---
url: "https://www.youtube.com/watch?v=ruvis-VWg2s"
title: This CPO regrets that product management exists | Tom Verrilli (CPO of Whatnot)
source_kind: youtube
author: "Lenny's Podcast"
captured: "2026-08-06T04:08:23+00:00"
comment_tree: false
topics: [ai-impact]
summary: CPO Tom Verrilli argues that product management as a specialist role is overused, advocating for fewer PMs and more direct ownership by engineers and designers.
status: ok
---

# This CPO regrets that product management exists | Tom Verrilli (CPO of Whatnot)

Channel: Lenny's Podcast

## Transcript

[0s] As tech companies scaled, somewhere
[1s] along the line, this HR ratio of a pod
[4s] popped into being. Every time you hire
[6s] six engineers, you add a designer, you
[8s] add a PM. Hiring so many PMs
[9s] infantilizes the engineers and the
[11s] designers who are perfectly capable of
[13s] making good decisions, but just never
[15s] had to because there was always a PM to
[17s] babysit them.
[17s] >> Something you wrote online that [music]
[19s] surprised a lot of people at Whatnot.
[21s] Product team was built on the somewhat
[23s] simple premise, we regret that product
[25s] management exists.
[26s] >> Not a thing you probably hear from a lot
[28s] of CPOs. We articulated that way to
[29s] force ourselves to remember that you
[31s] don't hire a PM just for the sake of
[33s] hiring one, you hire one with this
[34s] really specific need. It [music] is
[36s] better to not assume we need a PM in
[37s] every place. The only argument for why
[39s] you would want product management to be
[41s] a specialist function is really it's a
[44s] trade, not a qualification. It's
[45s] something you get good at by doing. It's
[47s] a muscle, but the flip side of that is
[50s] the more you abstract your engineers and
[52s] your designers from doing the same
[53s] thing, their muscle gets under
[54s] developed.
[55s] >> They also wrote this brutal quote, "In
[57s] the last 2 years, 31,832
[60s] people applied to be a product manager
[62s] at Whatnot. We hired one." What do you
[64s] look for in the folks that you hire?
[66s] >> I can tell you what's definitely
[66s] trending down. Folks who spend a lot of
[68s] the time in their interviews talking
[69s] about driving alignment and stakeholder
[71s] management because there's definitely a
[74s] group of PMs whose specialty wasn't
[76s] technical, it was politics.
[77s] >> You're very excited about this move to
[79s] IC work. PMs moving away from this big
[82s] org management world. If you were really
[83s] successful as a PM, you got promoted
[85s] into being a director. We took all of
[87s] our A players and then promoted them out
[89s] of doing things. Why wouldn't [music]
[90s] you want Messi playing for your team
[93s] rather than trying to have the academy
[94s] coming along all the time?
[97s] >> Today, my guest is Tom Verilli. Tom is
[100s] chief product officer at Whatnot, former
[102s] long-time chief product officer at
[103s] Twitch, and former director of product
[105s] growth at Twitter. He's also someone
[107s] I've been wanting to get on this podcast
[109s] for so long. Tom has a lot of hot takes
[112s] and really unique and important insights
[113s] on the future of the product role,
[115s] >> [music]
[115s] >> what he's seeing the best PMs doing
[117s] differently these days and what he's
[118s] looking for when he's hiring product
[119s] people for his team now. If you're not
[121s] familiar with Whatnot, it's a live
[123s] stream shopping platform. It's the
[125s] fastest growing US marketplace business
[127s] of all time. At one point Tom shares how
[129s] he bought a fresh lobster from a
[131s] fisherman on the platform. Before we get
[133s] into it, don't forget to check out
[134s] Lenny's product [music] pass.com for a
[136s] free year of the hottest and most
[139s] beautifully crafted AI product in the
[140s] world available exclusively to Lenny's
[143s] newsletter subscribers. With that, I
[145s] bring you Tom Birilli.
[149s] Tom, thank you so much for being here
[152s] and welcome to the podcast.
[153s] >> Thank you so much, dude. It feels kind
[155s] of surreal after years of watching.
[156s] >> Mhm. I hear that. I hear that people
[158s] come on the podcast. Here you are.
[160s] >> to be first time long time.
[162s] >> [laughter]
[162s] >> That's right.
[164s] Um I want to start with something that
[166s] you wrote online
[168s] that surprised a lot of people and I
[170s] think will surprise a lot of people uh
[172s] coming from a long time chief product
[175s] officer, long time product builder,
[176s] someone that's built a lot of very
[177s] successful products and teams.
[180s] What you wrote is since it's earliest
[182s] inception, the Whatnot product team has
[184s] built was built on the somewhat simple
[186s] premise, we regret that product
[188s] management exists.
[189s] >> Yes, sir. Not a thing you probably hear
[191s] from a lot of CPOs.
[192s] >> No.
[193s] Talk about why you feel this way. Talk
[196s] about how you got to this place. Talk
[197s] about what this means.
[199s] >> Sure. I mean, I always try and start
[201s] with like history in order to understand
[204s] kind of things. And one of the things
[206s] when you come to product management is
[207s] if you go all the way back, like product
[209s] management didn't exist, right? It was
[211s] like the business and you know, very
[213s] often founder CEO types talking directly
[216s] to engineering and design about what we
[217s] needed to build and then executing it
[219s] together.
[220s] And you know, internet businesses it
[223s] turns out scaled a lot faster than any
[225s] other businesses in history.
[227s] And so at some point scale meant
[229s] delegating the specifics of execution to
[232s] somebody or trusting somebody else to
[234s] work out what comes next because there's
[236s] just too many things on for somebody to
[237s] sit with. But, if you think about most
[240s] startups or you think about how that
[241s] evolution worked, that was really a
[242s] specialist role where somebody was
[244s] working it out. It was usually you said
[246s] less things to the engineer where you
[248s] had to describe in less detail what you
[250s] were trying to get done to to a designer
[252s] because
[253s] they understood or they they'd been
[254s] involved. And this idea that we need
[256s] this kind of specialist decision-making
[258s] class of humans in tech is really a more
[261s] modern function
[263s] than it is a like pure necessity.
[265s] You know, the way I would think about it
[267s] and the way that Whatnot has always
[268s] treated it is like
[269s] it would actually be way better if
[271s] engineering and design had the context
[272s] that they needed to just make great
[274s] decisions there if they were so in touch
[276s] with users and what they needed that
[277s] they could
[278s] kind of make the same decisions that a
[280s] product manager is. The only argument as
[282s] far as I'm aware for why you would want
[285s] product management to be a specialist
[287s] function
[288s] is really it's a trade, not a
[290s] qualification. And and what I mean by
[292s] that is it's something you get good at
[293s] by doing.
[295s] It's a muscle, for want of a better
[296s] term, and as every private trainer has
[299s] ever told me, muscles are built by reps.
[301s] And so, the more you do it, the better
[302s] you get at it. But, the the flip side of
[305s] that is the more you abstract your
[307s] engineers and your designers from doing
[308s] the same thing, their muscle gets
[310s] underdeveloped. And so, I think product
[312s] management plays a really important
[314s] role, and I think, you know, you can
[316s] deploy product management to have a
[317s] really important leverage on particular
[320s] things that need to go really well.
[322s] Um and in a lot of ways doing that lets
[324s] design and engineering be the best they
[326s] can be at their craft in those
[328s] situations. But, wherever possible, it's
[330s] kind of optimal to not have a product
[332s] manager
[333s] uh and instead have design and
[334s] engineering going through those those
[336s] kind of steps and making sure that their
[337s] muscles are well ripped.
[339s] >> Do you feel like product management was
[342s] very helpful early on and was important,
[344s] and then it kind of went through a
[345s] period of, well, there's way too many
[347s] PMs that are not that amazing. And now
[349s] there's kind of this coming back to
[351s] okay, what is actually an amazing PM and
[352s] maybe we need fewer of them.
[354s] >> Yeah.
[355s] Uh the funny thing is I think if you
[356s] talk to most engineers or even
[358s] designers, they will remember the great
[360s] PMs that they've worked with
[362s] mostly because they've worked with so
[363s] many bad ones.
[365s] Uh and I don't mean that as a
[366s] disheartening pejorative for folks, but
[368s] I think it's more, you know, you're
[370s] being kind of a really really
[371s] self-critical as a function. Does the
[374s] average PM add a ton of value to folks
[377s] around them in the way that having
[379s] really high-quality product management
[380s] can, you know,
[382s] disseminate clarity and absorb ambiguity
[384s] in the ways that they can or like help
[386s] people make really timely decisions.
[388s] And I think a lot of it is this function
[389s] of like as tech companies scaled and
[391s] winded up hiring so many engineers,
[393s] somewhere along the line um and I don't
[396s] want to pin it on like HR, but somewhere
[398s] along the line this kind of HR ratio of
[400s] like a pod popped into being. You know,
[402s] every time you hire six engineers, you
[404s] add a designer or you add a PM, you add
[406s] an EM and and the you know, the nucleus
[409s] exists.
[410s] And then when you're now serving billion
[412s] DAU products, you got a lot of
[413s] engineers. And so now all of a sudden
[415s] you got a lot of product managers. And
[417s] in most cases, you probably don't need a
[420s] PM for notifications infrastructure.
[422s] Right?
[424s] Engineers are perfectly capable of
[425s] understanding how that works. And in a
[427s] lot of ways, you know, as we just
[430s] described, hiring so many PMs
[431s] infantilizes
[433s] the engineers and the designers who are
[435s] perfectly capable of making good
[436s] decisions, but just never had to because
[438s] there was always a PM to babysit them.
[441s] >> This episode is brought to you by our
[442s] season's presenting sponsor WorkOS. What
[446s] do OpenAI, Anthropic, Cursor, Vercel,
[449s] Replit, Sierra, Clay, and hundreds of
[451s] other winning companies all have in
[452s] common? They are all powered by WorkOS.
[456s] If you're building a product for the
[457s] enterprise, you've felt the pain of
[458s] integrating single sign-on, SCIM, RBA,
[461s] [music] audit logs, and other features
[463s] required by large companies. WorkOS
[466s] turns those deal blockers into drop-in
[468s] APIs with a modern developer platform
[470s] built specifically for B2B SaaS.
[473s] Literally every startup that I'm an
[474s] investor in that starts to expand
[476s] upmarket and start working with WorkOS.
[479s] And that's because they are the best.
[481s] Whether you are seed stage startup
[482s] trying to land your first enterprise
[484s] customer or unicorn expanding globally,
[486s] WorkOS is the fastest path to becoming
[488s] enterprise ready and unblocking growth.
[491s] It's essentially Stripe for enterprise
[493s] features. Visit workos.com to get
[495s] started or just hit up their Slack where
[497s] they have actual engineers waiting to
[499s] answer your questions. WorkOS allows you
[501s] to build faster with delightful APIs,
[504s] comprehensive docs, and a smooth
[505s] developer experience. Go to workos.com
[508s] to make your app enterprise ready today.
[511s] >> Uh something that I find people run into
[513s] eventually when they think this way that
[514s] I want to get your take on is engineers
[516s] and designers don't necessarily want to
[518s] be doing the work of a PM because a lot
[519s] of the work of a PM is kind of annoying
[521s] and not fun. And you know, there's a
[523s] glamour part like making decisions.
[525s] >> less glamorous than people think, yeah.
[526s] >> Yes, exactly. How do you think about
[528s] that just like
[530s] o- uh especially as the company scales,
[531s] engineers having to be in alignment
[533s] meetings, having to write docs, uh
[534s] aligning everyone, uh taking note, you
[537s] know, all these like kind of minutia
[538s] part of PM. And also just the big like
[540s] they also want to you know, engineers
[542s] want to build, engineers want to code,
[543s] designers want to design. How do you
[544s] think about that element of
[546s] s- they may not actually want to be
[548s] doing that work.
[549s] >> I think this is where like
[550s] specialization works both ways, right?
[552s] It is useful to have folks who are well
[554s] honed in making decisions. It's totally
[556s] reasonable as well for someone to say,
[557s] "Listen, uh
[558s] I'm an infrastructure lead and I want to
[560s] think a lot about scale and I do not
[561s] want to have to spend my time debating,
[563s] you know, alignment or getting those
[564s] minutia right. And certain skill sets
[566s] don't necessarily translate super well,
[568s] right? If you're really good at building
[570s] big mental models in your head of how
[572s] infrastructure should scale, you may not
[574s] have the skill set of listening to a
[575s] customer and actually understanding the
[576s] core problem as opposed to the thing
[577s] that they said, which again, just a
[579s] thing that we've built over time. So,
[582s] my supposition of we regret isn't to say
[585s] that we don't want product managers,
[586s] just we recognize in those situations
[588s] that you do need to kind of help other
[590s] functions specialize, but I think it's
[593s] we we articulate it that way to kind of
[595s] force ourselves to remember that you
[596s] don't hire a PM just for the sake of
[598s] hiring one, you hire one with a really
[600s] specific need.
[601s] And I think what goes with it, Lenny, is
[603s] like you have to build the culture of
[605s] the organization around that. So, for
[607s] example,
[608s] when we say, you know, we regret product
[610s] management exists,
[612s] every time we write documents about how
[613s] we ship or what we're doing, we're very
[616s] kind of clear that anyone can bring, you
[619s] know, change forward that we can have
[620s] DRIs of new product development that
[622s] isn't an engineer or is a designer,
[624s] but that everybody goes through that
[625s] same level of function of like, you've
[627s] got to go through product review. You've
[628s] got to actually do the work because it
[630s] is real work.
[631s] And if people don't want to do that, if
[632s] they will kind of want to do something
[634s] else, we can always move product
[636s] management around. And so, rather than
[638s] mapping PMs to teams where you kind of
[641s] assume that the PM will always do that,
[643s] we tend to map them to kind of problems
[644s] or kind of like core projects. And that
[646s] means that there will be a year more
[648s] where there isn't a PM attached to a
[650s] particular engineering team even though
[651s] there's lots of ongoing product work to
[653s] do.
