is strong and crushing it, I'm just like I'm just like handing them more and more rope to hang themselves with. And I'm pushing them as hard as possible. Like to the to the point of breaking.
Peter Sellis
The mechanism is decision volume. Someone making good calls gets more calls until the quality drops.
Load them up until the decisions get worse
people who are making consistently good decisions I just want to load them up with more decisions till I start to see them make bad ones.
Peter Sellis
He paired it with a rule about public defense: he never criticizes his own people in front of others. And he was candid about the cost of the approach, which he framed as his own limitation.
He has no good method for improving people, so he lets them fail
I just don't have a really good mechanism for helping people improve. So, I essentially like let them fail on their own. Which could be I think probably pretty frustrating for me as a leader to work for.
Peter Sellis
The thing the model can get wrong
I think the thing you need to control for when you adopt this type of management philosophy is that not paying attention to somebody doesn't mean they're failing.
Peter Sellis
8. Growth Is In The Core
His third counterintuitive position is that growth comes from the core of the business, and he said he learned it twice, at Snap and at Discord.
Network businesses want to grow on their own
the nature of like network effect oriented businesses is that they want to grow implicitly.
Peter Sellis
The arithmetic he uses is deliberately simple. A monthly user is worth a thirtieth of a daily one, so the cheapest growth is moving the ratio of daily to monthly users rather than buying more monthly users and hoping they stay.
Thirty monthly users equal one daily user
if you're thinking about like a daily active user, if you're going for somebody who uses the product once a month, you're going to need 30 of those to equal somebody who uses it daily.
Peter Sellis
So move the ratio, do not buy the top of the funnel
a lot of growth can come from essentially moving that ratio just a few basis points versus trying to go acquire a bunch of new monthly active users and seeing if they if they retain.
Peter Sellis
He also made the point that the daily-to-monthly ratio is simply next-day retention, and that days are the only unit that maps onto how people actually live. Weeks are softer and months are mostly a business-to-business rhythm.
The Snap case study is the 2018 redesign.
The only quarter Snap's daily users went flat
at Snap, the lesson there came from that disastrous redesign in 2018, which was the only time I think that Snap's overall DAU actually flattened out quarter over quarter.
Peter Sellis
What fixed it was Android speed
what we like laser focused on after that was performance, particularly on Android.
Peter Sellis
That, he said, produced a growth renaissance that ran for years. The Discord case is more counterintuitive, because at the time everyone outside the company thought the story was Midjourney, which was then being built on Discord.
Midjourney does not show up in the usage chart
if I were to show you like a you can even look at this online, right? like Discord's use over time, you wouldn't be able to pick out midjourney.
Peter Sellis
So the team pointed itself at the users it already had
what we did in 2024 was completely focus the team around this idea of how can we be the best place for you and your friends to play intentional multiplayer games together.
Peter Sellis
Everyone assumed that group was already at full penetration. It was not: people were choosing in-game voice, or a FaceTime call, or nothing.
Moving a user from 10 days a month to 13 is the whole game
we were actually able to like identify a lot of things that brought somebody from using Discord 20 or let's say 10 days out of the 30 to 11 to 12 to 13.
Peter Sellis
And it produced the fastest growth since the pandemic
As we made Discord better, as we focused on performance, we were actually able to drive some of the fastest, I think, if not the fastest growth.
Peter Sellis
The side benefit was reputational, and he was blunt about which audience produces it.
Gamers are the most demanding users there are
the most vocal critical honestly like intelligent or like high IQ user base in the world are gamers
Peter Sellis
Telling that audience to post a bug thread on Reddit and promising to fix everything on it, however small, is what he says made Discord unrivaled in its niche.
9. When To Leave The Core
Asked when it does make sense to bet outside the core, Sellis gave three methods, in increasing order of difficulty.
The first is a grid: take what your product is genuinely great at, the camera at Snap or voice at Discord, and look for other groups who need that. He called it an easy exercise that should always be running.
The second is watching what users force the product to do.
Users going through hoops are telling you what to build
one of the greatest gifts is when people use your product in a way that you didn't expect or like didn't like didn't ordain because they're going through extra hoops to do what they want to do with the product.
Peter Sellis
They are voting with their feet
they're essentially voting with their feet.
Peter Sellis
At Discord that was people organizing communities far larger than the twenty-person gaming friend group the product was built for.
The third is the hard one, and it is the Snapchat method: start from why the company exists, imagine the end state with no reference to today's constraints, and work backwards.
