Adriel Frederick (Reddit, Lyft, Facebook)20 October 2022

Humanizing product development

7Frameworks

15Insights

Episode overview

Lenny Rachitsky interviews product leader Adriel Frederick about lessons from building growth, marketplace, pricing, and experimental products at Facebook, Lyft, and Reddit. Frederick argues that algorithm-heavy products still need people to establish intent, constraints, strategic judgment, and operational control. He also explains how diverse teams improve global product development, why leaders should stay close to users during controversy, and why durable growth comes from removing friction in core actions rather than chasing isolated hacks.

Key ideas

Product managers should explicitly decide what algorithms control, what people control, and which decision framework connects them.
Machine learning is best used to amplify human intent; people must still determine strategic objectives, constraints, and acceptable long-term effects.
Operational control should be treated as a first-order product requirement in marketplaces where local conditions and human judgment matter.
Globally diverse teams can identify assumptions and resolve product questions faster because their lived experience better reflects worldwide users.
During public controversy, product leaders should stabilize their teams, distinguish legitimate user problems from power struggles, and investigate firsthand.
Sustainable growth comes from repeatedly simplifying core actions such as discovering the product, registering, finding friends, and returning—not from growth hacks alone.
Data identifies where a marginal user's experience is failing, but observation and direct conversation reveal why it is failing.
As leaders become more senior, organization design and empathy become more important than individual technical contribution.

Transcript available · source text is retained privately and is not published

Frameworks in this episode

Leadership6 steps

Criticism Triage Under Public Controversy

Separate valid criticism from power-shift noise by going and being your own user.

Read →

Innovation6 steps

Human-in-the-Loop Algorithm Design

Decide explicitly what the algorithm owns, what the human owns, and how the handoff works.

Read →

Influence5 steps

Take Your Own Shoes Off

Empathy fails at the removal step, not the entry step — drop your agenda before entering theirs.

Read →

Strategy5 steps

The Cannonball / Lead Bullet Portfolio

Allocate growth energy between a few fundamental rebuilds and many small experiments — by product stage.

Read →

Leadership4 steps

The Galvanizing Activation Metric

The activation number's job is organizational momentum, not statistical precision.

Read →

Strategy5 steps

The Marginal User Method

Find the user on the cusp of converting, then go watch the worst-case version of them.

Read →

Leadership4 steps

The Organ-Rejection Test for Innovation Teams

Three conditions that stop the core org from rejecting your new-bets team like a foreign body.

Read →

People & resources mentioned

Attributed to the moment in the episode. Timestamps are approximate.

People · 11

Adriel FrederickMentionstoday my guest is Adriel Frederick Adriel is a VPR product at Reddit refocuses on incubating and scaling new products within Reddit
“today my guest is Adriel Frederick”
▶ Watch at ~0:30approx.
Lenny RachitskyMentionstoday we've got another very special compilation episode something I've been pulling on more and more with the podcast and the newsletter
“welcome to Lenny's podcast I'm Lenny”
▶ Watch at ~0:30approx.
Jules WalterMentionsJules is a product leader at YouTube. Before that, he spent 4 and 1/2 years at Slack, where he was their first growth PM
“I actually found out about you through a guy named Jules Walter who we both know he's a PM at YouTube”
▶ Watch at ~4:00approx.
Adriel FrederickMentionstoday my guest is Adriel Frederick Adriel is a VPR product at Reddit refocuses on incubating and scaling new products within Reddit
“I'm the vice president of product management for Reddit X”
▶ Watch at ~6:00approx.
Chamath PalihapitiyaMentionsthe original leadership of the growth team I think starting with chamat
“the original leadership of the growth team I think starting with chamat”
▶ Watch at ~16:30approx.
Elon MuskMentionsWhether you like him or not, Elon, this is similar to his um if you watch him, he has sets these really ambitious goals
“the way but why guy described it as elon's worried that AI will take over at some point”
▶ Watch at ~39:30approx.
Sam HarrisMentionsSam Harris talks about this a lot that it feels like once you get close then it could just accelerate so quickly
“Sam Harris talks about this a lot that it feels like once you get close then it could just accelerate so quickly”
▶ Watch at ~41:30approx.
Tom AlisonMentionsTom Allison I think no is responsible for Facebook app
“one of my favorites is something Tom Allison”
▶ Watch at ~46:30approx.
Mark ZuckerbergMentionsit was sort of a random decision for Mark Zuckerberg to move from a uh monthly active to a daily active
“at some point Zuck just said 10 14 days go”
▶ Watch at ~50:00approx.
Daniel YerginCoinedthe new map by Daniel Jurgen
“the new map by Daniel Jurgen”
▶ Watch at ~1:03:00approx.
Malcolm GladwellCoinedmy Talisman person was Malcolm Gladwell because he's an incredibly captivating speaker he's a self-described introvert
“revisionist history with Malcolm Gladwell just gives you a different look into things”
▶ Watch at ~1:03:30approx.