[654s] >> What I love about this is we often hear
[656s] uh
[657s] people at like eng-oriented products
[660s] talk like this. Uh like developer tools
[662s] where we don't need PMs. Why do we need
[663s] PMs? Uh and it's often comes from
[665s] companies like Linear and I don't know,
[667s] like companies that are building
[668s] developer tools where you could see why
[670s] engineers are enough. And so, it's
[673s] really interesting to hear from your
[675s] perspective because you built very
[676s] consumery products, very very consumery
[678s] products that require what you think are
[681s] very strong PM skills. So, it it means a
[683s] lot even more so coming from you that
[685s] you find that PMs aren't as necessary as
[688s] people may think in building something
[689s] great.
[690s] >> Yeah, I think there was like two parts
[692s] that that drove it. And I've certainly
[694s] like a lot of what I'm kind of talking
[695s] about here are things that I definitely
[697s] learned over time, you know, I I spent
[698s] seven years at Twitch prior to time at
[700s] whatnot, and that was very Amazon
[702s] two-pizza team, you know, ratio-driven.
[705s] I started in the valley at Twitter, and
[707s] that was similarly very like you always
[709s] had this kind of very tight alignment.
[711s] And there was God, there was a lot of
[712s] alignment meetings, so you can imagine
[713s] why engineers didn't want to be in them.
[715s] And it's kind of evolved over time to
[716s] understand that like actually a lot of
[719s] that is just a function of like
[721s] management not being able to see what's
[722s] going on. And so you you hire more
[724s] folks, and you build more layers, and
[725s] your systems beget systems.
[728s] And and I think what's really changing
[730s] is people are realizing that like it's
[732s] not true anymore that the only way to
[734s] get leverage is just to hire more more
[736s] PMs underneath you and and be more
[738s] senior. That
[740s] you know, we've got a better
[741s] understanding of what's happening in
[742s] businesses now. It's easier to kind of
[744s] like converse directly with the code
[745s] base and talk to your engineers than it
[747s] ever has been. And so you don't actually
[749s] have to get into this like all-scale
[751s] thing. And I don't really think consumer
[753s] or enterprise is the cutting point
[754s] anymore. I think it's more kind of the
[756s] culture of the organization itself.
[758s] >> How much of this shift in your mind is
[760s] AI-driven because AI now enables non-PMs
[763s] to do PM-y work?
[765s] And how much of it was like pre pre-AI?
[767s] And then I want to talk about just what
[768s] this actually looks like at your team,
[769s] but let me ask that question first.
[770s] >> I don't think it's explicitly because of
[772s] AI,
[773s] but I do think AI makes it a lot easier.
[776s] I think it's two things. I think
[778s] AI certainly means that there's an
[780s] enormous amount of leverage for an IC
[782s] now. I don't know how much other people
[784s] feel this, but I was talking to a couple
[785s] of our senior leads the other day, and
[787s] we feel that we're so capable of doing
[789s] things now with AI that you almost feel
[791s] a bunch of pressure of the stuff you're
[792s] not doing because you know that if you
[793s] carved out a couple more hours, you
[795s] could you could move a lot of stuff.
[797s] Um it certainly easier to move fast with
[800s] AI tooling if you've got well-honed
[802s] judgment. You know, you can call
[804s] data now on a hex thread that is
[806s] basically what it would take a week, two
[808s] weeks with an Amazon L7, you know, uh
[811s] data scientist in 2017. And so, if
[814s] you're empowered and have good judgment,
[816s] you can basically make really quick
[817s] decisions and just kind of keep people
[819s] moving, which is extraordinary.
[821s] But more than just AI, I think it's also
[823s] a lot of those kind of like ratio, you
[826s] know, like kind of pushes also came with
[827s] a bunch of other cultural pushes like,
[829s] you know, hire great people and get out
[831s] of their way and bottoms-up roadmaps and
[833s] and all of those pieces where I think
[834s] there's been a of a cultural shift in
[835s] tech over the last little while, which
[837s] is like, actually top-down's pretty good
[839s] because folks at the top,
[841s] generally speaking, can make quick
[843s] decisions and remove all of this
[844s] alignment debate. They have probably
[847s] more macro context than most folks. And
[849s] assuming that they are genuinely in
[851s] touch with ground truth and are good
[853s] enough,
[854s] it's actually a really efficient model
[856s] to be able to kind of have senior
[857s] leadership involved in a lot of those
[858s] decisions, which means you it's not just
[861s] the tooling that makes them efficient,
[862s] but you can have one very senior PM
[864s] across more things and they can be more
[867s] efficient than having, you know, three
[869s] relatively entry-level PMs. But
[871s] certainly AI makes all of that a lot
[873s] easier again.
[874s] >> So, just again of coming back to the
[876s] broad premise, which I think is very
[877s] important to clarify,
[879s] uh this point about uh regretting
[881s] product management exists isn't saying
[883s] we don't want or think PMs are useful.
[885s] It's that uh it is better to not assume
[888s] we need a PM in every place and uh it is
[890s] a great to enable other functions to do
[893s] the PM work. Uh
[895s] and also just PMs kind of take away the
[896s] reps from people being able to do the
[898s] things that PMs do. And if they can do
[900s] that work, they can actually execute
[902s] better, they'll build better products.
[903s] >> I think it's also better for PMs to not
[905s] be mapped specifically to a team than it
[907s] is to say, we go with the workers. You
[910s] will build more and better reps, you're
[911s] going to work out more muscle groups by
[913s] like moving around on the things you
[915s] work on as opposed to saying, I'm
[916s] attached to whatever this EM owns.
[919s] >> So, let's let's follow about Uh what
[921s] does your what does the team look like?
[922s] What does the PM {slash} and design team
[924s] look like at Whatnot? What What does
[925s] this org look like in this world view?
[928s] >> Yeah, we've we've just passed 20 PMs. We
[930s] have
[931s] kind of I think 21, 22 PMs in the
[934s] building today, which
[936s] for those following Whatnot's trajectory
[937s] is is pretty small considering the
[940s] volume of of GMV that our sellers move.
[942s] We're very loosely organized into three
[945s] groups, kind of buyer, seller, and what
[947s] we would call kind of trust and risk.
[949s] So, the folks looking after our
[950s] standards, you know, payments, kind of
[953s] safety, etc. And then within those kind
[956s] of like broad groups,
[958s] we basically reassign the PMs pretty
[959s] regularly. So, people have like loose,
[962s] you know, alignment. You might be like
[964s] broadly a growth PM or broadly kind of
[966s] like work on discovery, but even amongst
[969s] those groups, kind of gets allocated and
[971s] moved around pretty quickly. And that's
[972s] because the way that we do planning
[974s] is like every 6 months, you know, the
[977s] CEO, myself, some other folks and senior
[979s] leads will sit down and just define what
[981s] what needs to be true over the next 6
[982s] months. What do we have to get done as a
[983s] company?
[984s] Both in terms of like outcomes and
[986s] critical projects. And then we sit down
[988s] and we go through that list and say,
[990s] who's who's the DRI? Who's accountable
[992s] for that? And that's mostly how we end
[994s] up allocating PM work. And so, quite
[996s] regularly you'll have something, you
[998s] know, we've just finished a planning
[999s] cycle literally this morning, and you'll
[1001s] quite regularly go through those and
[1002s] you'll get to the end and say, cool,
[1003s] there's this thing that's like second or
[1005s] third priority in a bunch of different
[1007s] teams' road maps. That feels really
[1009s] important.
[1010s] Who owns that? We actually don't have an
[1012s] owner for that. And we'll go and grab a
[1014s] PM and be like, congratulations, this is
[1016s] the thing we need you to deliver over
[1017s] the next 6 months.
[1018s] It's rarely you need to build a feature
[1021s] that works exactly this way, that does
[1022s] this thing. It's a little more
[1024s] high-level than that. But like, how do
[1026s] you go and work out what needs to be
[1028s] true? And then you map the human beings
[1029s] that you think can kind of got that.
[1032s] Not every name is a PM name, but
[1034s] overwhelmingly when you go through that
[1035s] planning process, it tends to be PMs and
[1037s] it tends to be kind of PMs who have the
[1039s] right skillset vis-a-vis what it is,
[1041s] whether it's more financial, whether
[1043s] it's more kind of like, you know,
[1044s] algorithmic and recommendations-based,
[1045s] whether or not it's like core user
[1047s] feature.
[1048s] >> Let me follow that actual specific
[1049s] thread at the end there around what you
[1051s] look for in product managers that you
[1052s] hire. So, you also wrote this uh brutal
[1055s] quote. In the last 2 years, 31,832
[1059s] people applied to be a product manager
[1062s] at Whatnot. We hired one.
[1064s] >> Yes, sir.
[1066s] >> Pretty great.
[1068s] Pretty great. Um so, there's a few
[1070s] things here I want to talk about. One is
[1071s] just what is it you look for in the
[1073s] folks that you hire, especially these
[1074s] days? What do you find what's kind of
[1076s] like trending up in what you
[1078s] find you need in really successful PMs
[1080s] at Whatnot and what's maybe trending
[1082s] down.
[1082s] >> I can tell you what's definitely
[1083s] trending down. It's uh folks who spend a
[1086s] lot of the time in their interviews
[1087s] talking about those like alignment
[1089s] meetings and driving alignment and
[1090s] stakeholder management and and those
[1092s] pieces because
[1093s] uh
[1094s] there's definitely a group of PMs and I
[1097s] certainly used to be one of them earlier
[1098s] in my career whose specialty wasn't
[1100s] technical or customer-oriented, it was
[1102s] politics.
[1103s] And so, folks who tend to kind of like
[1106s] naturally lean towards like driving
[1109s] alignment, building building
[1110s] relationships, I you know, tell me about
[1112s] a time when you failed and like, "Oh, I
[1113s] didn't keep the CEO up to date with
[1115s] something and that led to a pivot." is
[1117s] definitely kind of a thing that is a bit
[1118s] of an anti-pattern that trends down.
[1121s] What we tend to find kind of really
[1123s] jumps in a PM kind of interview is over
[1126s] the course of of your interviews or your
[1127s] case study, can we see both the macro
[1129s] thinking and the micro thinking?
[1131s] I think the system works something like
[1133s] this and I can describe an end state,
[1135s] but can I,
[1136s] you know, almost exude impatience on
[1138s] like, "And here's how I would validate
[1140s] that very quickly. Here's where I would
[1142s] push to get that done." And are you
[1145s] specific about the things that you've
[1146s] built? All right, there's an awful lot
[1148s] of folks who've worked at, you know,
[1150s] Fang, Uber, pick any scale company where
[1153s] they baby sat things that existed and
[1155s] they maybe, you know, polished the edges
[1157s] of it as opposed to the idea of saying,
[1159s] we were given this problem and I had to
[1161s] go and come up with something unique and
[1162s] novel or I had to kind of really iterate
[1165s] our way through a complicated change.
[1167s] But I made decisions along the way and
[1169s] we moved because I think it's it's easy
[1171s] in a very large organization with
[1173s] inertia to kind of
[1174s] go along with what's happening and not
[1176s] necessarily be an agent of change or be
[1178s] a decision maker, which is ultimately
[1180s] what you need PM to do.
[1181s] >> There's something you said there that I
[1182s] just had Elizabeth Stone on the podcast.
[1184s] She's CPO at Netflix and asked her
[1186s] what's the trait she most that is also
[1188s] most trending up and she said exactly
[1191s] what you said initially, which was the
[1192s] systems thinking, thinking big picture,
[1194s] thinking about the bigger
[1196s] business. And her And her advice there
[1198s] to work on this and I want to ask you if
[1199s] you have any other advice here. Say
[1200s] someone hears this, they're like, oh
[1201s] wow, I got to work on my systems
[1203s] thinking skills.
[1204s] Her advice is take one click back from
[1207s] your problem and think about, okay, for
[1209s] my manager, what do they think about
[1210s] this problem and how does that impact
[1212s] the rest of the business? Um thoughts on
[1214s] just how somebody might develop the
[1215s] skill and get better at systems
[1216s] thinking?
[1217s] >> I mean, I think the skill is exactly the
[1218s] right way to say it. I think you can get
[1220s] good at it from just a mental exercise.
[1221s] So, like, you can do it in small ways.
[1223s] One of the things I ask a lot in product
[1225s] review when someone says, hey, we want
[1226s] to run an experiment, is, okay, what do
[1228s] we do if it's green? What do we do if
[1230s] it's red?
[1231s] And if folks are like, actually, I don't
[1233s] know how my strategy would change, like,
[1235s] cool, we haven't really thought about
[1236s] that. So, like, stop and go and do the
[1238s] mental exercise of how it would work.
[1240s] And then I think it's the same thing if
[1242s] you can build quick local solutions if
[1245s] you have done the mental exercise in
[1246s] your head of saying, well, what would
[1247s] happen if we had a thousand times more
[1248s] usage of this than we expected or what
[1250s] are the unexpected knock-on effects that
[1252s] this could have? And how do you just
[1254s] start doing all of that in your head
[1256s] before you get pen on paper, before you
[1258s] get code on the system? And I think it
[1261s] it helps actually unblock people to move
[1264s] faster, too, because I I the other thing
[1266s] that PMs are often slowed down by is
[1268s] like oh, risk, oh, legal might, finance
[1271s] might, another team might.
[1273s] And I think one of the monikers we use
[1275s] internally is know then go.
[1277s] As in just like think through all the
[1279s] things that could go wrong, understand
[1280s] where scale will break, understand the
[1282s] things that might happen
[1284s] and then make the like move on anyway
[1286s] because if you've thought through all
[1287s] the things that could happen at scale,
[1289s] you're probably going to preempt a bunch
[1291s] of them. So,
[1292s] uh I really like Elizabeth's quote
[1293s] there, but mine is like just do the
[1294s] mental exercise of just like playing out
[1297s] if it gets really widely adopted, if it
[1299s] happens, if there is something that goes
[1301s] wrong, what will it be? You don't have
[1302s] to solve all of them, you've just got to
[1304s] think through all of them and then you
[1305s] end up solving more than you think.
[1307s] >> I love that. So, it's essentially don't
[1308s] just focus on uh will this be an
[1311s] impactful experiment and think about
[1313s] what comes next and what comes next if
[1315s] this is true.
[1315s] >> Like, okay, it's I moved it. Um
[1318s] particularly in a higher growth
[1319s] environment, you're not really looking
[1320s] for a 5% statistic win. You're looking
[1323s] for something that kind of like totally
[1324s] moves the business um and that has kind
[1327s] of compounding vector over time. And so,
[1328s] you just got to think through what that
[1329s] what that is.
[1330s] >> Something else you said that uh is
[1332s] changing in how PMs operate that you're
[1334s] very excited about is this move to IC
[1337s] work, PMs moving away from this kind of
[1339s] big org management world to actually
[1342s] doing the work. Talk about that and what
[1344s] that means for the role of product
[1345s] management.
[1346s] >> I I I I mentioned a little bit earlier,
[1347s] but there was this thing in the you
[1349s] know, in the ratio land where like what
[1352s] you did if you were really successful as
[1353s] a PM is you got promoted into being a
[1355s] director and then all of a sudden it was
[1357s] like don't be hands-on anymore. Your
[1359s] goal goal is just to coach and guide.
[1361s] And so, we took all of our A players and
[1363s] then promoted them out of doing things.
[1366s] Uh and they spent all of their time in
[1367s] alignment and they spent all of their
[1368s] time
[1369s] kind of like coaching and tweaking what
[1372s] their team was doing and you get this
[1373s] really yo-yo development process where
[1375s] somebody does all this work, it goes
[1376s] through a review, it gets told no and
[1377s] you just kind of going back and forward
[1379s] in in reviews.
[1380s] You can see my scar tissue coming
[1382s] through.
[1383s] Um
[1384s] Uh on our team,
[1385s] uh everybody is like there are managers.
[1388s] There's like I think four or five people
[1389s] across the team who manage other PMs.
[1391s] All of them would spend 90 plus percent
[1393s] of their time doing IC work. I'm still
[1395s] probably 50% of my time doing IC work
[1398s] personally.
[1400s] And I think there's kind of a couple
[1401s] real advantages of it. The first
[1403s] um is
[1405s] if you are a kind of ZP product where
[1407s] you've got a decade plus, maybe 15 years
[1410s] of experience building things, hopefully
[1411s] your instincts as to what's going to
[1412s] work or not fairly well honed at this
[1414s] point, you can just make decisions more
[1416s] quickly than people. You can have real
[1418s] impact very, very quickly. And it's
[1421s] really great for the organization to
[1422s] have somebody who can do that as opposed
[1424s] to the idea of like working through
[1425s] three layers of, you know, we we divide
[1428s] the problem up against a most a couple
[1430s] PMs. Those folks need to get into
[1432s] alignment. There's different engineering
[1433s] teams debating stuff. You just tend to
[1435s] go.
[1436s] So, that's wonderful. I think the second
[1438s] thing is
[1439s] there's two ways that that ends up
[1441s] driving leverage. One is
[1443s] a VP in theory can, you know, handle the
[1445s] workload of multiple, kind of like more
[1447s] junior PMs just because, as I said,
[1449s] they're more efficient. And that means
[1450s] that you see more of the board at any
[1452s] given point in time, and so you're far
[1453s] more likely to make the intuitively
[1455s] correct decision for
[1457s] how should we tune the discovery
[1458s] algorithm vis-a-vis people who ship
[1460s] slowly, which is a you know, an
[1462s] evergreen thing in in e-com. Well, if
[1464s] you're thinking about how we, you know,
[1466s] manage
[1468s] uh sellers who ship slowly, and you know
[1470s] about the power of kind of like
[1471s] discovery, you can in either case make
[1474s] the kind of correct decision. You know?
[1476s] One of the things that I remember
[1477s] playing
[1478s] Twitch for a long time, and I was
[1480s] probably one of the people more at fault
[1481s] of it than ever was discovery team and
[1483s] the ads team were always at war for
[1485s] impressions.
[1486s] Right? Like, where do ads go in the
[1488s] feed? What's the impact to discovery
[1490s] metrics? What's the impact to kind of ad
[1492s] dollars? One of the first things I did
[1494s] when I got to Twitch was just put ads in
[1495s] discovery, and make sure that there's
[1497s] the same PM who's accountable for both.
[1500s] Cuz they're going to make the natural
[1501s] trade-off that say the goal is GMV
[1503s] generated from the feed. One of them is
[1505s] through organic, one of them is through
[1507s] kind of like a paid substitution.
[1509s] Solved. And when you put the same person
[1511s] across multiple things, they tend to
[1513s] organically align those things and you
[1514s] just cut out months and months and
[1516s] months of back and forth and the
[1518s] politics that tends to kind of take it
[1520s] from being company first to career
[1522s] first. And so, having
[1525s] VPs mostly in IC land, having directors
[1528s] mostly doing IC work, even having me
[1530s] having to grapple with IC work keeps
[1532s] everybody connected to the ground floor
[1534s] of what's actually true as opposed to
[1536s] what seems true in a review.
[1538s] But also means that you're more likely
[1540s] to make the kind of intuitively correct
[1542s] decision early. Just because why
[1544s] wouldn't you want, you know, Messi
[1547s] playing for your your team when you
[1549s] rather than trying to kind of have the
[1550s] the academy coming along all the time.
[1552s] >> I love this. So, when you talk about IC
[1554s] work for a PM,
[1555s] what does IC work for a PM in this
[1557s] context mean? Does it mean shipping code
[1558s] building or is it like running a team,
[1560s] running owning a road map, writing the
[1562s] strategy doc? Imagine it's the second
[1564s] bucket.
[1565s] >> Uh whatever is required to kind of like
[1567s] most effectively ship is the short
[1568s] answer. Um have I personally shipped
[1570s] some production code and whatnot? Yes.
[1573s] Uh do I think that's really the best use
[1576s] of my time? Not really.
[1577s] You know, I'm I'm certain that quietly
[1579s] somebody reworking most of my code in
[1581s] order to ensure that the linting was
[1582s] correct and the localization worked and
[1584s] all of the new ones that decades of
[1586s] software engineering has taught you
[1587s] that, you know, me and Claude code did
[1589s] not get right. But I do think it starts
[1591s] with like are you literally in the
[1592s] support tickets? Do you know what
[1594s] customer problems we're having? Have you
[1596s] pulled all of the data yourself so that
[1598s] you actually understand it? Have you sat
[1600s] with engineering and design, you know,
[1602s] have you queried the code base directly
[1604s] in order to understand how things work
[1605s] and then have you written the spec? Are
[1607s] you then running a stand-up in a week or
[1609s] I think all all that is just like core
[1610s] individual IC work.
[1612s] >> There's a lot of people that have worked
[1613s] their way up the ladder of product,
[1615s] become a VP, and
[1617s] it doesn't feel exciting to go back to
[1619s] being an IC. Some people like clearly
[1621s] you love it, you enjoy it. A lot of
[1623s] people are like, "Ah, I thought I was
[1624s] done with this. I could just work
[1625s] through people. I could think big
[1626s] picture."
[1628s] How do you feel about that? And what do
[1629s] you What would you say to folks in that
[1630s] in that bucket?
[1631s] >> Uh I think there are probably still a
[1633s] lot of organizations where that is
[1634s] really valuable and that they will go.
[1637s] My recruiting tends to be the folks who
[1639s] are, "Oh my god, I used to love product
[1640s] management and I'm so sick of sitting in
[1642s] alignment meetings and and I'm out there
[1644s] pitching CPOs and VPs of product to be
[1645s] like, "Don't you miss actually doing
[1648s] things? Do you want to come back?" And I
[1650s] think it's okay for us to acknowledge
[1652s] that there'll be a bifurcation across
[1654s] the industry. I do think that there are
[1656s] organizations that are sufficiently
[1658s] large that maybe everyone being hands-on
[1660s] isn't right. I also think, you know, I
[1662s] mentioned earlier that like you have to
[1663s] have a matching culture to kind of go
[1665s] with this kind of environment where
[1668s] that's the expectation, you know, where
[1669s] people are like, "I don't want to talk
[1671s] about it. I want to just, you know,
[1673s] let's go and do that thing."
[1674s] And in a lot of ways, you know, every
[1676s] every startup is a reflection of their
[1678s] founders. And so that naturally tends to
[1680s] be, you know, how do they think and how
[1681s] do they want to run the organization?
[1683s] But I've actually found that a lot for a
[1684s] a lot of the cases
[1686s] you go and talk to somebody who's spent
[1688s] the last five, six years as a as a
[1689s] senior director
[1691s] at, you know, Meta who spends their
[1693s] entire time in alignment meetings and
[1694s] they miss actually talking to customers
[1696s] and talking to engineers and shipping
[1697s] things.
[1698s] >> What makes makes me think about it, I
[1699s] imagine you've seen this list of all of
[1701s] these chief technology officers that
[1703s] have gone to become just engineers at
[1705s] Anthropic. I pulled up this list as you
[1707s] were talking. The CTO of Workday is just
[1709s] a member of technical staff at
[1711s] Anthropic. Uh CEO CTO of Instagram,
[1713s] Box's CTO, uh super.com's CTO, they're
[1717s] just like engineers now at Anthropic.
[1718s] Yep.
[1719s] >> Uh if this is my greatest desire that
[1721s] the Whatnot product bench basically
[1723s] looks like that.
[1724s] >> Mm.
[1724s] >> Um all of these people with like
[1727s] great skills and great understanding and
[1729s] actually end up coming and building as
[1730s] ICs.
[1732s] >> What about the like the the comp of this
[1734s] path, you know? That's people's dream,
[1737s] move up to VP, make millions of dollars.
[1739s] Is there a world where you can still do
[1741s] that and be an IC?
[1742s] >> I actually think it's easier.
[1744s] Spicy take, but like go and take the
[1748s] comp required to have five L5s reporting
[1751s] to one L7 and then four L7s reporting to
[1754s] one VP and now total the comp of that
[1756s] product org and turn around and say,
[1758s] what if I had three people?
[1760s] Why can't I pay them all
[1762s] you know, D2 VP money, particularly if
[1764s] they're having the level of impact that
[1766s] those folks are having there? Like why
[1768s] not?
[1769s] >> And just to fully understand why this is
[1771s] happening, why this should happen, what
[1773s] I'm hearing there's kind of many
[1774s] combinations. One is AI is enabling
[1776s] this, which is great, perfect timing.
[1778s] >> Yes.
[1778s] >> Uh
[1779s] What are the other motivations to do
[1781s] this? Is it just the product ends up
[1782s] being better? Is it fewer people?
[1785s] >> Uh I certainly think the product ends up
[1786s] being better. Like one of the things
[1788s] that folks have been telling us for a
[1790s] long time is yeah, that won't scale.
[1792s] Like oh, leadership isn't going to be
[1794s] able to stay hands-on with what's going
[1796s] on. You're going to need to go and hire
[1798s] tons more layers and what we found is
[1799s] that's actually not true.
[1801s] >> Right.
[1802s] >> It requires a different muscle. You have
[1803s] to make an effort to make sure you
[1804s] genuinely understand, you know, like
[1806s] ground truth. Uh an example we use all
[1808s] the time is we'll be talking in a growth
[1810s] meeting and someone will say, oh yeah,
[1812s] but that was fraud.
[1813s] You know, turn around and say, how do
[1815s] you know that was fraud? Oh, it's
[1816s] labeled in the data set as fraud. Okay,
[1818s] do you know how it gets labeled?
[1820s] I assume someone in ops does it. Okay,
[1822s] do you know the SOP or how they label
[1824s] that? No. Okay, so you don't know it's
[1826s] fraud.
[1827s] Uh and if you push, you know, a really
[1830s] experienced product director, like that
[1831s] they're going to go, good point, I
[1833s] don't. I'm going to go find out. And
[1835s] then you and invariably end up
[1837s] strengthening the the system that agents
[1839s] are using to kind of, you know, data
[1841s] label because suddenly there's a very
[1842s] smart person who's very invested in like
[1845s] understanding how we do that and helping
[1846s] guide it. And so,
[1848s] just that attitude that says like, we're
[1850s] going to do fewer things, we're going to
[1852s] make sure we execute the hell out of
[1853s] them, and we're going to kind of push
[1854s] our best people to be in the weeds
[1856s] everywhere means that you fix lots of
[1858s] things as you go, and you don't end up
[1860s] kind of just making loads of trade-offs.
[1863s] And I think there's a general belief
[1864s] that like
[1866s] it's too easy otherwise for growth to
[1868s] hide all sins. Like you get bigger and
[1870s] you just end up scaling and everybody's
[1872s] kind of like working off of averages.
[1873s] So, culturally I think you got to be
[1874s] really committed to like
[1875s] let's, you know, whole ass few things,
[1878s] as I tend to say sometimes. Shout out
[1880s] Ron Swanson.
[1881s] Uh, and then push your best people to be
[1884s] really in the weeds of stuff. And
[1886s] actually what you find over time is you
[1887s] end up being more efficient by doing
[1888s] that because you actually understand how
[1890s] things work the first time and you make
[1892s] the best decisions. And then listen, AI,
[1894s] huge leverage for all of this.
[1897s] I I can't think of how much time I spent
[1898s] as a junior PM asking my engineers how
[1901s] hard something would be
[1903s] and, you know, distracting actual
[1905s] velocity in order to help scope future
[1906s] stuff. And now I can sit and, you know,
[1908s] talk to Claude and understand roughly
[1910s] LOEs.
[1911s] Um, I can sit there and go through and
[1913s] be like, it feels like there's some car
[1915s] crash of models that must be hitting new
[1916s] users as they open up the app, and you
[1918s] can literally just go through and be
[1919s] like, let's let's load up the feed and
[1921s] talk to me about the logic of who sees
[1922s] what in what order and when does this
[1924s] thing fire, and you can get answers
[1925s] really quickly.
[1926s] And
[1928s] having, as I said before, folks with
[1930s] enough tenure and enough reps that they
[1932s] can see that and say, you know what,
[1933s] that's bad, let's just make a good
[1934s] decision and change the ordering of
[1936s] those.
[1937s] I've saved now a kickoff meeting, an
[1940s] alignment meeting, a week writing PRDs,
[1942s] experiment time, all of that by just
[1945s] having somebody who's kind of empowered
[1946s] to go and make a decision.
[1948s] >> So, coming back to people that are
[1950s] trying to get a job as a PM,
[1952s] whether you're new, let's actually hold
[1954s] off on new people, but people that are
[1955s] say managers,
[1956s] senior folks that are just like, wow,
[1958s] the market has really shifted. The a big
[1960s] thing we're hearing right now is you
[1961s] need to be comfortable with moving back
[1963s] into IC, giving up your fancy title.
[1966s] Is there anything more along those lines
[1967s] of just people looking for a job,
[1969s] struggling to find a job?
[1971s] Any other advice for them?
[1973s] >> Start doing IC work in the role you're
[1975s] in would be my push like get back to the
[1977s] basics of like make sure that you're
[1978s] taking on kind of practical work. Cuz I
[1981s] just think it's like
[1983s] good to make sure that you're keeping
[1985s] those muscles, you know, well honed. I
[1987s] think it also starts with like pushing
[1989s] internally for those things. Like I
[1990s] would I would bet that if you started
[1992s] bringing that level of productivity back
[1994s] into the role that you're in, probably
[1996s] helps you where you are in addition to
[1998s] kind of help you where you might move
[1999s] to.
[2000s] But I think
[2001s] there's a lot of chatter online
[2003s] obviously about like PMs are engineers
[2004s] now.
[2005s] And I think that's all well and good.
[2006s] Like it's a great muscle to go and go
[2008s] and hone. As I said, I've I've pushed
[2010s] some production code because I wanted to
[2011s] go through the exercise of understanding
[2012s] it.
[2013s] But I think there's also
[2015s] before you get to like building things,
[2017s] it's like how quickly can you get back
[2018s] into the muscle of scoping the correct
[2020s] thing, understanding the problem, being
[2021s] able to define what good looks like. And
[2024s] like take advantage of the scale and the
[2026s] scope that you've got that you can see
[2027s] more and bring more to things. Like
[2031s] I think most folks who have sat in a you
[2033s] know, a director plus role will know the
[2035s] pain of sitting there and watching a
[2036s] junior PM yo-yo back and forth on the
[2038s] same PRD back to review where you kind
[2040s] of know what the answer is.
[2043s] Somewhere along the way we decided that
[2045s] you know, lead a horse to water as
[2046s] opposed to kind of help them understand
[2048s] the answer and then keep moving. And I
[2049s] think there's like something in our
[2050s] coaching styles that we can get back to
[2052s] of like
[2053s] help somebody understand what good looks
[2055s] like relatively quickly as opposed to
[2057s] just endless review yo-yo.
[2059s] >> Uh this point you make about PMs not
[2061s] uh shipping to production, I so agree
[2063s] with. This my mind changed on this
[2065s] recently with a previous podcast guest
[2067s] um with this point that PMs are really
[2069s] like the leverage
[2071s] PMs have so much more leverage if
[2072s] they're not sitting there trying to ship
[2073s] to production. They can enabling the
[2075s] team to ship better and faster and
[2077s] making sure the things that are shipping
[2078s] are better is a much better use of PM's
[2079s] time than sitting there shipping stuff.
[2081s] >> I mean this sounds really silly, mate,
[2082s] but like it takes me substantially
[2084s] longer to go through the minutia of like
[2086s] getting get commit and all of the kind
[2088s] of like pieces that are second nature to
[2090s] engineer a line than it does to actually
[2092s] work out what problem is and be able to
[2094s] describe it. So yes, like good practice
[2096s] to try and you know make sure you're
[2098s] doing something
[2099s] for example I don't want to judge if our
[2100s] dev tools have gotten easier or not
[2102s] based on what someone tells me. I'm
[2103s] going to go and try it and be like, yep,
[2105s] that was easier than last time I did it.
[2107s] But I do think you can get an awful long
[2109s] way understanding the code base and then
[2111s] talking to somebody who can actually
[2113s] execute well, you know, otherwise you're
[2115s] just going to get you're going to you're
[2116s] going to fall a foul of like a thousand
[2118s] classic traps that every other engineer
[2120s] learned how not to do when they were in
[2121s] L4.
[2122s] >> So following that thread a little bit,
[2124s] what are some of the ways that AI has
[2126s] enabled you and or your team to move
[2129s] faster and be more productive
[2131s] other than prototyping, which is the
[2132s] very
[2134s] clear benefit of AI for PMs and product
[2136s] teams. What else would have been the top
[2138s] three of like, wow, this is really
[2139s] unlocked our productivity and
[2141s] and the quality of what we do?
[2143s] >> I mean the first one I think by a
[2144s] country mile is data science.
[2147s] You know, we use Hex Threads internally
[2148s] at Whatnot. I'm sure there are other
[2150s] kind of like comparable products, but I
[2152s] think it's almost hard to remember time
[2154s] as a PM before you had tooling like that
[2157s] where you could genuinely start pulling
[2159s] very nuanced cohorts of data where you
[2162s] could kind of grab an individual user
[2164s] where you've heard a report
[2165s] actually understand, you know, like
[2167s] let's pull logs. Help me understand
[2168s] exactly what this user did and saw. How
[2171s] many other users look like this? You
[2173s] know, what would impact be? And then all
[2174s] of a sudden you can build pretty
[2176s] meaningful sensitivity models or like
[2178s] forecasts of what might happen,
[2179s] regression models, etc. Really, really,
[2182s] really quickly, which is like incredibly
[2183s] powerful.
[2185s] Um the other thing that we found is it
[2186s] helps us move much more quickly with
[2188s] shipping things because you can spot
[2190s] regressions and weird knock-on effects
[2192s] of two products intermixing more quickly
[2194s] than you used to be able to.
[2196s] And I think in, you know, really large
[2197s] complicated systems, that's always one
[2199s] of the things that ends up slowing you
[2200s] down of like release trains and all of
[2202s] that versus like if you build the right
[2204s] AI tool, you can spot regressions really
[2207s] quickly, which basically lets people
[2208s] just kind of like go. So, I don't know
[2211s] what the future of data science looks
[2212s] like, but I think as a product manager,
[2215s] I've spent less time in the last year
[2216s] talking to a data scientist than I ever
[2218s] have in my career, even though I've
[2219s] probably spent 10 times more time in
[2222s] data and understanding actually how the
[2223s] product's working than I have ever have
[2225s] in my career. So, that one I think is
[2227s] really powerful.
[2228s] Second one I've already mentioned, which
[2229s] is like stop bothering engineers with
[2231s] how does the code base work and actually
[2233s] just go and talk to Claude and
[2234s] understand it, which is really helpful,
[2236s] you know, I used to say earlier on in my
[2238s] career that the goal was always to be
[2240s] understand your systems at the boxes and
[2242s] lines level of, you know, which system
[2243s] drives which thing and now there's no
[2245s] excuse not to understand that or a
[2247s] nuance layer.
[2249s] But the other one, and this might be
[2251s] very specific to Whatnot, so I don't
[2253s] know that this will help everyone, but
[2254s] one of the things that I've been lucky
[2256s] to do in my career is basically worked
[2258s] on live products for a decade now.
[2260s] And so, it's always been really cool to
[2262s] be able to ship a product and then watch
[2264s] a customer use it and watch them kind of
[2266s] figure it out. So, like, uh,
[2269s] I think people have just gotten this
[2270s] experience with like Listen Labs and
[2272s] and, you know, others in that cohort of
[2274s] watching people use your product, but
[2276s] I've always been able to sit and watch
[2277s] people use a thing for the first time
[2279s] and go through that new user
[2280s] comprehension gap. What's really cool
[2282s] with a bunch of the AI tooling right now
[2284s] is as they're describing, oh, I'm having
[2286s] a problem, you can literally be watching
[2287s] the code base live
[2289s] and work out is that actually a bug
[2291s] that's happening right now, right now,
[2294s] or is that a comprehension gap where it
[2295s] doesn't work as expected and suddenly
[2297s] you've got this video artifact of
[2299s] someone using your product. You can be
[2301s] analyzing the code base in real time.
[2303s] And you can be just talking kind of
[2305s] through AI to the code base to
[2306s] understand what's actually happening and
[2308s] it's this it's like
[2310s] you know, a feedback loop on steroids
[2312s] because all of a sudden you know exactly
[2313s] what's going on customer side, code
[2315s] side, and observe side as as like a
[2318s] viewer
[2319s] in real time, which is really cool.
[2320s] >> That sounds uh
[2322s] both awesome and very stressful to be
[2324s] building products that are that
[2326s] live in real time. I think about um
[2327s] Netflix where they invest in live now
[2329s] and that's just like all you've done for
[2331s] 10 years and how big of a deal that was
[2333s] for them. I know the scale is different,
[2334s] but
[2335s] >> Honestly, my second week at Whatnot, um
[2337s] I remember sitting in a room watching so
[2340s] uh
[2341s] uh one of those not familiar uh kind of
[2343s] commerce auction live auction platform
[2345s] but live commerce platform um and
[2347s] there's varieties of different ways to
[2348s] run an auction, but one of them is
[2349s] what's called sudden death, which is
[2351s] like when the timer ends, it ends.
[2353s] Right? Otherwise, the classic auction
[2354s] environment, somebody bids in the last 5
[2356s] seconds, it adds 10 seconds back on the
[2358s] clock.
[2359s] Uh and a seller can decide what auction
[2360s] model they want. And I was sitting there
[2362s] watching a seller who was like, "These
[2364s] are taking too long. I wish this
[2365s] 7-second timer was actually 3 seconds
[2367s] cuz I want to move more product." And
[2369s] they I I watched two engineers in the
[2370s] office look at each other and be like,
[2372s] "That's a config. We could totally do
[2374s] that." And so they went and updated it
[2375s] in real time and then jumped in the chat
[2376s] of a show and just said, "Refresh your
[2378s] app." And then all of a sudden, bang, it
[2380s] was operating that way and I was like,
[2382s] "Cool. I'm I'm with my people. I'm in
[2383s] like the right place." Because that's
[2385s] the level of like responsiveness that
[2387s] you can get uh in a live environment and
[2389s] obviously AI means that that's really
[2391s] easy for lots of people to take on. Not
[2393s] that I would touch production code
[2395s] uh in that kind of way cuz that's a
[2396s] disastrous idea. Uh but for qualified
[2399s] humans, it's a wonderful one.
[2400s] >> That is very cool.
[2402s] This episode is brought to you by
[2403s] Mercury, radically different banking
[2405s] loved by over 300,000 entrepreneurs. And
[2409s] now with Command.
[2410s] >> [music]
[2410s] >> I've been a customer of Mercury's for
[2411s] over 6 years. I have never once thought
[2414s] about leaving. Mercury is basically what
[2416s] happens when banking is built by product
[2418s] people, not by bankers. They make it so
[2421s] easy, dare I say fun, to send invoices,
[2424s] [music] move money around, set up
[2426s] virtual cards for folks on my team. Does
[2428s] your bank have an API, a terminal native
[2431s] CLI, or an AI-ready MCP server? I don't
[2434s] think so. And [music] just recently they
[2436s] launched Command, a conversational
[2439s] interface built directly into Mercury,
[2441s] which acts as your financial operator.
[2443s] I've been using Command to transfer
[2445s] [music] money around, to figure out what
[2446s] categories I've been spending the most
[2448s] money in, analyze my cash flows, and
[2451s] just today I used it to find out how
[2452s] much I've made from a specific sponsor
[2454s] over the past year. I just asked, "How
[2456s] much have I made from X over the past
[2458s] year?" 10 seconds later I have an
[2460s] answer. It is so freaking cool. Visit
[2463s] mercury.com to learn more and apply
[2465s] online in minutes. Mercury is a fintech
[2467s] company, not an FDIC-insured bank.
[2469s] Banking services provided through Choice
[2471s] Financial Group and Column NA, members
[2473s] FDIC.
[2475s] >> Coming back to this data science point,
[2477s] uh I I I had a I have a friend who's a
[2479s] data scientist, and he said it's a rough
[2482s] time for data scientists because of this
[2484s] exact thing. And to add a little more
[2486s] color, uh basically their time used to
[2488s] be
[2489s] asked to do some data analysis, do some
[2491s] work with the data, come back, here's
[2493s] the results, here's I'm confident in
[2495s] this conclusion. Now their time is
[2497s] basically uh seeing like half-assed data
[2500s] science work from non-data scientists,
[2502s] and and just like show me, is this
[2504s] right? And then and and half the time
[2505s] it's wrong. And they're like, what the
[2507s] hell is my job now? It sucks.
[2509s] >> Yeah, so uh I have a lot of empathy for
[2511s] that because I've certainly seen it.
[2513s] Another plug for why having fewer more
[2515s] senior PMs is helpful because there is a
[2518s] folks who've seen more of those reps and
[2520s] understand. But I also think a lot of
[2522s] the time that uh lack of clarity comes
[2524s] from the other thing, which is
[2526s] organizations have historically
[2527s] underinvested in data engineering
[2529s] and data structures and and good data
[2530s] labeling, and whether or not you've got
[2532s] your kind of um not just your tax on a
[2534s] maze right but whether or not you really
[2537s] understand whether or not your data
[2539s] systems and structures are set up well
[2541s] and so what we found is a lot of our
[2542s] best data scientists are pushing in that
[2543s] direction of like are we actually you
[2546s] know correctly updating
[2548s] all the ways in which our tracking and
[2549s] attribution works so that it's less easy
[2552s] for people to kind of like misunderstand
[2553s] those and then yeah
[2555s] using an AI tool to find a piece of data
[2558s] much like using an AI tool to write code
[2560s] doesn't absolve you of responsibility to
[2562s] make sure that that was good analysis
[2564s] good code it's just tends to be
[2566s] leveraged for those people who are
[2567s] naturally inclined that way.
[2568s] >> So if you think about these different
[2569s] functions data science user research
[2571s] design engineering PM what I'm hearing
[2574s] so far is we'll need probably fewer PMs
[2576s] we'll need fewer data scientists are
[2578s] there any other roles that you think
[2579s] they'll they're kind of trending down in
[2581s] terms of we'll need and are there any
[2583s] roles trending up like wow we're going
[2584s] to need a lot more of this kind of
[2586s] person?
[2587s] >> Well funnily enough like as I say I
[2588s] haven't spoken to a data scientist in a
[2590s] while we've we've certainly still hired
[2591s] plenty because I do think that kind of
[2594s] all that tracking and attribution and
[2597s] measurement really really powerful I
[2599s] also think one of the flip sides of we
[2601s] say we need fewer it's like
[2603s] fewer of for the same output
[2607s] doesn't necessarily mean fewer of in
[2610s] macro because if you are using these
[2612s] systems right you can just grow more
[2614s] quickly you can build more things you
[2615s] can take on more stuff so
[2618s] you know I would be surprised actually
[2620s] if we ended up with like net fewer I
[2623s] think it's more like net fewer vis-a-vis
[2625s] customer impact for both
[2628s] um
[2629s] certainly I think kind of like trending
[2631s] up over time within these roles
[2634s] are these kind of like and I don't
[2636s] remember the exact label that people
[2638s] used to use but this idea of like
[2640s] tech leads that are kind of like a
[2642s] hybrid EM where you're running a very
[2644s] small team kind of running at things
[2647s] because same way that we say
[2649s] you know, the cost of trying something
[2650s] has come down. The idea that says
[2653s] you can have kind of core focus, which
[2654s] is a thing we're very big on. We've also
[2656s] found incubating a lot of smaller teams
[2658s] to kind of go over and sit in the corner
[2660s] and just try and build this thing going
[2662s] you know, it's not quite prototyping,
[2663s] but it's like go and take a swing at a
[2665s] thing that we've historically thought
[2666s] was too hard
[2667s] is getting cheaper and increasingly has
[2670s] very high leverage. So, the idea of like
[2672s] engineering manager light
[2675s] for one of a better term, uh is a thing
[2677s] that I definitely think that we'll see
[2678s] more and more of. It's like not quite
[2680s] the technical member of staff, although
[2682s] that must be lovely, um but certainly
[2684s] not the kind of like full-on EM role. I
[2686s] think it's one that that definitely will
[2687s] come about more often.
[2689s] >> So, maybe just to close the loop on this
[2690s] part of the conversation, there's this,
[2692s] you know, trend towards everyone's kind
[2694s] of a builder, PMs are shipping a little
[2696s] bit, being a little more engineer,
[2698s] engineers are taking on more of the PM
[2699s] work. How do you think this just kind of
[2701s] maybe plays out over the next couple
[2703s] years in terms of what product teams
[2705s] look like broadly? Is it still PM,
[2708s] engineers, a designer
[2710s] data scientist somewhere? Is there How
[2712s] do you kind of envision the the
[2713s] canonical product team over the coming
[2715s] years?
[2716s] >> Great question,
[2717s] uh and I don't know that I know what it
[2718s] looks like everywhere.
[2720s] I think what it'll look like at whatnot
[2721s] is it probably still looks mostly like
[2724s] it has historically, which is, you know,
[2726s] there are reasons that you'd have a
[2727s] specialist designer, specialist
[2728s] engineers, specialist product
[2729s] management.
[2731s] I think those kind of very concrete
[2733s] teams will largely be reserved for like
[2736s] very specific projects or like things
[2738s] that we have high confidence or
[2740s] conviction that we need to solve or
[2743s] we're pretty high confidence conviction
[2744s] that we have a path forward on a thing
[2746s] and we want to make good progress.
[2748s] And I think at the edges around that is
[2750s] going to be a lot more free space for
[2751s] people to play on, like, "Hey, I'm
[2753s] reasonably sure I can go and make a
[2755s] meaningful improvement to this thing and
[2757s] it doesn't matter if you are
[2759s] a designer, an engineer, a product
[2761s] manager, a data scientist, you can and
[2764s] you should.
[2765s] Right? If you're sitting there on a
[2766s] Friday afternoon and you can't focus on
[2767s] the PRD you're writing, but you're
[2768s] pretty sure you can go and fix
[2769s] something, go for it. Um and so I think
[2773s] it probably doesn't morph
[2775s] in the more formal sense, but I do think
[2777s] there's just a lot more free space for
[2778s] people who are
[2780s] well-versed in the customer problems,
[2781s] well-versed in the code base, and
[2783s] understand some of that macro context
[2785s] will just be empowered to do more and
[2787s] more things.
[2788s] >> Two or two and a half years ago I had
[2789s] this post I put out where I said, "Why
[2791s] PMs are the best positioned role in tech
[2793s] to thrive in an AI world?"
[2795s] And I feel like
[2797s] even though the beginning of our
[2798s] conversation was like, uh we should live
[2800s] in a world where we regret PM exists, I
[2802s] feel like
[2803s] we agree on this idea that the skills
[2805s] that seem to matter most and are going
[2807s] to be most valuable, whether it's a PM
[2808s] doing them or an engineer or designer,
[2810s] is very PM-y skills. I'll share a few
[2812s] examples in this from this post. Like
[2814s] what what do we Who is really good at
[2816s] this stuff? These these things.
[2818s] Identifying what to build, distilling
[2820s] and communicating requirements,
[2821s] prioritizing everyone's ideas for the
[2822s] highest ROI opportunities,
[2824s] uh giving feedback on designs to improve
[2826s] impact, developing go-to-market
[2828s] strategy, understanding business
[2829s] strategy. Like to me this is what PMs
[2831s] do, and it feels like that's becoming
[2833s] more and more important. So I guess the
[2834s] question to you is if you agree with
[2836s] this idea that the PM-y skills seem to
[2838s] be the most valuable now as AI takes on
[2840s] the building.
[2841s] >> Definitely agree. Also props for
[2843s] bringing in the receipts even with a
[2844s] timestamp, my friend. Well done.
[2846s] >> Mhm.
[2847s] >> Um
[2848s] I think what I my only build on that
[2850s] would be to say I think those core PM
[2853s] skills aren't necessarily the things
[2855s] that we have rewarded PMs for over the
[2856s] last 5 years versus storytelling,
[2860s] alignment, strategy. And so I do think
[2864s] you're 100% correct that like can I
[2866s] genuinely understand the customer, can I
[2868s] genuinely understand the business, can I
[2869s] genuinely understand the tech, and can I
[2871s] translate the three together for, you
[2873s] know, optimal efficiency?
[2875s] Is the point of leverage when doing
[2878s] things gets cheaper, trying things is
[2880s] cheaper, etc. Um so absolutely agree
[2883s] that product skills are probably the
[2886s] most durable. My build would just be
[2888s] there's a lot of people who have the
[2889s] title PM who haven't spent a lot of time
[2891s] building those skills in the last 5
[2893s] years,
[2894s] but have gotten really good at
[2895s] communicating frameworks to leadership.
[2898s] Uh and so I just kind of push us back as
[2900s] a function into like that core work.
[2903s] >> Yeah, Marty Cagan calls this product
[2904s] theater.
[2905s] A lot of people just do the things that
[2907s] PMs should be doing.
[2908s] >> Or and listen, I'm guilty of this. We
[2910s] have we rewarded it for so long,
[2912s] >> Yeah.
[2912s] >> uh that like it doesn't surprise me that
[2914s] that product theater is a is a core
[2916s] skill set for a lot of folks. I just
[2917s] think that there's not a lot of place to
[2919s] hide in that anymore.
[2921s] >> And this comes back to that 32,000
[2923s] people applied for jobs. I imagine a lot
[2925s] of that is just people who think they're
[2927s] PMs or have the title PM, but don't have
[2929s] but are exactly what you described. They
[2931s] just focus a lot on alignment, writing
[2932s] docs, meetings, things like that, and
[2934s] not actual building, understanding what
[2936s] it takes to build a successful product.
[2937s] >> Yeah, the number of people who do really
[2939s] well in a product interview Lenny, and
[2941s] then you give them a case study. So
[2942s] everyone who gets hired at Whatnot in
[2943s] any role has to do an actual hands-on
[2945s] case study, but the number of people who
[2948s] present incredibly well, and then you
[2949s] give them a prompt and some data and ask
[2951s] them to come back with a PRD on
[2952s] something, and then we make them
[2954s] verbally defend it,
[2955s] how quickly the thinking decays from
[2958s] folks who are good at the theater, but
[2959s] not the specifics, uh is kind of really
[2962s] telling, I think.
[2963s] >> Okay. So kind of going uh beyond the
[2965s] hiring step, say you hire somebody,
[2968s] what are some things you've learned
[2970s] about how to get the most out of the
[2971s] people you hire? You mentioned this kind
[2973s] of uh contrarian take that you don't
[2976s] agree with this hire great people get
[2977s] out of their way. So I want to hear more
[2979s] about that. And just is there anything
[2980s] else you've learned about just uh
[2982s] elements to building a very successful
[2984s] world-class product team?
[2985s] >> I mean,
[2987s] nuance required in my life hire people
[2989s] and get out of the way is wrong. I will
[2990s] say um
[2991s] but couple couple pieces to build off of
[2994s] it. I think in general the higher great
[2996s] people and get out of their way became
[2999s] this kind of like macro saying for
[3002s] let them work out what the road map is.
[3004s] Let them work out what the problems are.
[3006s] Just like completely
[3008s] devolve, you know, like what's going on.
[3010s] And I think the real answer is obviously
[3012s] the better people you hire the more you
[3014s] can kind of like totally trust that they
[3016s] know what they're doing. But we tend to
[3018s] live in a verify then trust land as
[3020s] opposed to a like totally trust or even
[3022s] trust but verify, which is like
[3025s] I'm probably in a better position than
[3027s] any of my kind of directs to understand
[3029s] how all of the different pieces of our
[3030s] system buyer seller kind of trust fit
[3033s] together.
[3034s] They're almost certainly in a better
[3036s] position than me to understand the
[3038s] nuance of like how any of those
[3039s] individual features work. You know, if
[3041s] somebody quizzed me today on exactly all
[3043s] the waitings in the Whatnot Discovery
[3045s] model, I would definitely be wrong
[3047s] vis-a-vis any of the engineers on that
[3048s] team, vis-a-vis any of the team, you
[3050s] know, uh, the PMs on that team. Good.
[3053s] But like it's actually kind of incumbent
[3055s] on me to learn that and understand that
[3057s] over time because I'm asking them to
[3058s] make decisions and I am proving things
[3061s] that they're doing. And so
[3064s] time spent actually working alongside
[3066s] those teams like in the trenches trying
[3068s] to solve something really really
[3069s] powerful.
[3071s] One of the things that I saw earliest
[3072s] when I joined Whatnot that I've seen
[3074s] kind of Grant, who's our founder CEO, do
[3076s] is he'll sit in a review and be like, I
[3078s] don't think this is right.
[3081s] And then he'll pause and say, I'm going
[3082s] to clear the rest of my day. Let's sit
[3084s] and figure it out.
[3085s] And he'll actually end up sitting with
[3087s] the team and going through, you know,
[3089s] again, easier with AI data tools. Let's
[3092s] literally pull up the tickets. Let's
[3093s] literally pull up the code. Let's like
[3095s] go through the data line by line and
[3097s] understand what's actually happening so
[3099s] that we can make a decision there.
[3102s] And it means he's very up to date with
[3104s] what's going on. Kind of very culturally
[3106s] sets the tone for the team that like
[3108s] we're just seeking truth.
[3110s] And it makes it very much a like us
[3112s] versus you. Like reviews got very into
[3115s] like listen for yes for a little while
[3118s] there where all you're trying to do as a
[3119s] PM is just get a a green light so that
[3121s] you can go back to your your engineers
[3123s] and say I have some credibility. I can
[3124s] get, you know, the CPO to approve what
[3127s] we're building as opposed to this idea
[3129s] that says we're just trying to find out
[3130s] the right answer and in theory everyone
[3133s] wants us to come up with the right
[3134s] answer. So
[3135s] you know, we use planning to to align
[3138s] the company on like what are the really
[3139s] important things we have to solve.
[3141s] That's mostly a resourcing
[3143s] discussion, right? Like if we pick the
[3145s] right things stack rank them all,
[3147s] ultimately I'm most accountable for
[3149s] making sure that we have the right
[3150s] resources in the right places to hit
[3152s] things. But like I don't know if those
[3154s] are the right places if all I do is
[3155s] delegate to the team to go figure it out
[3157s] and I'm not actually periodically very
[3159s] deep with them on like exactly how does
[3161s] that work, you know?
[3163s] How do our like you know, referrals
[3166s] work? Literally what is the logic that
[3169s] fraud might use in order to invalidate
[3171s] one? Oh, based on address signals. How
[3173s] do we calculate address signals? Is that
[3175s] like a Google normalized thing or is
[3177s] that like, you know, free text that's
[3179s] put in? And if you don't actually push
[3181s] yourself down to sit alongside your IC
[3184s] engineers and your IC designers and your
[3186s] IC PMs, you don't actually know that
[3188s] stuff. So you can't make good macro
[3190s] decisions without the micro.
[3192s] So increasingly I think
[3195s] I you know, I I push myself to I try and
[3197s] be T-shaped. I can go very very deep
[3199s] when required but I'm mostly broad
[3201s] across pieces and I think the idea of
[3203s] just like hiring people and then like
[3204s] not asking any more questions and
[3206s] delegating all of the detail just isn't
[3208s] really
[3209s] the most successful model for getting
[3211s] the most out of an organization.
[3212s] >> With a very product-minded founder, CPO
[3216s] classically is a very challenging role
[3217s] for people because you're basically with
[3220s] person between a very opinionated
[3221s] founder and the team building it. Uh
[3224s] what have you
[3224s] >> Yeah. found works in creating a, you
[3227s] know, environment where you are happy in
[3229s] that role?
[3231s] >> Yeah, and kind of funnily that's been
[3233s] most of my career. Actually, I worked
[3235s] for three founder founders in a row who
[3237s] are all kind of very product-minded. Um
[3239s] and at Whatnot I've got two, which is a
[3240s] blessing, actually.
[3242s] In general,
[3243s] uh I try not to kind of double up if,
[3246s] you know, Grant or Logan, our our
[3248s] founders, are on a thing, probably
[3250s] doesn't need me. Like, what's the
[3251s] advantage of an extra layer? Um you
[3255s] know, I think jokingly one of the PMs on
[3256s] the team has referred to it as the two
[3258s] dads problem, where you've just got like
[3259s] two people dishing kind of like
[3260s] conflicting instructions or somebody
[3262s] wants to review, and then you do all
[3264s] this work to present it to me, and then
[3265s] you go back and it gets a different
[3266s] thing. So, my first thing is like,
[3269s] if Grant or Logan are on it, I check
[3271s] that they're watching it, they're
[3272s] accountable for it, and I step out. So,
[3274s] there'll be long periods of time where
[3275s] like fully half my team could be working
[3277s] on something,
[3278s] and I couldn't tell you day-to-day where
[3279s] it is,
[3280s] because, you know, it's with Grant, it's
[3282s] with Logan, and that's totally fine. I
[3284s] don't have to be across
[3285s] all of the things they're doing. We just
[3287s] need to make sure that there is somebody
[3288s] doing that bar raising.
[3290s] So, we spend a lot of time doing that
[3292s] alignment. I think the second one is
[3293s] like, ultimately, if you are a product
[3295s] leader in a founder-led company, you
[3297s] have to understand it's not your
[3298s] company, it's theirs, and you just find
[3300s] the right balance of like, you know,
[3302s] "Hey, are you open to feedback on this?
[3303s] Have you made up your mind? You know,
[3305s] are you open to a push on this?" And you
[3306s] just find that rhythm kind of working
[3308s] with folks. Generally speaking, you
[3311s] know, the there's a reason that
[3312s] founder-led companies do so well in our
[3314s] industry, you know, the insight required
[3316s] and the kind of customer intuition to
[3318s] make the thing in the first place and
[3320s] work tends to be really important. Um
[3322s] and then I just view my job as, you
[3323s] know, making sure that we've got
[3325s] coverage on the places where our
[3326s] founders are.
[3327s] >> Awesome. So, a couple things you've
[3329s] learned here for building and this I
[3331s] think in consumer this is especially
[3333s] important is uh counterintuitively to
[3336s] how maybe people think things should
[3337s] work, you're finding that the best
[3340s] teams, companies, products end up coming
[3342s] from top-down, founder-led almost you
[3346s] know, micromanagement is a is a dirty
[3347s] word to a lot of people, but it's
[3348s] basically being in the weeds is uh in
[3351s] spite of how people may feel, this
[3353s] actually ends up being leading to better
[3354s] stuff.
[3355s] >> Top-down works well if leadership is
[3357s] good enough to be in the weeds and be
[3359s] specifically correct. I think where it
[3360s] falls apart is where you don't actually
[3363s] know
[3364s] ground truth and then you attempt to
[3367s] manage people from above and that's
[3368s] where I think the term micromanagement
[3369s] comes from. Otherwise, if you're working
[3372s] from the same data and you have it, I
[3373s] don't know a junior engineer or a an
[3376s] entry-level designer who isn't stoked to
[3378s] sit there and work alongside the CPO or
[3380s] the CEO and ship something cuz you're
[3382s] just unblocked. There's no alignment,
[3383s] meetings there's nothing to do. They
[3384s] also find that like you can give way
[3386s] better feedback if you're literally in
[3388s] the detail. I think the nuance is just
[3390s] like how do you make sure you can do
[3392s] that in enough places? There's never
[3393s] been a better time to be in to to
[3395s] attempt to be in the detail on things
[3397s] because you can literally query it in
[3399s] real time.
[3400s] >> I think that's such an important nuance
[3401s] here. The story told is so um powerful,
[3403s] this idea of Grant just okay, I'm going
[3405s] to clear my day, I'm going to spend time
[3406s] going deep on this stuff. That feels
[3408s] like a very necessary ingredient for
[3410s] someone at the top to be making right or
[3412s] wrong decisions because your point if
[3414s] they don't have all the details, they
[3415s] don't they're not making decision out of
[3417s] real data.
[3418s] >> And we can do that because, you know, as
[3420s] I said, we we plan what we're doing
[3422s] and then we allocate and then we're
[3424s] dividing and conquering. So, like he's
[3426s] probably trying to nail three or four
[3427s] most important things at a time and you
[3429s] know, he's the CEO, he's going to call
[3430s] the ball and say these are the four
[3431s] things I own right now and I'm going to
[3432s] be like, "Great. I'll be over here
[3435s] then." And then within those, like what
[3438s] am I doing with the rest of my day that
[3440s] is more important than nailing the five
[3441s] things I've said we'll get done this
[3443s] half. Like if it's anything other than
[3445s] maybe hiring
[3447s] standing meetings, any of that stuff,
[3448s] you can just clear it and set the tone
[3450s] for the team that like until we
[3451s] understand it
[3452s] we can't do anything else.
[3454s] >> Say somebody is looking for a CPO role
[3457s] or a first PM role, kind of similar,
[3460s] basically working for a founder.
[3462s] Uh what would be your advice for them to
[3465s] land in a place where they're happy and
[3466s] not just super frustrated by by this
[3468s] kind of middle layer where they just
[3470s] don't actually have any agency.
[3472s] >> Yeah, I mean I think the first thing
[3473s] you've got to work out is like why do
[3474s] you want it?
[3475s] Right? There's this idea that like the
[3477s] CPO's job you get to decide all of the
[3479s] road maps and all these things and I've
[3480s] got bad news for you that's like not
[3481s] strictly true.
[3483s] But I think it also comes down to like
[3484s] spending some time with that person to
[3485s] work out like how do we jam on a topic?
[3488s] How do they like to get pushed? How do
[3490s] they not? And then you spend a lot of
[3491s] your time calibrating. So like before I
[3493s] joined Whatnot, I think Grant and I had
[3495s] like five or six different coffees where
[3497s] we talked about like how do you get, you
[3500s] know, this type of team to move? How do
[3502s] you work on those things?
[3503s] And then I obviously went through a
[3505s] series of interviews and actually came
[3506s] down to kind of LA where Grant and Logan
[3508s] were based at the time and spent a whole
[3510s] day with him.
[3511s] Just like in a room going through a
[3513s] couple different problems, talking
[3514s] through different things in the road
[3515s] map, just like really getting into it.
[3517s] And I tried through that to kind of be
[3520s] my most ordinary self as opposed to like
[3522s] interview self, if that makes sense. Cuz
[3524s] you kind of got to ask yourself, do I
[3525s] really want to spend
[3527s] my whole time like having this
[3529s] discussion and this debate? But like
[3532s] I like being a CPO or any kind of
[3534s] product lead because in a lot of ways
[3536s] what I'm doing is I'm like helping
[3537s] translate that vision and that intuition
[3540s] into reality. And then, you know, there
[3542s] is an art to learning how to push
[3543s] somebody
[3544s] without kind of competing with them. And
[3546s] I think a lot of the times I've seen the
[3548s] CPO CEO relationship go badly and ends
[3550s] up with like the CPO is competing with
[3552s] the CEO for vision and they end up at
[3554s] like loggerheads. And I think, you know,
[3556s] that's not your job.
[3557s] >> I want to ask more about that, this art
[3559s] of pushing back and and, you know,
[3561s] nudging things in a
[3563s] direction. Is there kind of one trick or
[3565s] one tip you might share with folks to
[3567s] get to be good at cuz a lot of people
[3568s] deal with execs and they're always
[3569s] trying to you know
[3571s] get them to agree to what they want.
[3572s] >> I think the first thing is like don't
[3573s] treat it as a trick. I mean you're not
[3575s] trying to get
[3577s] an answer and a yes. You're trying to
[3579s] kind of seek truth is like my first
[3580s] statement. And so
[3583s] I wouldn't say it's especially common
[3585s] that we start out of alignment, but I do
[3587s] think
[3588s] you know, part of your role if you're
[3589s] one of the more senior product folks in
[3590s] the room and the the CEO or you know, if
[3592s] you're a director and and it's the CTO
[3594s] in the room is like pushing the team in
[3596s] a way that you don't really expect or
[3597s] you don't really understand is like
[3598s] start from a place of curiosity.
[3601s] Does that person have more context than
[3602s] you or is there a thing that you're not
[3604s] aware of that is guiding it? So I often
[3605s] try and start with like can you can am I
[3608s] hearing you right that this is your
[3610s] prior?
[3612s] Is there a you know, piece of context or
[3614s] something that I don't have that helps
[3615s] inform that prior? Cool. Make sure
[3617s] everybody's on that same baseline and
[3619s] you know, I coach my PMs to do this with
[3620s] me all the time too. Like if I'm coming
[3622s] at you from an angle you don't expect
[3623s] pause and make sure that you understand
[3625s] why.
[3626s] And then after that you know, I try and
[3629s] do a couple different things. You're
[3630s] obviously just
[3631s] people are people, right? They have good
[3632s] days. They have bad days.
[3634s] Sometimes it can be as simple as like is
[3635s] your mind made up on this or are you
[3637s] open to input?
[3638s] If the answer is like no, I'm pretty set
[3640s] that this is the answer shut up. Like
[3643s] don't pick a fight for the purposes of
[3644s] picking a fight. If the answer is like
[3646s] actually yes, you're welcome to push me
[3649s] but I'd need to see data.
[3651s] If I don't have any data in the room, if
[3652s] it's just my opinion then I've the terms
[3654s] of the debate are pretty clear.
[3656s] If I have fresh data, bring it. Uh if I
[3659s] really believe a thing
[3660s] and I don't have data, question why.
[3662s] Um or go get it and then go back to them
[3665s] with the data. But if the answer is like
[3667s] if it nobody's got data, it's just two
[3669s] opinions, the CEO's opinion is going to
[3671s] win.
[3672s] And like that's okay. Check your ego at
[3674s] the door, get to the answer, but if
[3676s] you've got data, bring it.
[3678s] And then sometimes it's like just make
[3680s] sure that you're debating for the right
[3681s] reasons. I think it's, you know, easy in
[3683s] your PM theater that you're describing
[3684s] of like wanting to make sure that you
[3686s] set the framing and the tone.
[3688s] And there are genuinely times in our
[3690s] industry where like nomenclature matters
[3693s] and exactly what word is used can be
[3695s] really important.
[3697s] But a lot of the time it doesn't.
[3699s] >> There's a concept that I heard
[3701s] uh come up a bunch when I talk to people
[3702s] who work with you.
[3703s] Uh the phrase was uh play the accordion.
[3706s] >> [laughter]
[3707s] >> Explain.
[3708s] >> So, I mentioned earlier I'm not huge on
[3710s] frameworks, but I do think uh there's a
[3711s] mental model which is quite useful that
[3713s] goes to the thing that you and I
[3715s] discussed earlier on like uh
[3717s] think through a problem
[3719s] mental do the mental work even if you're
[3721s] not shipping at all, which is like
[3723s] obviously the magic of code is that you
[3725s] can kind of ship things and iterate
[3726s] really quickly and get data as you go.
[3728s] And I think one of the the trappings
[3730s] that comes with that is uh you're just
[3732s] iterating forward and throwing spaghetti
[3734s] at the wall and you don't really know
[3735s] what direction you're going in.
[3736s] I think there's a second failure mode,
[3738s] which is people sit down and write out
[3739s] these long road maps and strategic
[3742s] vision docs of what we'll do over the
[3743s] next two, three years that kind of loses
[3746s] the comparative advantage we have over
[3748s] every other industry, which is learning.
[3750s] Like AB tests mean that you can update
[3753s] your understanding of a problem and
[3754s] therefore change your direction as you
[3756s] go.
[3757s] So, uh the point of the analogy of play
[3759s] the accordion is if you think about like
[3760s] a piano accordion
[3761s] before you can play a note you got to
[3763s] stretch it all the way out, bring the
[3764s] air in. So, like what is it we're trying
[3765s] to get done here?
[3767s] But, you don't make music until you
[3769s] press the key and push it all the way
[3770s] back into V1.
[3771s] And then when you go to play the the
[3773s] next kind of like progression you pull
[3775s] it all the way out again. Okay, given
[3777s] what we just learned
[3779s] what do we do?
[3781s] And then you go and build the next thing
[3782s] again. And like you've got to get used
[3783s] to this motion that says like this isn't
[3786s] creating value. Like pulling it out
[3788s] doesn't actually do anything.
[3790s] Uh all the value is created here, but
[3792s] until you are going through this motion
[3793s] constantly of saying we'll re-evaluate
[3795s] what we understand, how does this change
[3797s] what we're doing? You're probably not
[3799s] playing the right thing.
[3800s] And I know I'm probably bastardizing
[3802s] somewhere in the in the comments there's
[3804s] going to be a musician who points out to
[3805s] me that this is actually also one of the
[3806s] ways you make music, but I found it just
[3808s] a generally very valuable like memory
[3811s] device for PMs to be like you've got to
[3813s] constantly be zooming out and pushing
[3815s] back in, zooming out and and pushing
[3817s] back in.
[3818s] >> Yeah, that is
[3819s] >> [laughter]
[3819s] >> such a fun analogy cuz you know, it's
[3821s] like the even though you make music
[3822s] expanding it, it's like inward
[3825s] or it's like learning from the
[3828s] >> Yeah, the compression
[3829s] >> What are we learning? How do you
[3829s] communicate differently?
[3831s] Back to shipping.
[3832s] >> a different kind of music, internal
[3833s] music, external
[3834s] experiment.
[3835s] >> Yeah, cuz genuinely there are people who
[3837s] are really big on road maps and there
[3839s] are people who are really big on
[3839s] iteration. I I think the honest answer
[3841s] is you've got to do both. You can't
[3843s] over-index on either.
[3845s] >> So the lesson here is
[3846s] run experiments and but then make sure
[3849s] you think about the implication of the
[3851s] of the result at the bigger picture and
[3854s] then
[3854s] >> What what are we learning?
[3855s] >> Have a plan. Like
[3856s] >> system works this way. I I have this
[3858s] belief that if we change the way that
[3860s] discovery algorithms work for XYZ
[3861s] reasons it will have this impact.
[3863s] What's the smallest thing I can do to to
[3865s] prove that? Okay, it worked. Great, no
[3868s] change to plan. V2.
[3870s] Oh that didn't work.
[3872s] V3 is going to have to be different and
[3874s] you just kind of want to always be going
[3875s] through that motion.
[3876s] >> People not watching on YouTube, Tom is
[3878s] moving his hands.
[3879s] >> Just gesticulating wildly.
[3881s] >> What's like
[3883s] what's like a context where you had said
[3885s] that to someone, hey, go play the
[3885s] accordion or we're playing the
[3887s] accordion. Like Like what are they
[3888s] usually doing wrong?
[3890s] >> It tends to be that you're shipping
[3891s] something with a in a very local sense
[3894s] without necessarily understanding
[3896s] impact. So
[3897s] an example right now is, you know, the
[3900s] core of most marketplaces is listings,
[3902s] right? Like if you try and think about
[3904s] amazon.com without listings, there's
[3905s] basically nothing there. It's a bunch of
[3907s] you know, it's a left rail or right rail
[3908s] and some videos.
[3910s] Um
[3911s] in video commerce and live commerce
[3913s] whatnot, you don't really historically
[3914s] need listings. If I want to sell you a
[3916s] pair of AirPods, I could literally hold
[3918s] them up to screen and show you them and
[3919s] describe them and say they're AirPods.
[3922s] I'm going to start them at a dollar and
[3923s] as a buyer, you now have all the
[3925s] information you need in order to kind of
[3928s] like make a purchase decision, which is
[3930s] great.
[3931s] And so, you may not have to invest in
[3933s] making listings the way that someone
[3935s] else does and that's probably net good
[3937s] for a seller because it takes like 3 and
[3939s] 1/2 minutes to make a listing.
[3941s] Uh and it takes 0 minutes to describe a
[3944s] thing and hold it up.
[3945s] Uh but you then zoom out and say, "Oh,
[3948s]  New buyers are going to come to
[3950s] live commerce and expect search to
[3952s] function.
[3954s] If I don't know what you're selling
[3955s] until after you've sold it, there's no
[3957s] possible way that I can put somebody in
[3959s] the right stream for AirPods because
[3961s] they didn't we didn't know you have them
[3963s] and I therefore can't direct people to
[3965s] it. And so, you go through this exercise
[3966s] of like we don't need to fix it, it
[3968s] works. And then you're like, "Okay,
[3969s] great. What does that have implications
[3971s] for other things?
[3972s] Okay. Then what would we do? Oh, we're
[3975s] going to make everyone make listings.
[3976s] And you're like, zoom back out again. If
[3977s] every seller is now spending 3 minutes
[3979s] for every listing that they're making,
[3981s] the number of things they can sell per
[3983s] hour is now dramatically, dramatically
[3985s] lower, which means it's bad for sellers.
[3988s] Okay, We can't do that again. And
[3989s] so, you just you have to keep stretching
[3991s] through the like what are the
[3992s] longer-term implications or what are the
[3994s] knock-on effects of things we're
[3995s] building.
[3995s] >> Uh that is extremely helpful. What's
[3997s] interesting about Whatnot is it's
[4000s] there's like this spectrum of like
[4001s] Whatnot and then there's this whole
[4003s] trend of agentic commerce where agents
[4006s] are going to be buying the things for
[4007s] you and just working with each other,
[4008s] create a whole new agentic economy. And
[4011s] this is like the opposite, humans live
[4013s] talking to each other, buying from each
[4014s] other. Uh thoughts on that on that trend
[4017s] and and how may how that might impact
[4019s] you guys?
[4020s] >> Uh listen, I for one welcome our agentic
[4022s] overlords for things like I don't want
[4025s] to have to think about light bulbs,
[4027s] air filters, any of the programmatic
[4029s] stuff that I need to run my life, right?
[4032s] Um I'm also pretty great with it for
[4034s] like really high intent things. I need a
[4036s] particular cable for my computer. I'm
[4038s] looking for, you know, outdoor lights
[4040s] for my house. Things where there are
[4043s] really taxing searches.
[4046s] I'm thinking about this these you know,
[4047s] got to go to a wedding which means I
[4048s] need black shoes but delivered by
[4050s] Thursday cuz if they don't get here
[4051s] before Thursday, they're no good for me
[4052s] kind of things like absolutely.
[4055s] But most of commerce in America isn't
[4057s] actually high intent. Like how many
[4059s] years are we into e-commerce now? Like
[4061s] 30 years into e-commerce and e-commerce
[4064s] has never exceeded 20% of retail spend
[4066s] in America.
[4067s] Right? The vast vast vast majority of
[4070s] retail shopping in the United States is
[4071s] still people going in person and buying
[4074s] things. Uh in the UK, it's somewhere
[4075s] it's like 75 25.
[4078s] And it turns out actually that a lot of
[4080s] shopping is low intent.
[4083s] I'm going to the mall. I might be
[4085s] because I've got a wedding coming up and
[4087s] I don't have anything to wear and I'm
[4088s] going to wander around and see what
[4090s] people have available because I don't
[4091s] actually know specifically what I want.
[4094s] And actually the value of stores is that
[4097s] that person who runs that shoe store has
[4100s] agency and taste and she curates a
[4102s] selection of shoes. She has, you know, a
[4105s] level of customer service. She has a
[4107s] window display which can show me the
[4109s] types of things she has and I can wander
[4111s] around the mall and actually work out
[4113s] what it is I want to buy or be educated
[4115s] by it.
[4116s] And as it turns out, it's quite
[4117s] pleasant. Like there's a reason that the
[4119s] mall is a social thing. And so I think
[4123s] you know,
[4124s] that agentic is going to be huge.
[4126s] Retail is a seven and a half trillion
[4128s] dollar industry in the US though, so I
[4130s] don't think it's a like winner takes all
[4133s] thing. What I think live commerce does
[4135s] is it's the first time that we've ever
[4136s] managed to kind of like bring together
[4138s] scale and convenience of the internet
[4140s] and also that same kind of like social
[4142s] cultural experience
[4144s] all of physical shopping because
[4147s] you know
[4148s] at at Twitch we used to basically assume
[4149s] that if it's less than a thousand people
[4151s] on the stream it's non-economic cuz
[4152s] you're you're operating on CPMs.
[4155s] So but you can go to a stream on whatnot
[4156s] and see 30 50 people are in that stream
[4158s] and you imagine for a moment that you're
[4160s] running a shoe store at a mall and you
[4162s] had 50 people in your store you'd never
[4164s] close. Like you would literally never
[4167s] shut down the store because that's so
[4168s] much more foot traffic than you can ever
[4170s] imagine being in a store because
[4172s] commerce has totally different economics
[4174s] to entertainment and CPMs aren't the
[4176s] thing that you have to worry about. So I
[4178s] think it's not really in kind of like
[4180s] competition with a gentle I think it's a
[4181s] completely different kind of like
[4184s] customer need.
[4185s] >> Amazing. There's room for everyone. I
[4187s] want to end maybe with a a question
[4189s] about Twitter. So you were at Twitter
[4191s] PM at Twitter working on growth.
[4194s] I feel like everybody that worked at
[4195s] Twitter as a PM was just scarred from
[4197s] the experience of Twitter. Everyone's
[4200s] do not do things this way. What's what's
[4202s] something that stuck with you from that
[4203s] experience? What did you learn? What did
[4204s] you unlearn?
[4205s] >> It is funny. I uh
[4207s] I have said to a friend before that
[4209s] asking someone who was at Twitter in
[4210s] like the 2015-16 era is a little bit
[4212s] like a therapist asking somebody to tell
[4214s] them about their childhood. It's like
[4215s] you know the trauma that you're you're
[4216s] bringing up. Yeah. For those who not
[4219s] familiar with that lore I think we had
[4220s] nine heads of product in the two years
[4222s] that I was there. It was that level of
[4223s] kind of chaos. I think my favorite quote
[4225s] was from someone on the partnership team
[4227s] who said it felt like chaos which had
[4228s] lived in the events. That was like how
[4230s] often drama was going on about the
[4232s] place.
[4233s] Um
[4234s] I think learning I probably took two
[4235s] overwhelming things away from from my
[4237s] time at Twitter and other than like some
[4239s] incredible friends and I will say that
[4241s] that the product diaspora from that era
[4243s] of Twitter is pretty incredible and and
[4245s] everywhere.
[4246s] The first one is like if you truly find
[4248s] product market fit like if you manage to
[4250s] bottle lightning doesn't matter how
[4252s] badly you screw up the organization uh
[4255s] of it like it's huge. And like
[4258s] Twitter genuinely had that level of
[4260s] product market fit that you could
[4261s] emotionally feel how much people loved
[4264s] your product. And it's a really powerful
[4265s] litmus test to learn relatively early in
[4267s] your career that that's what PMF is, not
[4269s] like
[4270s] "Hey, the graphs look okay." Right?
[4272s] There's like a level of fervor that goes
[4273s] in.
[4274s] Um
[4275s] I think maybe on the flip side of like
[4277s] less positive is like what I definitely
[4279s] learned is like most of the time you
[4280s] hear it's really complex.
[4282s] It isn't.
[4283s] Leadership's just weak.
[4285s] So,
[4286s] you know, the two years that I was
[4288s] there,
[4289s] everyone knew we were going to have to
[4291s] lift the 140 character limit, right?
[4293s] There was working group after working
[4295s] group, like the project beyond 140 was
[4297s] like everywhere, because we knew for
[4299s] example that like uh people in Japan
[4302s] tweeted six times more often than people
[4304s] uh in Western markets, and it was
[4306s] largely when we spoke to them because
[4307s] kanji let you say heaps more
[4310s] uh in the same number of characters than
[4312s] uh you know, a Romance language did.
[4314s] We knew that was the inevitable end
[4315s] state, but there's obviously a bunch of
[4317s] like work that was needed and
[4319s] trade-offs, and just no one wanted to
[4320s] make the call, and so there was just
[4322s] like yet another design sprint, yet
[4324s] another cycle.
[4325s] And it was another year and a half after
[4327s] I left. I think it was maybe maybe even
[4329s] almost two years after I left before
[4330s] someone actually did it.
[4332s] And it turns out, you know, nobody died.
[4335s] The the soul of the place didn't fall
[4336s] apart. I imagine the same discussion
[4338s] went on for editing tweets, which took
[4340s] another two and a half years. Like,
[4341s] sometimes things aren't actually that
[4343s] complicated. It's just weak leadership.
[4346s] >> Yeah, it's funny how far it's come. Now
[4347s] you can write an entire blog post on
[4349s] Twitter, like articles. It's a big bet.
[4352s] On the product market fit piece, I think
[4353s] even more important there is the network
[4354s] effects of a Twitter, which you spent a
[4357s] lot of time thinking about building
[4358s] marketplaces. Like to me, watching Elon
[4360s] basically change everything. I forget
[4362s] who tweeted this, but just like
[4364s] everything changed. The brand, the name,
[4365s] the website,
[4366s] >> uh number of people working there,
[4368s] everything.
[4368s] >> people, the team, like what what was the
[4371s] thing that that stayed the same? and it
[4372s] was basically the network effects of
[4374s] Twitter.
[4374s] >> Yep.
[4375s] >> It's the place to be so everyone's there
[4377s] and that's hard to break.
[4378s] >> And even in spite of how much you tried
[4380s] to mess it all up
[4381s] it's still kicking.
[4383s] That's it.
[4384s] >> If you bought a lightning genuinely you
[4385s] can tell.
[4386s] >> Yeah. Okay, I'm going to take us to a
[4388s] recurring corner of the podcast, fail
[4390s] corner.
[4392s] Why I like doing this is because people
[4394s] see you see see people like you coming
[4396s] on the podcast have the have this
[4398s] illustrious career everything's going
[4400s] great constantly you're just killing it
[4403s] and they don't see the things that don't
[4405s] work out and the times that you failed
[4408s] and in their life things often go wrong.
[4410s] So So the question for you is just
[4412s] what's an example of a time in your
[4413s] career where things failed something you
[4415s] built some career movie made that didn't
[4417s] work out and then what what did you
[4418s] learn from that experience?
[4420s] >> Honestly mate and it's very nice of you
[4422s] to say nice things but I feel like I
[4423s] fail more often than I succeed across
[4425s] the course of my career. Genuinely to
[4427s] the to the blog you referenced right at
[4429s] the start
[4430s] I published in the in the back of that
[4431s] the actual document we use internally to
[4433s] talk about how we build and it starts
[4435s] with like batting 500 is like the goal.
[4438s] So like you're hoping to be right as
[4439s] often as you're wrong.
[4441s] So there's probably just too many
[4442s] specific examples of of times I've
[4444s] screwed up in my career but there
[4445s] probably is a really common thread to it
[4447s] and I think it's probably an easy trap
[4449s] for for any PM to fall into which is
[4451s] like
[4452s] averages mean nothing to the individual
[4455s] is probably the thing that I've like
[4456s] really scarred by.
[4458s] In any sizable population it's really
[4461s] attractive to go and look at like
[4462s] average utility or average adoption of
[4464s] something and then you find that like
[4466s] you know only 3% of people use
[4468s] something. And you're like cool we can
[4470s] probably get rid of that feature it's
[4471s] not used widely.
[4473s] But if you don't go a layer deeper and
[4474s] be like actually four like you know it's
[4476s] only 3% of something but there's a group
[4478s] of people for whom it's 100% of what
[4480s] they do. This is their core use case and
[4483s] for expediency sake because somebody
[4484s] doesn't want to maintain a feature
[4485s] anymore you're just going to deprecate
[4487s] it.
[4488s] And then it turns out you like blow up
[4490s] the use case of that group of humans and
[4492s] then to your last point about network
[4493s] effects, the ongoing spiral effect of
[4496s] that can be enormous. You know, I I
[4499s] think about it a lot in e-commerce of
[4500s] like this is somebody's business.
[4502s] Right? If we're just not reliable or
[4505s] like deprecating a feature, it's kind of
[4506s] like a Westfield Mall just turning off
[4508s] the power in the lead-up to Christmas
[4510s] without thinking about it. And so like
[4512s] there are real downstream impacts to
[4513s] people's businesses that often come from
[4515s] just like a lack of nuance in
[4517s] understanding metrics, particularly
[4519s] averages. Like they they just lie to you
[4521s] all the time and I think
[4523s] you know, I've probably screwed up in
[4524s] all of the ways in my career, but most
[4526s] of the time I've made
[4528s] genuinely like I'm disappointed in
[4530s] myself levels of decisions, it's
[4531s] typically that I've relied on averages
[4533s] without thinking about the individual
[4534s] use cases that that are hidden
[4535s] underneath.
[4536s] >> Makes me think about Jeff Bezos has a
[4538s] quote, "When you have data and an
[4539s] anecdote, trust the anecdote."
[4541s] >> Yep.
[4543s] That's exactly right.
[4545s] >> Well, Tom, we've gone through everything
[4548s] I wanted to talk about. Is there
[4549s] anything else that you wanted to share,
[4551s] anything you want to leave listeners
[4553s] with before we get to our very exciting
[4554s] lightning round?
[4555s] >> I think all I'd add, Mike, I've really
[4557s] enjoyed the discussion, so thank you, is
[4558s] like
[4559s] I don't think there is one way to to do
[4561s] product management and I don't think
[4562s] there is one way that AI will shape the
[4564s] industry. So like we're pretty confident
[4566s] that for Whatnot, the product we're
[4567s] building and the culture of the company
[4569s] that we have, that you know, this model
[4571s] of like fewer PMs who are more senior a
[4575s] lot more autonomy is right for us. Uh I
[4577s] don't pretend to, you know, presume that
[4579s] that will be true for the entire
[4581s] industry, but I do think there has never
[4582s] been a better time to go back to the
[4585s] roots of like the actual product work
[4586s] and getting out of that theater and I
[4588s] think that probably is true everywhere.
[4591s] Um even if you are still, you know,
[4593s] there are people who are wonderful
[4594s] people managers and really derive their
[4596s] satisfaction from doing that and growing
[4597s] and coaching and I'm sure there'll be
[4598s] loads of places where that's still
[4600s] valuable. So uh assume that at least
[4602s] half of what I've said is wrong in the
[4604s] same basis that half of the things I've
[4605s] probably ever shipped are not correct.
[4607s] >> That's why I love these conversations
[4608s] and why I think this work that we do
[4610s] together is important is we are living
[4613s] through the wildest time in our careers.
[4615s] So much is changing. So much is being
[4617s] rethought as you've described and it
[4619s] feels like the only way we can make our
[4621s] way through this successfully is to
[4623s] learn from how other people are
[4625s] approaching it, see what they've
[4626s] learned, see what they've has not
[4627s] worked.
[4627s] >> Take bits that resonate, ignore the bits
[4629s] that seem bombastic or not applicable.
[4632s] >> Exactly, cuz like no one knows exactly
[4634s] where
[4635s] Elizabeth Stone had this great way of
[4636s] putting it. We're in this kind of
[4637s] there's there's a storming phase and a
[4639s] norming phase and we're in the storming
[4640s] phase of holy Like I remember in
[4643s] my PM career as I started writing and
[4645s] stuff everyone was always asking me,
[4646s] "How has product management changed in
[4648s] the last decade?" I'm like, "It hasn't
[4649s] changed. It's the same basically." But
[4651s] it feels like now it actually has
[4652s] significantly changed.
[4654s] >> Although to your point earlier, actually
[4656s] the core thing
[4657s] will be the same.
[4659s] >> Yeah.
[4659s] >> Yeah.
[4660s] >> Yeah. So So that's why these are so
[4662s] useful just for people to see. Here's
[4663s] how a team is operating and what they've
[4665s] learned and here's things to try and it
[4667s] may not work for you.
[4668s] But this is how we learn from each
[4669s] other.
[4670s] With that, we have reached a very
[4672s] exciting lightning round. I've got five
[4674s] questions for you. All right, here we
[4675s] go. What are two or three books that you
[4678s] find yourself recommending most to other
[4680s] people?
[4681s] >> Okay, uh, not to be clichéd, but The
[4683s] Hard Thing About Hard Things I still
[4684s] think is the best book written about
[4685s] product management. Um, just from a like
[4688s] breadth of things that you have to go
[4690s] through and try and do and screw up a
[4692s] lot.
[4693s] Slightly left field, um, but the best
[4696s] book anyone's ever recommended to me to
[4697s] read which I now recommend to others,
[4699s] um, is The Purpose Driven Church
[4701s] by a pastor called Rick Warren. Uh,
[4703s] Emmett Shear, the CEO at at Twitch used
[4705s] to basically make sure that people read
[4706s] it. It's this wild examination of why
[4709s] people emotionally invest in something
[4711s] and how to engineer emotional investment
[4713s] into it from a group of people.
[4715s] Uh, Uh it's literally a like how to go
[4716s] and build a church
[4718s] uh, guidebook written in the '90s.
[4720s] Definitely worth a read if you're in a
[4722s] community product of any form.
[4724s] Uh and the last one is if you're looking
[4726s] for kind of like fiction or fantasy,
[4727s] which is where I tend to go in the
[4729s] evening, Babel by R.F. Kuang.
[4731s] >> 11 books have never been mentioned
[4733s] before get added to the canon of
[4734s] recommended books.
[4735s] >> It's very fun.
[4737s] >> Favorite recent movie or TV show that
[4740s] you've really enjoyed?
[4741s] >> Uh Star City on Apple TV.
[4744s] Uh if you liked For All Mankind, it's
[4746s] kind of like the flip of that, but it's
[4747s] the Soviet side. It's like watching the
[4749s] Americans and For All Mankind mixed
[4752s] together.
[4753s] >> Favorite product that you have recently
[4755s] discovered that you really like?
[4757s] >> Uh this is a very deep cut. So I will
[4759s] apologize to most listeners. Um
[4762s] uh hopefully you've detected the accent.
[4764s] I'm told constantly that my Australian
[4765s] accent is going, but one of the things
[4767s] that I love is every time I go home, I
[4769s] realize actually a bunch of the
[4771s] government services apps in Australia
[4772s] have become phenomenal. You ever had
[4774s] that kind of concept where you're like,
[4775s] I wish the government has all my data,
[4776s] but I wish there was just like one place
[4778s] where I could with one click get my
[4779s] driver's license renewed, I could like
[4781s] transfer titles, and do all of the admin
[4783s] that slows you down in life? Services
[4785s] New South Wales actually nailed bizarre
[4787s] to me that I would ever come on a
[4789s] podcast and say, actually a
[4790s] government-run app in Australia, of all
[4792s] places, is it. But I was home recently
[4794s] and had to do all of my life admin, and
[4796s] it's incredible.
[4798s] >> Wow.
[4799s] Something I heard recently about
[4800s] Australia, while we're on that topic
[4801s] real quick tangent from lightning round,
[4802s] is
[4803s] with a solar panel buildout that has
[4805s] happened there, there's more
[4808s] electricity available in Australia than
[4810s] they can use.
[4811s] >> Yep. Solar Solar is real huge in
[4813s] Australia.
[4814s] >> So it's like
[4815s] they're giving people free electricity
[4816s] in the middle of the day cuz there's so
[4817s] much available, and they're like, use
[4819s] all your stuff in the middle of the day
[4820s] cuz this is otherwise it's going to go
[4822s] to waste.
[4822s] >> There is a pitch somewhere that says if
[4824s] AI actually needs a loads of electricity
[4826s] in order to feed data centers,
[4827s] Australia's entire 21st century economy
[4829s] should be power.
[4830s] >> Incredible. That's just like such good
[4832s] news that we are finding ways to
[4834s] generate so much energy from solar
[4836s] panels.
[4837s] >> A small piece of regulation in what 20
[4839s] years ago that said if you're building a
[4840s] new property you got to put solar panels
[4841s] on the roof and it turns out it works
[4843s] great.
[4843s] >> Oh my god. I love this. I love this
[4845s] optimism of the future because you know
[4847s] climate energy
[4847s] >> Technology is the way through. Not not
[4849s] the problem.
[4850s] >> Mhm.
[4851s] Here, here.
[4852s] Okay, two more questions. Do you have a
[4854s] favorite life motto that you often come
[4856s] back to in work or in life?
[4858s] >> In my uni days or back in college for
[4861s] American translation
[4863s] I used to have a party trick where I'd
[4864s] memorized if by Rudyard Kipling cuz I
[4865s] thought it was like deep and really
[4866s] meaningful.
[4868s] Um, but I think if if I'm being really
[4870s] honest I'll figure it out is probably
[4873s] the closest.
[4874s] You know, like it turns out most things
[4876s] are not as hard as people think like
[4877s] we'll work it out. I'll figure it out.
[4879s] If you're willing to devote the required
[4882s] time, money, effort, energy, you can
[4884s] solve almost anything.
[4885s] And if you're not then it's probably not
[4887s] that big of a problem.
[4888s] >> Final question. I imagine people ask you
[4890s] this a lot but I'm also just curious.
[4892s] What's something you bought on Whatnot
[4894s] in the past month or so that was just
[4896s] awesome, delightful, surprising?
[4898s] >> The most fun thing I bought recently I'm
[4900s] not joking is a live lobster.
[4902s] So
[4904s] >> Was that a hook?
[4904s] >> We've recently one of the things that's
[4906s] really taken off on Whatnot is our like
[4908s] fresh and specialty foods kind of
[4910s] category and so there's this wonderful
[4912s] seller who goes by e-fish-co who has a
[4915s] seafood store down on the dock in San
[4916s] Diego and every morning as the boats
[4918s] come in he literally goes out and live
[4920s] streams all of the crates of seafood
[4922s] coming in off the boats and then he
[4923s] auctions them off live on Whatnot and
[4925s] and ships next day to your door. So
[4928s] I got a California spiny tail lobster
[4930s] shipped direct to my door
[4932s] courtesy of a live stream.
[4933s] >> And how's this work? They put in ice
[4935s] ship it next day same day?
[4936s] >> and ice kind of like container and it's
[4938s] UPS overnight on my door the next day.
[4941s] And that is why I think e-commerce is
[4943s] going to be great but not all
[4944s] encompassing because I had no intention
[4946s] of buying a spiny California lobster
[4948s] that morning.
[4949s] >> And my free agent decides he needs a
[4950s] lobster today and we
[4951s] >> I got to be honest with you, it was
[4952s] delicious.
[4954s] >> Amazing. I didn't know that you could
[4955s] buy stuff like that. Tom, where can
[4957s] folks find you online if they want to
[4958s] follow your writing / hiring? Talk about
[4962s] what you're hiring for and finally, how
[4964s] can listeners be useful to you?
[4965s] >> If you're looking for me online, I'm TD
[4967s] Robo, t d r o b b o on Twitter, which is
[4970s] or X, I guess,
[4971s] which is where I do most of my musing
[4974s] and some mix of product management and
[4976s] yelling about Warriors games, so
[4977s] apologies in advance.
[4979s] And then on LinkedIn is the other place
[4980s] I've got most of my kind of like work
[4982s] writing. I will say it's
[4984s] I try for quality over volume, so don't
[4986s] expect daily daily drops from me on
[4988s] either.
[4989s] >> Just bang those. Just bang those once a
[4990s] year.
[4991s] >> That was the nicest thing anyone said
[4992s] about me in ages.
[4993s] Just just dropping casual bang those.
[4995s] And how can folks be be helpful?
[4997s] Honestly, would always welcome feedback
[5000s] around what not and how people are
[5001s] finding it and what more we can do
[5002s] better. So like hit me up on either of
[5004s] those platforms with, you know, hot
[5005s] takes, feedback or thoughts.
[5008s] >> And then you're hiring PMs even spite of
[5010s] the many strong opinions about product
[5011s] management. Maybe just talk about that
[5012s] and where folks can apply.
[5014s] >> Sure. Listen, we are constantly looking
[5016s] for PMs. The the intent of kind of like
[5018s] explaining how many people applied
[5019s] versus hired was not to discourage
[5021s] people. It was more along the lines of
[5023s] like the proliferation of product
[5025s] management doesn't itself naturally lend
[5028s] to the people with the skill set that
[5029s] you and I spent the better part of 90
[5030s] minutes talking about right now.
[5032s] But, you know, we hired two people
[5033s] yesterday. I'm very excited for them to
[5035s] come start
[5036s] and we've always got a variety of of PM
[5038s] roles open.
[5039s] The Whatnot, if you just kind of Google
[5041s] Whatnot jobs,
[5042s] the the right kind of application
[5044s] process will will pop up there. Or I've
[5046s] found that, you know, the valley is
[5047s] small enough and product management is
[5049s] not that kind of obscure that you can
[5050s] probably find one of the 20 or so humans
[5052s] who work at Whatnot and and reach out to
[5054s] them. But we're looking at a variety of
[5055s] things right now. Payments very high on
[5058s] on our list of what we're doing as well
[5059s] as logistics. So if anyone is really
[5061s] excited to come work on the future of
[5063s] shipping,
[5064s] lobsters overnight,
[5066s] it turns out there's quite a lot of
[5067s] nuance product we have to happen there.
[5068s] >> Amazing. Tom, thank you so much for
[5070s] being here.
[5071s] >> It was a pleasure, mate. Thanks for
[5072s] having me.
[5073s] >> Bye, everyone.
[5074s] Thank you so much for listening. If you
[5076s] found this valuable, you can subscribe
[5078s] to the show on Apple Podcasts, Spotify,
[5080s] or your favorite podcast app.
[5082s] Also, please consider giving us a rating
[5084s] or leaving a review as that really helps
[5086s] other listeners find the podcast. You
[5089s] can find all past episodes or learn more
[5091s] about the show at lennyspodcast.com.
[5092s] [music]
[5094s] See you in the next episode.