Backcasting, not forecasting
more like future casting like backcasting instead of forecasting
Peter Sellis
His worked example is capture itself. Pointing a phone at a moment is a constraint of current hardware; ask what the best way to share a moment would be with no such constraint and you arrive at glasses.
Almost nobody can actually do it
In my experience there are very few people that have this type of like blank sheet creativity to think about the future totally unencumbered from the present.
Peter Sellis
Most people, including creative ones, remix the present, and he was rude about the standard route: the trip to Japan, the retreat, the return with the answer. The deeper obstacle is that finding product-market fit once makes a team think it can do it again.
Finding the first thing that works was luck, and teams forget that
you found product market fit, something's working. You almost take for granted how lucky you are that you found something that works.
Peter Sellis
The host offered Airbnb, where he worked for years, as the case in point: nothing beyond the original idea has moved the growth trajectory, and what has worked is making the core experience better with more supply and more demand.
Sellis named the risk in staying at the core, using Nassim Taleb's turkey: everything improves until something ten times better arrives and takes the business to zero overnight.
10. SpaceX Burns The Ships
His answer for how a company guards against that is SpaceX, where his wife is an engineer. He was explicit that he has no inside information.
His rating of the company
SpaceX is the greatest American company right now. Greatest company in the world.
Peter Sellis
The example he gave is the decision to retire the Falcon 9, which he described as the most cost-effective reusable launch system in the world and one that any nation on the planet would trade for, in favor of Starship.
Retiring the best thing you have, before anything beats it
they're sunsetting the Falcon 9 and this is like Cortez level burn the ships upon arriving
Peter Sellis
This is the innovator's dilemma solved rather than described
just because we have the thing that is the best, we're completely willing to build the thing that will compete against ourselves and make the thing that we built irrelevant.
Peter Sellis
He said Clayton Christensen, who wrote the book on the problem, would have wept over it. What he could not explain is what produces the behavior beyond Elon Musk himself and the culture around him, though he noted one visible mechanism: both SpaceX and Tesla published their long-range plans years in advance.
The roadmap is public, a decade out
you look and like on the website, now I don't know where it is, but like they publish essentially like 10-year plans. Like nothing SpaceX is doing is not something that they didn't talk about 10 years ago.
Peter Sellis
He said Tesla described the roadster funding the S funding the 3 exactly as it happened, and then released the patents as well, and that he aspires to that level of confidence as a product leader.
The host connected it to a theme running through the show: that AI makes building easy, so ambition becomes the scarce input. Sellis put it in physics terms.
He does not want the position or the velocity
I don't even want the speed. I don't want the velocity. I want the acceleration.
Peter Sellis
Getting acceleration takes an impulse, he said, and that is not natural for people, because the world is full of systems designed to keep tomorrow looking like today. The host relayed a point from a previous guest, Tara Seshan, who has worked on product at ChatGPT: the job of product management now is increasingly to raise the team's ambition and ask what a thousand times the current idea would look like. Sellis agreed, with a caveat about the cheap version of it, the senior person who arrives in a meeting and asks whether it could be done faster.
11. Long-Term Greedy
Asked for the phrases his teams ended up repeating, Sellis named one he used at Snap, Discord and in advisory work: in advertising, be long-term greedy.
The problem it solves is that an advertising business is revenue-driven, hand-to-hand and relentlessly short term. The temptation it blocks is the cheap way to improve an advertiser's return, which is to charge them less.
Price is an output of the auction, not a lever you pull
The issue is that the easiest way to drive ROI is actually to decrease the denominator the I so you can you could lower prices. right now in modern auctions price is an output not an input.
Peter Sellis
So the way to grow the business is to deliver return for advertisers today and collect the budget increases later, even where that costs revenue in the current quarter. The value of having the phrase, he said, is that it gives cover for long-term thinking at exactly the moments when long-term thinking is hardest: a public company missing numbers, with a layoff on the table if things do not turn.
And it is not charity
we're not a charity like we're not giving this stuff away because we want to make people happy. We're giving this we're giving this away. We're making this happen today because we know they'll come back tomorrow even bigger.
Peter Sellis
12. Core Product Value
The second phrase is a framework both Snap and Discord ran on, and it has two halves that have to exist together: a sentence anyone can repeat, and a set of metric definitions underneath it.
Define, conceptually and numerically, why someone comes back daily
you need to define both from a theoretical or like a conceptual level as well as a metrics level like what it means for somebody to come back to your product every single day.
Peter Sellis
Snap's sentence
Snap is the fastest way to share a moment with the people you care about.
Peter Sellis
Each part is decomposable: fastest becomes time to load, share becomes opening into the camera, people you care about becomes best friends. Then a roadmap item either serves that sentence or it does not.
