LLenny's Podcast
← All episodes
Brian Tolkin (Head of Product at Opendoor, ex-Uber)04 August 2024

Lessons from scaling Uber and Opendoor

7Frameworks
15Insights

Episode overview

Brian Tolkin draws on his operations and product leadership at Uber and Opendoor to explain why technology and operations work best as mutually respected, tightly connected functions. He discusses product reviews, a pragmatic implementation of jobs-to-be-done, experimentation in low-volume environments, and finding the highest-leverage product opportunity amid noisy signals. Stories about launching Uber Pool in China, virtualizing Opendoor during COVID-19, competing and partnering with Zillow, and learning to remain calm under pressure ground the advice in operating experience.

Key ideas

  • Product and operations should function like two engines: local teams supply customer context and rapid experimentation, while product and engineering create scalable leverage.
  • Technology investment should concentrate on the few systems that create disproportionate business leverage, with explicit decisions about what will remain manual for now.
  • Products that touch the physical world need flexibility and fail-safes because people, schedules, GPS, homes, and other real-world inputs are not deterministic.
  • Product reviews should primarily improve the product, stay psychologically safe, combine leadership breadth with team depth, and usually keep the live discussion under ten people.
  • Jobs-to-be-done is most useful as a flexible customer-empathy lens that includes the customer's broader context, not as a rigid template applied to every problem.
  • Low-volume experimentation begins with power analysis; when conventional A/B tests are impractical, teams can triangulate with customer research, observational data, geographic comparisons, holdouts, and informed judgment.
  • Strong product judgment means finding the kernel of truth among many signals, recording ideas so contributors feel heard, and choosing what genuinely advances the customer.
  • Calm leadership improves outcomes under pressure; repeated exposure, reflection, and learning from other operators' stories help build that composure.

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

Frameworks in this episode

Strategy3 steps

Competition-Aware, Not Competition-Focused

Stay aware of rivals but anchor on the customer job, because the real competitor is the default old way

Read →
Innovation4 steps

Conviction-Building for Low-Volume Experiments

When you can't run a clean A/B test, stack alternative signals to raise conviction instead of faking precision

Read →
Leadership5 steps

Depth-Meets-Breadth Product Review

Run reviews to make the product better, not to interrogate — you bring breadth, the team brings depth

Read →
Peak Performance4 steps

Even-Keel Under Pressure

Don't reflect stress onto your team — mantras plus lived exposure keep your head clear when everything's on fire

Read →
Strategy5 steps

Iterate-Then-Automate Scaling Ladder

Let ops manually iterate a process until it breaks at scale, then hand exactly that broken step to technology

Read →
Productivity4 steps

Kernel-of-Truth Signal Triage

Product is finding the one thing that really matters in a sea of inputs — capture everything, then ruthlessly filter

Read →
Leadership3 steps

Product Operations Feedback Loop

A function that reports to Ops but sits with Product to close the two-way gap between HQ builders and field teams

Read →

People & resources mentioned

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

People · 7

  • 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

    “today my guest is Brian tolken”

    ▶ Watch at ~0:30approx.
  • Brian TolkinMentionsBrian is currently head of product and design at Open Door before that he spent nearly 5 years at Uber

    “Brian is currently head of product and design at Open Door before that he spent nearly 5 years at Uber”

    ▶ Watch at ~0:30approx.
  • Kayvon BeykpourMentionsKavon was the Beloved and longest tenured head of product at Twitter and also GM of the consumer business at Twitter

    “just a huge thank you to Kavon bakour for connecting us introducing us”

    ▶ Watch at ~2:00approx.
  • Casey WintersMentionswho's had the most influence on you in your career well we mentioned him before so I gotta bring him back up Casey Winters

    “a previous podcast guest shared in a newsletter post Casey winers”

    ▶ Watch at ~24:00approx.
  • Ben AffleckMentionsmentioned

    “wasn't there that Nike documentary recently with Ben affle”

    ▶ Watch at ~1:07:00approx.
  • Jeff HoldenMentionsat Uber our first estate product Officer Jeff Holden was someone who often talks about kind of fast forwarding to the Future

    “Jeff Holden who was the chief product officer at Uber back in the day”

    ▶ Watch at ~1:09:00approx.
  • Travis KalanickMentionsFormer Uber leader discussed as a visionary product leader during the company's early era.

    “was Travis involved in that interview”

    ▶ Watch at ~1:11:30approx.

Resources · 26

  • UberMentionscompany · Uber

    “you started your career in operations at Uber”

    ▶ Watch at ~2:00approx.
  • PendoRecommendssoftware · Pendo

    “visit pendo.io Lenny to create your free pendo account today”

    ▶ Watch at ~4:30approx.
  • AirbnbMentionscompany · Brian Chesky, Joe Gebbia, and Nathan Blecharczyk

    “this was definitely true at Airbnb where the product team kind of looks down a little bit on the Ops Team”

    ▶ Watch at ~6:00approx.
  • ExploRecommendssoftware

    “try it for free at expo. co/ Lenny”

    ▶ Watch at ~6:00approx.
  • GoogleMentionscompany · Google

    “at the time I believe Google had had a function”

    ▶ Watch at ~9:00approx.
  • ChengduMentionsplace

    “were in changu China which is the the first Chinese market that we were launching Liverpool in”

    ▶ Watch at ~13:30approx.
  • OpendoorMentionscompany

    “we're digital platform to buy and sell real estate”

    ▶ Watch at ~17:00approx.
  • ReplicateMentionscompany

    “join modal replicate 11 labs and more”

    ▶ Watch at ~31:30approx.
  • AttioRecommendssoftware · Attio

    “head to ato.com Lenny and you'll get 15% off your first GE”

    ▶ Watch at ~31:30approx.
  • ModalMentionscompany

    “join modal replicate 11 labs and more”

    ▶ Watch at ~31:30approx.
  • ElevenLabsMentionscompany · ElevenLabs

    “join modal replicate 11 labs and more”

    ▶ Watch at ~31:30approx.
  • Stripe AtlasMentionsproduct · Stripe

    “with Jeff Weinstein from stripe who's building stripe Atlas”

    ▶ Watch at ~52:00approx.
  • AngelListMentionscompany · Naval Ravikant and Babak Nivi

    “they had a similar experience with angelist launch direct competitor to Angel to Atlas”

    ▶ Watch at ~52:00approx.
  • StripeMentionscompany · Patrick Collison and John Collison

    “with Jeff Weinstein from stripe who's building stripe Atlas”

    ▶ Watch at ~52:00approx.
  • FoundersRecommendspodcast

    “one of the podcasts that I I love it Founders podcast”

    ▶ Watch at ~55:30approx.
  • Drive to SurviveRecommendsdocumentary

    “so full swing draft to survive break points tennis golf F1 all of them”

    ▶ Watch at ~1:06:30approx.
  • Full SwingRecommendsdocumentary

    “so full swing draft to survive break points tennis golf F1 all of them”

    ▶ Watch at ~1:06:30approx.
  • Break PointRecommendsdocumentary · Netflix

    “so full swing draft to survive break points tennis golf F1 all of them”

    ▶ Watch at ~1:06:30approx.
  • NetflixUsescompany · Netflix

    “I like the sports sort of docy ones on Netflix”

    ▶ Watch at ~1:06:30approx.
  • ParticleRecommendssoftware

    “for news and stuff is a particle which uh is great news news aggravation tool AI news tool”

    ▶ Watch at ~1:07:30approx.
  • Uber SUVMentionsproduct · Uber

    “was launching Uber's second ever product type which was Uber SCV”

    ▶ Watch at ~1:11:00approx.
  • TwitterUseswebsite · Twitter

    “they can find me online on Twitter or LinkedIn”

    ▶ Watch at ~1:13:30approx.
  • LinkedInUseswebsite · LinkedIn

    “they can find me online on Twitter or LinkedIn”

    ▶ Watch at ~1:13:30approx.
  • Apple PodcastsRecommendssoftware · Apple Inc.

    “you can subscribe to the show on Apple podcast Spotify or your favorite podcast app”

    ▶ Watch at ~1:14:00approx.
  • SpotifyRecommendssoftware · Spotify

    “you can subscribe to the show on Apple podcast Spotify or your favorite podcast app”

    ▶ Watch at ~1:14:00approx.
  • LennyPodcast.comMentionswebsite · Lenny Rachitsky

    “you can find all past episodes or learn more about the show at Lenny podcast.com”

    ▶ Watch at ~1:14:30approx.

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

Insights & moments

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

Hot Take· 1

Hot Take31:30

Don't Be a Hammer Looking for Nails With Frameworks

Brian likes Jobs To Be Done but warns against dogmatism: pick the right framework for the right job rather than forcing everything into one. He gives an Open Door example where an early framing of the job ("get an offer from Open Door") was too narrow, and the truer job was price discovery for the customer. The value is cultural internalization, not just wedging content into a template.

  • Have more tools in the toolbox and know when to apply each; avoid the hammer-and-nail trap
  • JTBD forces empathy, useful at Open Door where customers transact only ~once every seven years
  • A shallow job ("get an offer") can mask the real one ("price discovery")
  • Templates don't make content better; cultural internalization does

“we try to avoid um being a hammer and everything's a nail”

Brian Tolkin · 32:00

“the broader job to be done might be like price Discovery for the customer right”

Brian Tolkin · 37:30
#jobs-to-be-done#frameworks#product#opendoor

Explainer· 6

Explainer02:30

Why Starting in Ops Made Him a Better Product Leader

Brian argues that beginning his career on Uber's operations team gave him a ground-level understanding of what actually moves a business. Running a city day-in and day-out, onboarding drivers, and answering support tickets put him close to the customer. That foundation, he says, is what lets you then decide what's worth building in a scalable technology way.

  • Operating a city directly teaches you what drives the metrics (down to whether it rained that weekend)
  • Early Uber had no centralized support team, so operators were the ones closest to the customer
  • Deep operational understanding is a strong foundation for later product decisions

“starting on the operation side gave a really deep understanding of like how the business actually works”

Brian Tolkin · 03:00
#operations#product-leadership#uber#career
Explainer06:30

The Twin-Turbine Model: Product and Ops as One Engine

Brian describes how Uber and Open Door treated product and operations like a twin-turbine jet plane: you can fly on one engine for a while, but it runs best when both work together. Local ops teams iterate faster and gather far richer qualitative insight than a PM sitting in San Francisco ever could. The goal is harmony, not competition, with a strong feedback loop between the field and the builders.

  • Product and ops run best as harmony, not a competition
  • Local ops teams iterate faster and get better qualitative insights on the ground
  • A PM in SF can't be in 50 markets walking houses or 1,000 cities understanding local safety nuances
  • Fostering a good field-to-builder feedback loop is the lever

“kind of like a twin turbine jet jet plane where you can like fly the plane on one engine for a little bit if you…”

Brian Tolkin · 07:00
#operations#product#org-design#feedback-loops
Explainer21:30

Do Things That Don't Scale, Then Scale Them

Brian walks through how Uber driver onboarding evolved: from 90-minute one-on-one sessions, to small classes, to a video, until validating thousands of licenses a week broke the manual system. That's the signal to move from the iteration stage to the scale stage, where technology (like OCR license validation) uniquely excels. Automating it also freed hundreds of people to go solve the next problem.

  • Onboarding scaled from 1-on-1 → small classes → 20–30 at a time → onboarding video
  • License validation broke once volume hit ~1,000/week, signaling the shift from iteration to scale
  • Tech (OCR/auto-recognition) is uniquely good at scaling once the process is understood
  • Automation freed hundreds of people to tackle the next optimization

“do things that don't scale and then scale the things that you're doing”

Brian Tolkin · 23:30
#scaling#operations#automation#uber
Explainer26:00

How to Run Product Reviews That Aren't Firing Squads

Brian sees product reviews as serving two goals: accountability/informing an audience, and, most importantly, making the product better by helping teams think through problems. The primary goal means reviews should feel intellectual, not scary. He keeps the live conversation small (under 10), while the resulting documents can go to wide distribution and are especially powerful for onboarding new people.

  • Two stated goals: accountability/informing, and (primary) making the product better
  • Reviews shouldn't feel like firing squads; that environment doesn't improve products
  • Keep the conversation small, ideally under 10 people
  • The written artifacts are valuable for the whole team and great for onboarding new hires

“product reviews hopefully are not feeling like firing squads”

Brian Tolkin · 27:00

“the best conversations happen when they're relatively small so try and keep it under under 10”

Brian Tolkin · 29:30
#product-reviews#leadership#meetings#culture
Explainer40:00

How to Experiment When You Have Low Sample Sizes

Open Door does few, very large transactions, so classic A/B testing is often underpowered. Brian's advice: run a power analysis before committing, and be honest about runtime, sometimes a six-month set-and-forget experiment is the right call. Where a canonical test isn't possible, raise conviction other ways (talk to customers, observational data, twin cities, 80% confidence, long-term holdouts) and don't chase false precision.

  • Always run a power analysis before starting an A/B test; know the detectable effect and runtime
  • Some important experiments justify a six-month runtime; the real mistake is faking an answer in a month
  • Alternatives to raise conviction: talk to customers, observational data, sister/twin cities, segmenting, 80% confidence, long-term holdouts
  • When nothing works, trust intuition and ship; don't waste time chasing false precision

“don't just for for yourself into AB testing without running the power analysis”

Brian Tolkin · 41:00

“you shouldn't spend time trying to get false Precision”

Brian Tolkin · 44:00
#experimentation#ab-testing#statistics#opendoor
Explainer56:30

Product Is Finding the Kernel of Truth in a Sea of Signals

Signals about your product come from everywhere, CS teams, customers, executives, a YouTube video, and the core job is deciding what actually matters versus what is noise. That means saying no to things that sound like good ideas. Brian adds a practical tip: write down every idea and piece of feedback, both to reference later and to make sure the people who raised them feel heard and respected.

  • Good ideas come from everywhere; the job is separating what truly matters from noise
  • Doing the job well means saying no to some seemingly good ideas
  • Write down all ideas and feedback in whatever backlog you keep
  • Capturing ideas signals that contributors were heard and their input considered

“the core job is to understand what really matters right like what is noise what is a good idea what is a suggestion”

Brian Tolkin · 57:00

“making sure the people who present those ideas are heard and respected”

Brian Tolkin · 59:30
#product-strategy#prioritization#feedback#focus

Story· 7

Story08:00

How Product Operations Was Born at Uber

Uber had a centralized product team mostly in San Francisco and a globally distributed ops team, with a weak bidirectional feedback loop between them. The fix was a new function, Product Operations, whose members reported into operations but physically sat with and operated like part of the product team. Brian is modest about credit, noting Google and others had dabbled in similar models before him.

  • Problem: SF-built features didn't translate cleanly to global markets, and market input didn't flow back
  • Product Ops reported into operations but embedded with the product team
  • The role formalized and built out an organization others had only dabbled in

“our solution at the time was to start up a new function called Product operations who had accountability and reported into operations but physically sat…”

Brian Tolkin · 08:30
#product-operations#org-design#uber#scaling
Story10:00

Uber's Surge Pricing Was Once Controlled by Hand

For much of 2012, surge pricing was a manual, human-in-the-loop system: city GMs set the parameters for when and where surge could turn on. A GM might know a baseball game lets out at 10pm and set surge at 9:45, something the algorithm couldn't yet predict. Reasons included caution around a powerful new lever, trust in local knowledge, and the technical difficulty of a fully dynamic geospatial pricing system.

  • GMs controlled whether surge was on/off, the geographies, the hours, and the cap; the algorithm optimized within those bounds
  • Local teams could anticipate events (e.g. a ballgame ending) the algorithm couldn't
  • Three drivers: caution with a powerful new tool, trust in local knowledge, and the hard technical build of always-on dynamic pricing

“GMS in every city would control basically the parameters in which surge would operate”

Brian Tolkin · 10:30
#surge-pricing#uber#operations#pricing
Story12:30

"UberX" Was Just a Placeholder That Stuck

The product that became UberX had no name during modeling, so someone on the ops team used a placeholder: X. The company moved fast enough that it launched under that placeholder, and a decade later UberX is still the name. Brian notes many products start this way, with a temporary name becoming permanent because it's too expensive to rebrand.

  • UberX started as an all-hybrids concept with no agreed name
  • "X" was literally a placeholder in the model
  • It launched fast and the placeholder name stuck permanently

“so it was going to be a placeholder so what do you put as some placeholder X so Uber X”

Brian Tolkin · 13:00
#uber#product-naming#startup-stories
Story13:30

The All-Nighter Launching Uber Pool in China

Launching Uber Pool in Chengdu, the team flipped on the testing infrastructure the weekend before go-live and the matching algorithm didn't work. Brian slept roughly 30 minutes that night while the team debugged with the US, finally getting it working around 5:30am for a 6am launch. Afterward, exhausted, a mediocre street-food pancake became, in his memory, the best meal of his life.

  • The matching algorithm broke the day before a press-backed 6am launch
  • Brian slept ~30 minutes; the fix landed around 5:30–6:00am, just in the nick of time
  • Extreme, sleep-deprived, stressful moments become the best memories and stories

“I remember I slept it up 30 minutes that night between 2 and 3:00 am”

Brian Tolkin · 14:30

“in my mind that was like the best meal I've ever had in my life”

Brian Tolkin · 15:00
#uber#launch#china#startup-stories
Story15:30

Open Door Shut Off Its Core Business When COVID Hit

In March 2020, with people unwilling to let others into their homes and Chinese real-estate data looking frozen, Open Door turned off its core business and stopped buying homes for a few months. It used that time to virtualize the entire process and came out the other side transformed. Brian frames it as a very stressful moment that became a fond memory in hindsight.

  • Open Door physically enters homes to buy and sell, which COVID made impossible
  • They stopped buying homes for a few months rather than operate blind
  • They used the pause to virtualize the whole process

“we actually turned off the core business and we stopped buying homes for a few months”

Brian Tolkin · 16:00
#opendoor#covid#crisis-management#operations
Story47:00

Why Zillow Struggled to Beat Open Door at Its Own Game

Zillow tried to do what Open Door does, failed, and now partners with it instead, sending high-intent audiences to Open Door's selling solution. Brian attributes the difficulty to the business being far more than a software product: you must be excellent at pricing, product, operations, risk discipline, and capital markets simultaneously. That vertical integration has been in Open Door's DNA from day one; a software-driven company can't just bolt it on.

  • Zillow went from competitor to partner, funneling its browsing audience to Open Door
  • The business demands excellence across pricing, product, operations, risk, and capital markets at once
  • Vertical integration was in Open Door's DNA from day one
  • A software-first company underestimates how hard bolting on operations and capital is

“you have to be really good at pricing you have to be really good at product you have to be really good at at the…”

Brian Tolkin · 49:00

“being um competition aware but not necessarily competition focused”

Brian Tolkin · 50:30
#opendoor#zillow#competition#vertical-integration
Story1:00:00

Failure Corner: The Early Uber Pool Launch and the Liquidity Lesson

Early Uber Pool in San Francisco tried to match commuters by specific companies and corridors, but the team quickly learned that liquidity is the only thing that matters and a company-based approach could never generate enough of it. They pivoted, running a $5-anywhere-in-SF pool promotion to juice liquidity and find the product's limits. The lesson: don't overthink it or get too cute with a beta, sometimes you just need a lot more people in it.

  • The initial commuter/company-matching strategy couldn't generate enough liquidity
  • Liquidity was the only thing that mattered for the carpooling product
  • A $5-anywhere-in-SF promotion was used to juice liquidity and probe the product's limits
  • Lesson: don't overthink the beta or get too cute; you just need enough people in it

“liquidity is the only thing that matters”

Brian Tolkin · 1:01:30

“don't overthink it don't try to get too cute”

Brian Tolkin · 1:02:30
#uber#uber-pool#failure#liquidity#marketplaces

Takeaway· 1

Takeaway52:30

Reflecting Stress Onto Your Team Doesn't Produce Better Outcomes

Brian's calm was sharpened in Uber's constant fire-drills, but the core lesson is that projecting stress onto teams makes everyone tense and counterintuitively worsens outcomes. He leans on mantras like "you're never as good, or as bad, as you think you are" to keep a clearer head. The uncomfortable truth is you often only gain this perspective by living through stressful situations, and learning from them, so the next one feels familiar.

  • Reflecting stress onto teams makes them tense and lowers performance
  • Mantra: you're never as good as you think, never as bad as you think, aim for an even keel
  • Perspective that "things pass" mostly comes from having been through it before
  • Advice: expose yourself to stressful situations rather than running from them, and learn from them

“when you reflect the stress onto your teams everybody tenses up and tightens up”

Brian Tolkin · 53:30

“You're Never As Good As you think you are you're never as bad as you think you are”

Brian Tolkin · 53:30
#leadership#resilience#stress#mindset