Resources · 27

Lenny's PodcastCoinedpodcast · Lenny Rachitsky
“welcome to Lenny's podcast I'm Lenny”
▶ Watch at ~0:30approx.
RedditMentionswebsite · Steve Huffman, Alexis Ohanian, and Aaron Swartz
“Adriel is a VPR product at Reddit”
▶ Watch at ~0:30approx.
LinearRecommendssoftware · Linear
“visit linear.app Lenny to try a linear for free with your team and get 25 off when you upgrade”
▶ Watch at ~2:30approx.
FlatfileRecommendssoftware · Flatfile
“if you'd like to learn more or get started check out flat file at flatfile.com Lenny”
▶ Watch at ~3:30approx.
YouTubeMentionswebsite · Google
“he's a PM at YouTube”
▶ Watch at ~4:00approx.
Trinidad and TobagoMentionsplace
“I'm originally from Trinidad and Tobago an island in the Caribbean”
▶ Watch at ~5:00approx.
LyftMentionscompany
“then worked on Marketplace at Lyft so Rider pricing real-time driver incentives matching writers with drivers”
▶ Watch at ~5:30approx.
FacebookMentionscompany · Meta
“worked at a startup growth team at Facebook for four years working on user acquisition”
▶ Watch at ~5:30approx.
RedditMentionswebsite · Steve Huffman, Alexis Ohanian, and Aaron Swartz
“I'm the vice president of product management for Reddit X”
▶ Watch at ~6:00approx.
Reddit XMentionscompany · Reddit
“I'm the vice president of product management for Reddit X”
▶ Watch at ~6:00approx.
Reddit Avatar MarketplaceMentionsproduct · Reddit
“we have a an avatar Marketplace that we've been working on recently where creatas have been able to make art”
▶ Watch at ~7:30approx.
McKinsey & CompanyMentionscompany
“from the beginning of my career McKinsey to today at Reddit”
▶ Watch at ~17:30approx.
Prop 22Mentionsevent
“when there were all the pr was going on about EB size and prop 22”
▶ Watch at ~25:00approx.
EppoRecommendssoftware · Eppo
“check out EPO at getfo.com get eppo.com and 10x your experiment velocity”
▶ Watch at ~28:00approx.
NeuralinkMentionscompany · Elon Musk and co-founders
“it makes me think about a neural link and what elon's trying to do”
▶ Watch at ~39:30approx.
DALL-EMentionssoftware · OpenAI
“just like Dolly come in absolutely co-pilot”
▶ Watch at ~42:00approx.
GitHub CopilotMentionssoftware · GitHub
“just like Dolly come in absolutely co-pilot”
▶ Watch at ~42:00approx.
People You May KnowMentionsproduct · Facebook
“I thought about the way the people you may know algorithm worked”
▶ Watch at ~45:30approx.
The PrizeRecommendsbook · Daniel Yergin
“the prize and probably now the new map by Daniel Jurgen”
▶ Watch at ~1:03:00approx.
The New MapRecommendsbook · Daniel Yergin
“the prize and probably now the new map by Daniel Jurgen”
▶ Watch at ~1:03:00approx.
Revisionist HistoryRecommendspodcast · Malcolm Gladwell
“revisionist history with Malcolm Gladwell just gives you a different look into things”
▶ Watch at ~1:03:30approx.
HP AcademyRecommendspodcast
“there's this esoteric one called HP Academy that I'm into”
▶ Watch at ~1:03:30approx.
MoRecommendstv show
“last night I discovered Mo on Netflix”
▶ Watch at ~1:04:00approx.
NetflixUsescompany · Netflix
“last night I discovered Mo on Netflix”
▶ Watch at ~1:04:00approx.
Radiant NuclearRecommendscompany
“it's the team behind radiant nuclear”
▶ Watch at ~1:05:00approx.
LinkedInUseswebsite · LinkedIn
“you can find me on LinkedIn Adriel Frederick”
▶ Watch at ~1:06:00approx.
lennyspodcast.comRecommendswebsite · Lenny Rachitsky
“you can find all past episodes or learn more about the show at lennyspodcast.com”
▶ Watch at ~1:07:00approx.

Spot an error or want something removed? Request a correction or removal.

Insights & moments

The myth-busts, hot takes, explainers, and tools worth keeping.

Myth Buster· 1

Myth Buster43:30

Growth Hacks Can't Save a Product With No Real Value

Adriel pushes back on the idea that you can growth-hack your way to lasting success. Hacks help get initial traction, but assuming they alone will work is disrespectful to users, who eventually catch on. The real work at Facebook was mundane and hard: make the product easy to find, easy to get into, and stupid easy to find your friends.

Hacks give outsized short-term impact but can only produce a one-hit wonder
Assuming hacks alone will work disrespects users, who catch on eventually
Without fundamental value, growth disappears after a few months
The durable wins were grinding on: easy to find the product, easy to get in, stupid easy to find friends

“so if you don't have a product that's providing real fundamental value to people you can be a one-hit wonder”
Adriel Frederick · 43:30

“one make it easy to find the product number two make it easy to get into the product three stupid easy to find your friends”
Adriel Frederick · 44:30

#growth#growth-hacking#facebook#retention

Hot Take· 3

Hot Take16:00

Diverse Teams Build Global Products Way Faster

Adriel argues that a globally diverse team lets you resolve product debates in 15 minutes that would otherwise take two weeks of user recruiting, because the team already contains the perspectives of the world. He also makes the case that diversity only sticks when a company recognizes concrete business value from it, not when it's done for PC or social-pressure reasons.

A diverse team can settle a design argument in 15 minutes vs. two weeks recruiting a user panel
Your teams should "look like the world" so you can build products the world will use
Diversity sticks only once the company recognizes concrete business value from it
Value-recognition drives retention, which drives easier recruiting

“you gotta have your teams look like the world it just makes you so much faster”
Adriel Frederick · 17:00

“when you recognize that you get business value from it then it all of a sudden becomes something that you look out for”
Adriel Frederick · 17:30

#diversity#team-building#product-management#growth

Hot Take21:00

Some Controversy Is Necessary — The Skill Is Spotting Valid Criticism

Adriel takes the contrarian view that you can't meaningfully change the world without disrupting some invested pattern of behavior, which guarantees conflict and press. The hard part is separating legitimate criticism from complaints that are really just about a power source being threatened.

Meaningful influence requires changing a behavior pattern someone is invested in
Conflict makes the best press stories, so controversy is part of the game
The hard skill is knowing which criticism is valid vs. displaced power
Example: little sympathy for medallion owners, but real, legitimate complaints about driver-pay structure

“I don't think you're going to have any meaningful influence on the world without changing some pattern of behavior”
Adriel Frederick · 21:00

“the really hard part about dealing with this is understanding what criticism is valid and how much of it is just because a source of…”
Adriel Frederick · 21:30

#product-management#controversy#press#lyft

Hot Take40:30

AI Can Help You Write the Paper — It Can't Write It For You

On AI fears, Adriel says current systems still seem like tools that need human guidance; even cutting-edge text and image generation lack real decision-making and judgment. Using a book report as an example, AI can help you write faster once you've chosen the topics, but it can't choose the topics your background and interests make compelling.

Cutting-edge models can fool you into thinking they're near-human, but lack judgment
AI systems still seem like tools that need human guidance to be useful
AI can speed up the writing once you've decided the topics
It can't choose the topics your own background and interests make compelling

“there is a lack of decision making and judgment that I see coming out of them”
Adriel Frederick · 41:00

“you can help you write the paper faster for sure but it can't write the paper for you”
Adriel Frederick · 41:00

#ai#judgment#future

Explainer· 4

Explainer35:30

Algorithms Need People to Make the Judgment Calls

Adriel's central lesson across Facebook and Lyft: algorithms optimize a given objective brilliantly but don't understand long-term effects, human reactions, or your intent for the product. A PM's job on ML-heavy products is to decide what the algorithm owns, what people own, and to design the human-computer interface that lets a person's strategic intent be amplified.

Techno-utopians believe you can just feed all data to an algorithm and it does the right thing
Algorithms miss long-term effects, human responses, and product intent
The PM's job is to divide responsibility between machine and human and set the decision framework
Give people the information and tools to make the strategic calls; let the algorithm amplify that intent

“the algorithms don't understand a long-term effects often nor do they understand how people might respond to it nor do they understand your intent for…”
Adriel Frederick · 35:30

“when you are working on algorithmic heavy products your job is figuring out what the algorithm should be responsible for what people are responsible for”
Adriel Frederick · 36:00

#ai#machine-learning#product-management#decision-making

Explainer38:00

Software (Including ML) Is Just a Screwdriver

Adriel frames all software, including machine learning, as a tool like a screwdriver — the design question is how much you build into the tool versus how much you leave to the person. ML is amazing at optimizing an objective but can't understand the constraints and strategic choices, which must always be decided by a person and then amplified across millions of decisions.

All software, including ML, is just a tool like a screwdriver
The design craft is deciding how much goes in the tool vs. left to the person
ML optimizes objectives but can't set constraints or strategic choices
Make the human decision intuitive, then let the algorithm amplify it across millions of decisions

“all software in any form including ml is just a tool like a screwdriver”
Adriel Frederick · 38:00

#ai#machine-learning#product-design#tools

Explainer49:00

Facebook's "10 Friends" Milestone Wasn't Magic — It Was a Rallying Cry

Adriel demystifies Facebook's famous activation milestone: there was nothing magical about the specific number or timeframe. Its brilliance was being memorable and communicable — a discrete number, a discrete time, and a graph everyone could rally around to cut past academic debate and galvanize the org.

The specific number and deadline weren't magic — the design and communication were
A concrete number + timeframe + shared graph is organizationally galvanizing
Zuck cut through the academic debate by just saying "10, 14 days, go"
One funny downside: a coded retention limit at 10 friends caused a kink in the retention-vs-friends graph

“it's literally nothing magic about the number or the Deeds but basically it was a way of seeing like get people as many friends as…”
Adriel Frederick · 49:30

“at some point Zuck just said 10 14 days go”
Adriel Frederick · 50:00

#growth#facebook#activation#metrics

Explainer51:30

Optimize for the Marginal User — and Go Watch the Worst Case

Adriel defines the marginal user as the person right on the cusp of taking the action you want. To find them, look for a country with high traffic but terrible conversion, then go to the extreme worst case — a user on a feature phone on EDGE far from a data center — because that user reveals everything wrong with the product. Crucially, data only hints where the problem is; you must go watch and talk to people to find the real, often orthogonal, cause.

The marginal user is the person just on the cusp of taking the action you want
Find them by looking for high-traffic, low-conversion segments
The worst-case user surfaces the full list of what's broken (language, country detection, latency)
Data alone won't give the answer — a real problem is often orthogonal to the funnel; go watch and talk to users

“for me it's a person who is just on the cusp of taking the action you want to take”
Adriel Frederick · 51:30

“something I caution people against though is don't use the data alone to figure out who the marginal user is”
Adriel Frederick · 52:30

#growth#user-research#facebook#product-management

Story· 3

Story13:00

Why Growing Up in Trinidad Reshaped Facebook Registration

Adriel describes how his upbringing in Trinidad and Tobago gave him a perspective most of tech lacked: the assumption that one phone number equals one device equals one person is simply false in much of the world. Prepaid numbers change constantly, dual-SIM phones were common, and expensive phones were often shared. He built that knowledge into Facebook's user acquisition and registration design.

Most colleagues assumed one phone number + one device = one person
Prepaid users' numbers change often; dual-SIM and shared phones were common
This insight shaped how Facebook designed phone-number handling in registration
Diversity of lived experience directly produced better global product design

“people make an assumption that one phone number plus one device title one person and growing up in Trinidad I just knew that wasn't true”
Adriel Frederick · 13:30

#growth#facebook#product-design#diversity

Story23:30

Driving 15 Minutes for Rick: Feeling the Problem Yourself

To understand driver complaints on Lyft, Adriel picked up shifts himself. He drove 15 minutes unpaid to reach an 80-year-old rider named Rick whose trip was only two minutes, then realized Rick had called because he'd had too much to drink and feared breaking his hip. The experience showed him the real value of the ride and exposed how driver pickup time wasn't reflected in Lyft's pricing structure.

Adriel drove real Lyft shifts to feel driver pain points firsthand
He drove 15 unpaid minutes to reach Rick for a two-minute trip
Rick had genuinely high need — too drunk, afraid of breaking a hip
The trip revealed that pickup/deadhead time wasn't compensated or priced in

“so I went from wanting to curse Rick out for making me drive 15 minutes to come pick him up to feeling like all right…”
Adriel Frederick · 24:30

“hey I had a little bit too much to drink I'm worried about breaking my hip so that's why I call her right”
Adriel Frederick · 24:30

#lyft#user-research#pricing#story

Story28:00

The Most Stressful Lyft Project: Rebuilding a Pricing Algorithm

Adriel's most stressful memory was a sophisticated pricing model, built with revenue-management PhDs, that worked well in three cities but became nearly impossible to operate at scale — it took months just to change prices. The emotionally hard but simple answer was to own the failure publicly and rebuild it. The lesson: treat operational control as a first-order requirement, not an afterthought.

A textbook-grade pricing model worked in 3 cities but broke down operationally at ~100
Changing prices took months because operational flexibility wasn't a design requirement
They got seduced by the algorithmic "secret sauce" and ignored day-to-day control
The fix required admitting the mistake and rebuilding with operations as first-order

“the answer was like yo we gotta rebuild it”
Adriel Frederick · 29:30

“you have to think about operational requirements and operational control as a first order requirements”
Adriel Frederick · 30:30

#lyft#pricing#operations#product-management

Takeaway· 4

Takeaway08:30

How to Set Up an R&D Team Without "Organ Rejection"

Adriel explains why long-horizon innovation teams inside established companies often get culturally rejected by the rest of the org, and the three conditions that prevent it. The core idea is that the new team must feel integral to the mission rather than a privileged group playing off in a corner.

New teams risk "organ rejection" because they look so different from the rest of the organization
The rest of the company must see the work as core and critical to the mission, not a side project
It has to feel like everyone's success, not just the R&D team's win
Set up the work so other teams don't believe all innovation happens only on that team

“the rest of the company needs to see what you're doing as being core and critical to the mission”
Adriel Frederick · 09:00

“everyone should feel like they win”
Adriel Frederick · 09:30

#product-management#innovation#org-design#reddit

Takeaway20:00

A PM's Job Is to Buffer the Team From the Hype and the Hate

Working at companies constantly in the news (Facebook, Lyft, Reddit), Adriel says a PM must provide a damping effect that cuts both ways. When everyone thinks the work is amazing, pull them back to earth; when the press says you're the worst thing in the world, pull the team up and keep charging toward the mission.

A PM is a leader who provides a buffering/damping effect on the team
When something feels amazing, remind the team there's far more value still to deliver
When the press is brutal, reassure the team and keep charging forward
The effect must go both directions, not just in the bad times

“as a PM you are a leader you have to provide a buffering or damping effect on the team”
Adriel Frederick · 20:00

#product-management#leadership#controversy

Takeaway56:00

Run a Portfolio: Cannonballs vs. Lead Bullets

Adriel estimates roughly 60% of experiments succeed and 40% should be turned off, but warns of a hidden cost: fussing with easy small experiments crowds out the few big bets. He recommends treating experiments as a portfolio — weighting energy toward a couple of "cannonballs" and fewer "lead bullets" — and notes the right balance shifts with product maturity (early stage can be 100% cannonballs).

Roughly 60% of experiments succeed, 40% should be turned off
The hidden cost of small experiments is the big work you didn't do instead
Laziness pushes teams toward many small, easy wins that don't add up to something big
Weight the portfolio toward cannonballs; early-stage products can be nearly 100% cannonballs

“there's a laziness that can creep in where you're just finding a lot of little things because they're easier to come up with”
Adriel Frederick · 57:00

#experimentation#growth#prioritization#product-management

Takeaway59:30

The Skills That Matter Most as You Get Senior: Org Design and Empathy

Asked what he had to grow most moving from IC to VP, Adriel names organization design and empathy. He had to unlearn the idea that the smartest or best individual contributors should rise, and instead focus on building great teams and processes. The hardest part of empathy, he says, isn't getting into someone else's shoes — it's taking your own off first.

The two skills he grew most: organization design and empathy
Nothing meaningful gets done by a single person, despite "hustle porn" online
Org design means clear goals, safety, and smoother processes and interactions for the team
The hardest part of empathy is taking off your own shoes before stepping into someone else's

“the hardest part is taking my own shoes off”
Adriel Frederick · 1:01:30

#leadership#career#empathy#org-design

Key ideas

Frameworks in this episode

Criticism Triage Under Public Controversy

Human-in-the-Loop Algorithm Design

Take Your Own Shoes Off

The Cannonball / Lead Bullet Portfolio

The Galvanizing Activation Metric

The Marginal User Method

The Organ-Rejection Test for Innovation Teams

People & resources mentioned

People · 11

Resources · 27

Insights & moments

⚡Myth Buster· 1

Growth Hacks Can't Save a Product With No Real Value

◆Hot Take· 3

Diverse Teams Build Global Products Way Faster

Some Controversy Is Necessary — The Skill Is Spotting Valid Criticism

AI Can Help You Write the Paper — It Can't Write It For You

✶Explainer· 4

Algorithms Need People to Make the Judgment Calls

Software (Including ML) Is Just a Screwdriver

Facebook's "10 Friends" Milestone Wasn't Magic — It Was a Rallying Cry

Optimize for the Marginal User — and Go Watch the Worst Case

❝Story· 3

Why Growing Up in Trinidad Reshaped Facebook Registration

Driving 15 Minutes for Rick: Feeling the Problem Yourself

The Most Stressful Lyft Project: Rebuilding a Pricing Algorithm

▲Takeaway· 4

How to Set Up an R&D Team Without "Organ Rejection"

A PM's Job Is to Buffer the Team From the Hype and the Hate

Run a Portfolio: Cannonballs vs. Lead Bullets

The Skills That Matter Most as You Get Senior: Org Design and Empathy

Myth Buster· 1

Hot Take· 3

Explainer· 4

Story· 3

Takeaway· 4