Discord's sentence
I think Discord was like the best way to talk and hang out with your friends before, during, and after playing games.
Peter Sellis
He said the metric definitions behind both were internal and proprietary, and that they ran through the growth books and the experiment write-ups, which is what got new data scientists productive quickly. Neither half works alone: metrics without the sentence mean nothing to people, and the sentence without metrics never gets acted on day to day.
He gave the same onboarding talk every new hire for seven years
for core product value at Snap we actually had an onboarding presentation which I delivered to every single new hire for seven years straight every week or every other week later that was about this
Peter Sellis
Because it is the answer to the competitive question
This is why people choose Snapchat every single day in the face of tons of alternatives and a bunch of savage competitors.
Peter Sellis
He added that the hard part is the causal work: finding metrics that are actually demonstrated to drive retention, which he said makes him unpopular with data scientists.
13. Three Pairs Of Wolves
Sellis said a great product manager is an oxymoron, and named three contradictions each one has to hold.
The first is confidence against humility.
Confidence that borders on arrogance
to be a really good really good PM you kind of have to be like really confident and that borders on arrogance a lot of the time.
Peter Sellis
The humility half is credit: naming the designer who made the last-minute change and the engineer who stayed up for the partner launch. His explanation for why the two go together is unusual.
You can give the credit away because you know you will get too much of it anyway
having the humility and the attention to detail to know who to give credit to actually comes from this sense of confidence that like you kind of know that PMs are going to get too much credit no matter what.
Peter Sellis
The second is process against ambiguity, and he said this is the one he tries hardest to test for in interviews.
A checklist for everything, and no problem throwing it out
you want these people who are like extremely organized and very like process driven like they understand the value of good process and not too much not too little etc. and then are just like super comfortable with ambiguity almost to the point of like despising authority
Peter Sellis
He wants the candidate to demonstrate both through a story. The fashionable position, that process is useless, only carries one wolf.
The third is a long horizon against short-term urgency.
They know what the world looks like in two years, and which blocks come first
they just know how they want the world to look 6 months, a year, two years from now. And they're patient and they know what building blocks need to come into place for the system to work.
Peter Sellis
And they set the pace of the whole company
I think that combination of like urgency and bias for action with long-term thinking yeah those are my three my three oxymorons the three pairs of wolves inside every great PM
Peter Sellis
14. Contrarian Corner
Asked what he believes that most people do not, Sellis sorted his answers by heat.
The mild one is about who you live with.
The biggest career decision is who your partner is
I think the most important career decision you make is who your partner is, like who you marry, who you spend your time with personally.
Peter Sellis
No amount of podcast listening, he said, substitutes for a partner who supports you and acts as an intellectual sounding board.
The middle one extends the curator argument to collecting and to taste generally.
People with taste stop one step before everyone else
the ones with taste like kind of like stop one step before everybody else does.
Peter Sellis
The skill is knowing when a thing turns
people with great taste seem just like know when to take something and get passionate about something and then kind of like when it gets cringe.
Peter Sellis
The third he said he would not be comfortable giving with an employer behind him.
Some traits that matter for consumer product cannot be learned
There are certain inalienable traits about those product managers that cannot be learned cannot be acquired that are that are essential to creating truly great consumer products. And by that I mean literally gender, race, ethnicity, socioeconomic background.
Peter Sellis
He was specific that he does not mean matching a product's audience to its team's demographics. His argument is that consumer products work against basic human needs, and that upbringing and background affect how well someone understands a given set of them.
And he says it is mostly not actionable
That doesn't mean you can hire for it. That doesn't mean you can judge a book by its cover, things like that.
Peter Sellis
That, he said, is why it is a spicy take rather than a useful one: he can state it and not tell anyone what to do with it.
Bonus Insights
Why he had never done a podcast before
His first reason was unflattering about the format: early on, when he saw someone go on a well-distributed product podcast, he assumed they were about to get fired and were marketing themselves for the next job. His second was that how he talks at work is better suited to closed rooms. Being between jobs removed the constraint.
So I'm no longer beholden to any corporate comms team right now. So yeah, I can I guess I can I can let it rip a little bit.
Peter Sellis
What systems thinking actually requires
Sellis said the underrated part of systems thinking is being able to write the system down mathematically, in stocks, flows and the distribution behind each variable, rather than reading the popular book about it. He pointed at Jay Forrester's system dynamics work from the 1970s and at modelling it in a spreadsheet, which he said is clarifying even if the spreadsheet goes nowhere. His view of how most people use the term is less generous: asking why five times and calling the answer a first principle.
His three book recommendations
I'm going to butcher the title a bit. How to get filthy rich in rising Asia and when genius failed.
Peter Sellis
And then my third one which I actually give to a lot of consumer PMs is Einstein's Dreams by Alan Lightman the physicist from MIT.
Peter Sellis
The AI product that changed his workflow
The one that I found that changed my workflow the most in at work was Hex.
Peter Sellis
He said it was one of the only multiplayer AI experiences that worked, and described sitting in a meeting with Discord co-founder Stanislav Vishnevskiy inside a Hex thread, querying data as the discussion went on.
The quote he lives by
this quote from Steinbeck it's about like how time passes differently when you experience more or less. And essentially the quote is that like duration has no signposts sorry time has no signposts upon which to drape duration. From nothing to nothing is no time at all.
Peter Sellis
the best way to slow the passage of time is essentially to do something with it.
Peter Sellis
His first product job was a livestream of puppies
Sellis ran product at Ustream from around 2009, alongside Livestream and Justin.tv in what he called a golden age of live video opened up by better bandwidth. Ustream's breakout was a Logitech camera pointed at a litter of Shiba Inu puppies.
Ustream was essentially a failure because we should have focused on gaming. Justin TV focused on gaming became Twitch etc.
Peter Sellis
He said the site's other always-on channels at the time were an eagle's nest in Decorah, Iowa, and Jason Calacanis broadcasting his startup show, and that Ustream's first product adviser was Josh Elman and its first product manager was Matt Schlicht, who dropped out of high school to take the job and later sold a social network for AI agents to Meta.
Live streaming is still kind of like an uncracked product category.
Peter Sellis
The game he recommends instead of a TV show
you should give last year's game of the year, Clair Obscur: Expedition 33, a shot.
Peter Sellis
Sellis's bottom line is that the growth is almost always in the product you already have and the users you already have, that the scarce skill in product is refusal rather than invention, and that Snap's advertising problem was never a failure of execution but three structural facts about the audience, the opening screen and the format.
Products, Companies & Tools Mentioned
Snapchat and Snapchat+ (Where he was the first product manager for seven years. He names three structural caps on its advertising business, and says the subscription product has been executed well but launched about four years later than it should have)
Discord (Where he was head of product. The 2024 decision to re-aim the whole team at multiplayer gamers produced what he says was the fastest growth since the pandemic)
Midjourney (Built on Discord and impossible to find in Discord's usage data, which is his evidence that the attention a partner draws is not the same as growth)
Instagram (Copied Snapchat's features, including the ones that were not working; he says its version of Stories is now the better product)
OpenAI and ChatGPT (Speedrunning the ads infrastructure Meta built between 2010 and 2020; he says the open question is how much trust its auction trades for revenue)
Meta (Whose team he credits with the modern ad auction, pricing an ad's organic value to the user into the bid, and with the growth playbook he borrowed at Discord)
Amazon (His example of the trust trade going too far: 16 ads before the first search result)
SpaceX and Tesla (His answer for how a company avoids being taken to zero: retiring the Falcon 9 for Starship, and publishing ten-year plans in advance)
Pinterest, Reddit and X (The comparison set for Snap's advertising business: all started earlier, all with easier advertising surfaces)
Hex (His pick for the AI product that changed how he works, and one of the only multiplayer AI experiences he says works)
Ustream (His first product job. He calls it a failure because it did not focus on gaming, which is what turned Justin.tv into Twitch)
Twitch (What Justin.tv became by making the choice Ustream did not)
Roblox (Where he says the next cohort of Snapchat users is graduating from)
Airbnb (Raised by the host as the case for growing the core: nothing beyond the original idea has moved the trajectory)
Books & Resources Mentioned
How to Get Filthy Rich in Rising Asia – Mohsin Hamid (One of the two books he recommends most)
When Genius Failed – Roger Lowenstein (On the rise and fall of Long-Term Capital Management in 1998, which he says caused or nearly caused a crash)
Einstein's Dreams – Alan Lightman (The book he gives to consumer product managers)
East of Eden – John Steinbeck (Source of the line about time he uses as a motto)
Jay Forrester's system dynamics (The 1970s manufacturing-era work he would rather people read than the popular systems-thinking book)
Clair Obscur: Expedition 33 (His recommendation in place of a film or TV pick, from a French indie studio)
Listen to the full episode
🔴 YouTube | 🔗 Episode page
Watch the full episode:
More Peter Sellis interviews and podcast appearances
If this was worth your time, send it to someone closer to the industry than you are.
Get the latest market chatter as it happens: