A social network built around connecting people and sharing.
Marketplace came from data showing 40% of Facebook group posts were buying and selling, and it now generates over $100 billion in GMV a year. A seller-count gate for region launches hid the power sellers making a region work, and a Chile trip exposed a feed that barely loaded on slow LTE. Build for how users already misuse your product, revisit metrics, and visit customers.
- 1Spot forced behavior40% of group posts were buying and selling
- 2Launch buy and sell groupsServed the demand inside Groups first
- 3Build full MarketplaceDedicated product after groups proved demand
- 4Scale region by regionOver $100 billion GMV a year
From Boris Cherny on latent demand; Fiona Fung
What was said about Facebook
1How Facebook launched at Harvard and spread school by school
Facebook seeded one dense Harvard network so every newcomer found it alive. It launched at Harvard on Feb 4, 2004. 650 people joined in the first 24 hours and over half of undergraduates were active within two weeks.
70% of everyone who had ever signed up was active on a given day, and that held for years as schools were added. Friendster was open to everyone and sent newcomers to an empty site. Quote: "almost nobody can use the Facebook but for the people who can it's an unbelievably great experience"
Real names and user-entered content made trust the foundation from day one. Sign-up required a harvard.edu email, so profiles were verified classmates. After the Facemash scraping trouble, Zuckerberg had users enter all their own content, including AIM screen names and phone numbers.
MySpace and Friendster were open to anyone and grew faster at first. Quote: "an authentic identity is a part of the company from its first moment"
Zuckerberg attacked the campus with a rival first, then walled off neighbors. Three weeks after Harvard, Facebook opened at Columbia, where a competitor already existed, to learn right away whether it could displace an incumbent. It reached over 100 schools in about four months.
Each school was a separate network on its own servers, which kept trust high and contained server load. A school opened only when demand was strong, with waitlists building in between.
Copy the Harvard-only Facebook, not the Facebook of today. Facebook began by letting people open a profile and connect with classmates in two or three steps; its polish comes from about 20 years of stacked features. Founders who copy the finished product overbuild.
Pick a comically narrow first customer, as Facebook did with college students and Amazon with books. Mark Pincus makes the same case from Zynga, which began as a single Facebook poker game. The founders were still working on side projects like webcam gestural controls while it grew.
Every early feature fed the same engagement loop. Zuckerberg's summer 2005 roadmap listed photos, events, a personalized newspaper, local business and games. Events led to photos, tagging and feed stories, which pulled people to the next event and back onto Facebook.
Photos was written by one engineer in two weeks. By November 2005 Facebook served 230 million page views a day, more than Google, with under 10 million users.
2The growth team and seven friends in 10 days
Facebook treated growth as a product and engineering job, not marketing. When growth stalled in 2008, Chamath Palihapitiya, Alex Schultz, Naomi Gleit and Javier Olivan formed a team to grow Facebook using Facebook itself. Users went from about 100 million in August 2008 to almost 150 million by year end and 350 million in 2009.
They worked on registration, invites and new-user onboarding. Quote: "the product is the biggest lever to drive growth"
Growth accounting showed retention, not acquisition, was the biggest lever. Net growth is new users, minus stale users (no login for 30 days), plus resurrected users. At Facebook the churn and resurrection lines dwarfed new registrations.
The framework came from Danny Ferrante. It pushed the team toward engagement and getting people to the magic moment fast so they would not churn.
One shared activation goal mattered more than its exact number. Friending correlated most with retention, so the team rallied around seven friends in 10 days. Colleagues later remembered it as 10 friends in 14 days, and it still worked.
Both are points on the same retention curve. Quote: "the majority of the value is in the latter is just having extreme Clarity around the goal"
Onboarding was built to hit the friend goal. Open registration in 2006 brought people like Naomi's dad and grandma who did not know how to use it. The flow became: upload a profile picture, then find your friends via contact import and People You May Know.
The early college site needed no onboarding because students learned it from the person next to them. Find-your-friends became one of the biggest drivers of activation.
The team paused its roadmap to instrument every signup step. In January 2009 the growth team built nothing but data instrumentation on registration and onboarding, which showed where users dropped off.
The loop was understand, identify, execute. Alex Schultz's shirt read "why guess when you can know".
Counting a notification click as email confirmation removed a 20% drop-off. Instrumentation showed 20% of new users never confirmed their email. An SMS test did not fix it, so unconfirmed accounts kept getting notifications and any click counted as confirmation.
A click proved ownership of the email. Quote: "it's just removing a micro barrier of having to like go find the confirmation email"
Removing macro barriers widened the audience step by step. Eligibility went from college students to high schoolers, work networks, then anyone with an email. The team also built for feature phones, worked on internet access and crowdsourced translation into over 100 languages.
Translation happened inline with ranking of the best versions, like Wikipedia. The team kept one graph rather than a separate Facebook High.
Naomi Gleit doubts how much growth the team really caused. She sees most growth work as at the margin of demand that product-market fit already created, with the clearest impact in macro barrier work like translation.
Colleges were on waiting lists to join in 2005. Quote: "I'm actually not sure how much credit we deserve"
In India, engineers watched people friend each other at home. People recommendations underperformed in India, where friends shared about 7 mutual friends versus 22 in the US. Bangaly Kaba took engineers to homes in Delhi and Mumbai every three months.
Profile fields were irrelevant, so people checked photos, and common names made search useless: Amit Kumar matched about 250,000 people a month. The team then designed around name frequency.
Switching the North Star to daily users pushed daily return. After reaching 1 billion monthly users, Facebook moved from monthly to daily active users, giving the team a reason to bring people back every day.
Sean Ellis uses it as an example of a North Star that counts delivered value, not revenue.
Even Facebook never sustained a viral factor above one. In its heyday Facebook's viral factor was about 0.7, for perhaps a year. LinkedIn's address book import had about 0.4 lifetime.
Rahul Vohra's conclusion is to aim for word of mouth, the virality no feature mechanic captures.
3News Feed: launches, revolts and redesigns
Zuckerberg shipped News Feed to everyone and trusted engagement over the revolt. News Feed launched September 5, 2006 with no opt in. Users sent 30,000 angry emails and 10% joined a protest group, but engagement soared, so Zuckerberg posted "calm down, breathe, we hear you" and added privacy controls.
A hack that highlighted profile changes in yellow had already shown demand. News Feed cut page views, hurting revenue. Within two weeks users adapted, though the lesson that users always adapt did not always hold later.
Explain trade-offs publicly for onlookers, not to convert critics. Adam Mosseri led the 2009 News Feed redesign, which drew vicious criticism he likens to rearranging someone's desk without warning. He began debating openly, laying out what was done, why and the trade-offs.
Quote: "I'm just trying to enumerate all of the different puts and takes for the rest of the people watching the conversation."
Feed designs had to be presented with real, messy data. Mockups looked perfect in Photoshop but ugly once real photos, post lengths and comments filled them. Designers were required to present with real feed data.
Quote: "you need to actually design a system that produces both in content and visual design, like a delightful experience given input you don't control."
A carefully designed sharing loop kept News Feed stable for years. Peter Deng, Facebook's 4th PM, says News Feed has changed little in about 12 years because the sharing loop was designed carefully. News Feed and photo tagging came from watching what people wanted.
His broader point: Facebook won by polishing a simple database of human connections, not by lab breakthroughs.
4Facebook Platform and the open-then-close cycle
Facebook Platform thrived on the web but could not survive mobile. At f8 in May 2007 Facebook opened its API and hit 5,000 developers in two days. On mobile no app can run inside another, credits could not carry a second 30% fee, and Steve Jobs declined Zuckerberg's 2008 pitch to embed the platform in iOS.
Facebook later cut organic app distribution, and what remained was selling distribution as app install ads, a large share of early mobile ad revenue.
Platforms open to win content, then close; plan your exit on entry. Facebook opened in 2007 at about a quarter of MySpace's size, then about five years later took revenue cuts, suppressed organic reach and absorbed uses like events and photos. Brian Balfour sees the same four steps at Google, iOS, LinkedIn and Udemy.
Zynga's strategy was tuned to Facebook's graph and channels. Build defensibility right away: own part of the workflow, gather data the platform lacks, or build micro network effects. Startups should bet on one platform, not several.
Facebook beat bigger rivals on retention, not distribution. In 2007 Facebook was smaller than MySpace and Friendster but had better retention. Balfour applies the same test to ChatGPT, whose retention curves level off higher than rivals'.
Quote: "it was never the person who had the biggest distribution at the moment of time. It was the one that had the best retention and engagement."
5Founder control and the bet on mobile feed ads
Permanent founder control let Zuckerberg refuse every exit. Sean Parker, ousted at Plaxo, steered Facebook to investors who accepted Zuckerberg's board control. He turned down Yahoo's $1 billion in 2006 and a Microsoft offer reported at $24 billion in 2007.
Microsoft instead took 1.6% for $240 million at a $15 billion valuation, plus exclusive international banner ad rights until 2011 to feed its ad marketplace. The stake returned 7x by the 2012 IPO.
Facebook sacrificed desktop ad revenue to rebuild around mobile feed ads. After the May 2012 IPO at $38, with no meaningful mobile revenue, the stock fell 53.5%. Sandberg moved engineering to mobile and accepted missed quarters; by Q4 2013 mobile was 53% of ad revenue and total ad revenue grew 76%.
Feed ads fill the phone screen, so they had to be good, which made a more valuable unit than desktop ever had. Sandberg: "well Mark nobody can fire you and only you can fire me so if you're in I'm in"
6Marketplace, Dating and building for latent demand
Marketplace came from people already using groups to buy and sell. 40% of Facebook group posts were buying and selling, so Facebook built buy and sell groups and then Marketplace. Dating began from data that 60% of profile views were between non-friends of the opposite gender.
Boris Cherny calls it the clearest example of latent demand and used the same signal to build Cowork in about 10 days. Marketplace now generates over $100 billion in GMV a year.
A seller-count launch gate nearly blocked a region that was working. Marketplace gated region expansion partly on number of sellers. The first region had few sellers, but power sellers meant buyers found what they needed, so the metric was changed.
Quote: "make sure that you're not just having blinders on that's blindly following a metric that used to make sense"
A trip to Chile found the growth blocker no dashboard explained. Marketplace underperformed in Chile. Fiona Fung went with a few Android phones and saw the feed barely loaded on slow LTE.
Her rule: use the product daily, and visit customers when you cannot.
7How Facebook answers competitors
Facebook's Poke clone failed to dent Snapchat. Around 2012 and 2013 Facebook copied Snapchat as Poke and put a download banner atop every Facebook app. On Christmas Day Snapchat was still number one in the App Store.
Evan Spiegel took away that software has no moat and invested in what is hard to copy: a messaging network, an AR platform and a creator ecosystem.
Meta buys or copies rival mechanics and runs them on its own AI. Zuckerberg's 2012 email framed buying Instagram as buying time to deploy rivals' mechanics at scale. The AI lab started in 2013 for feed and ad ranking later let Reels answer TikTok quickly.
For live video it reportedly put 300 people on the copy, built live into the core product and paid streamers like Kevin Hart. Live worked as a layer on an asynchronous product, while live-only Periscope failed.
Big-company incentives favor copying proven apps, slowly. Nikita Bier says incumbents take 12 to 24 months to respond, because staff protecting reviews pitch copies of a number one app they do not own. Facebook put 80 people on the TBH deal against him alone.
That delay is the opening for startups with novel ideas.
Facebook's rise scared Google into an all-in Google+ bet. Google built Google+ with about 1,000 people and integrated it into Gmail for years; it shut down in 2019. Facebook went to a "code DEFCON one" response, yet neither company's ad revenue moved.
Itamar Gilad's lesson is not to go all in on an opinion-backed idea, however strategic.
8Getting your funnel ready before Meta ads
Hit funnel targets first, because Meta amplifies the funnel you have. Steve P. Young's targets: 5 to 10% install-to-trial (10%+ with a hard paywall), over 30% trial-to-paid, at least 2% install-to-paid, and at least $3 LTV per install measured around month 11.
Put the trial on the highest-LTV plan and keep a paywall in onboarding, possibly hard for all Meta traffic. Quote: "if you got a leaky bucket, you might as well write Meta a big check cuz you're wasting money."
Prove strangers will pay through search before paying for Meta. ASO and Apple Ads are pull traffic and cheaper, while Meta interrupts people and converts worse. Validate there first, unless your category cannot work on search, like a new social network.
Ladder learned this the hard way: Facebook ads brought the right persona with no context about its coaches, the funnel could not activate them, and spend stopped after about three months.
Run the budget math; $1,000 rarely buys enough signal. At $2 per install, $1,000 buys 500 installs, 35 trials at 7% and about 10 to 11 subscribers, roughly $95 each. Meta needs about 50 conversion events a week.
Plan weekly budget to reach 50 trials. Quote: "that $1,000 is not going to get us very far."
Cheap early results at small budgets are not product-market fit. At a few hundred dollars a day the algorithm finds the low-hanging fruit, so an MVP looks like it fits. Companies raise, scale to about $10,000 a day, and costs blow up.
Fix one funnel step at a time. On Facebook, Ladder tried to fix clicks, first workouts and paid conversion at once and went deep on none. On TikTok it mapped the funnel and fixed creative, then clicks, the homepage, then the quiz.
Strong ad metrics with non-converting installs may call for a web funnel. Steve's Facebook ads to an older audience get strong hook rates, hold rates and cost per install, but installs do not pay. He suspects older users avoid in-app purchases.
He plans to test paying on the web first through RevenueCat's funnels. This is untested.
9Setting up and running Meta app install campaigns
Install the SDK and verify Meta's events against App Store Connect. The Facebook SDK reports results so Meta can find likely installers. Send install, paywall view, checkout, trial and subscribe events, then compare counts with App Store Connect for the same dates.
In Steve's check Meta showed about 276 trial starts vs 312, and a little over 4,000 installs vs about 3,000: close enough to trust. Aim for near 100% install-to-paywall-view.
Optimize as far down the funnel as event volume allows. Meta delivers whatever you optimize for, so install-optimized campaigns get cheap installs that never pay. Facebook needs at least 50 events in a rolling seven days.
Steve goes straight to trial starts if volume allows. Vivek Girotra starts at installs to gather signal, then walks down to trial start and subscription; the event must be recent, frequent and correlated with revenue.
Go broad on targeting and let creative and placements do the work. Interest targeting never delivered scale; broad targeting with strong creative does. Use dynamic placements across Facebook, Instagram, Messenger and Audience Network.
Seed lookalikes with first-party lists and test 1%, 5% and 10%. Keep Advantage Plus expansion within 50% of your estimated audience; 4 million growing to 15 million is too far.
Don't count on measurable retargeting on iOS. With no device ID there is no suppression list, so ads reach people who already have the app. Use lookalikes for prospecting and custom audiences for retargeting, where Android has more options.
Run paid ads yourself rather than through an agency. Ladder hired the wrong agency during iOS 14, and the math failed. Greg Stewart now runs TikTok Ads Manager himself every day.
Quote: "you have to be black belt at growth and product"
10Ad creative for Meta
Mine the Meta Ads Library for long-running competitor ads. Filter to active ads and find the oldest; an ad running for months is probably profitable. A Cal AI ad started June 3rd was still running about four months later.
It was a 9-second video with a calorie-tracking hook, in-app demo and App Store end card. Study complementary apps too and save examples with start dates.
Test distinct concepts before small variations. Launch three to five concepts aimed at different customer profiles in one ad set, such as food tracking, before-and-after weight loss, and tracking snacks. Vary details only after a concept wins.
Every ad needs a hook, demo, offer and App Store end card. The opening must stop the scroll, then show the app, the benefit and end on the App Store page. Keep the promise consistent through the store screenshots.
In later phases, tie each ad to a custom product page, such as an ADHD screenshot set for an ADHD ad.
Winning TikTok ads transfer to Instagram Reels almost unchanged. A campaign getting 10,000 users at a $1 CAC on TikTok performed about the same run as Reels through Facebook ads. Ladder also saw Meta perform well during a TikTok outage.
11How Facebook alumni run teams
Assign one person to grow the product and another to guard craft. Peter Deng gives two people opposing charges and adjudicates the tension himself, rather than relying on one person to balance both.
He set this up at Facebook, Instagram, Airtable and ChatGPT.
Make the final interview test only growth mindset. After a bad Facebook hire, Deng's last interview asks for the candidate's most painful mistake and how it changed their work, after sharing one of his own.
His bar: within six months, if he is still telling you what to do, he hired the wrong person.
Build extreme clarity with one canonical doc per project. Naomi Gleit keeps one doc listing workstreams, a single-threaded owner for each, meetings and channels. Clarity means shared facts, not agreement.
She sends pre-reads 24 hours ahead with three options and a recommendation, scored in a red, yellow and green table.
Ask each morning what the most impactful thing is today. After blunt feedback from Sheryl Sandberg, Bret Taylor stopped shaping his job around work he enjoyed and improved at running Facebook's mobile and platform org of over 1,000 people.
Choose and leave roles by learning and impact. Deng joined Facebook to learn the most and moved teams roughly every two and a half years over nine and a half years. Bangaly Kaba scores manager, resources, scope, team, pay and culture yearly.
Too much scope and several manager changes made Kaba leave a great Facebook team.
12Facebook as a reference point elsewhere
Facebook's data advantage pushed AppLovin to buy game studios. Advertisers shared spend data with Facebook but not a small startup, so Adam Foroughi bought 14 or 15 studios to train Axon, held them about five years, then sold them to Tripledot.
Facebook Ads' shift to mobile native is a template for AI native. Jag Duggal cites Fidji Simo's mobile native push to argue for building AI at the heart of a product, not its corners.
From ChatGPT to Instagram to Uber: The quiet architect behind the world’s most popular products
Deng spent nine and a half years there as the 4th product manager, building News Feed, Messenger, photos, groups and profiles. News Feed is little changed after about 12 years because the sharing loop was designed carefully.
Hit funnel targets first, because Meta amplifies the funnel you have. Steve P. Young's targets: 5 to 10% install-to-trial (10%+ with a hard paywall), over 30% trial-to-paid, at least 2% install-to-paid, and at least $3 LTV per install measured around month 11.
Put the trial on the highest-LTV plan and keep a paywall in onboarding, possibly hard for all Meta traffic. Quote: "if you got a leaky bucket, you might as well write Meta a big check cuz you're wasting money."
Prove strangers will pay through search before paying for Meta. ASO and Apple Ads are pull traffic and cheaper, while Meta interrupts people and converts worse. Validate there first, unless your category cannot work on search, like a new social network.
Ladder learned this the hard way: Facebook ads brought the right persona with no context about its coaches, the funnel could not activate them, and spend stopped after about three months.
Run the budget math; $1,000 rarely buys enough signal. At $2 per install, $1,000 buys 500 installs, 35 trials at 7% and about 10 to 11 subscribers, roughly $95 each. Meta needs about 50 conversion events a week.
Plan weekly budget to reach 50 trials. Quote: "that $1,000 is not going to get us very far."
Install the SDK and verify Meta's events against App Store Connect. The Facebook SDK reports results so Meta can find likely installers. Send install, paywall view, checkout, trial and subscribe events, then compare counts with App Store Connect for the same dates.
In Steve's check Meta showed about 276 trial starts vs 312, and a little over 4,000 installs vs about 3,000: close enough to trust. Aim for near 100% install-to-paywall-view.
Optimize as far down the funnel as event volume allows. Meta delivers whatever you optimize for, so install-optimized campaigns get cheap installs that never pay. Facebook needs at least 50 events in a rolling seven days.
Steve goes straight to trial starts if volume allows. Vivek Girotra starts at installs to gather signal, then walks down to trial start and subscription; the event must be recent, frequent and correlated with revenue.
Mine the Meta Ads Library for long-running competitor ads. Filter to active ads and find the oldest; an ad running for months is probably profitable. A Cal AI ad started June 3rd was still running about four months later.
It was a 9-second video with a calorie-tracking hook, in-app demo and App Store end card. Study complementary apps too and save examples with start dates.
Test distinct concepts before small variations. Launch three to five concepts aimed at different customer profiles in one ad set, such as food tracking, before-and-after weight loss, and tracking snacks. Vary details only after a concept wins.
Every ad needs a hook, demo, offer and App Store end card. The opening must stop the scroll, then show the app, the benefit and end on the App Store page. Keep the promise consistent through the store screenshots.
In later phases, tie each ad to a custom product page, such as an ADHD screenshot set for an ADHD ad.
Meta’s head of product on working with Mark Zuckerberg, early growth tactics, and more | Naomi Gleit
Growth accounting showed churn and resurrection dwarfed new registrations, so the team chased an activation goal of seven friends in 10 days. Fixing a 20% email-confirmation drop-off, crowdsourced translation into 100+ languages and removing macro barriers helped growth, though much came from product-market fit.
Strong ad metrics with non-converting installs may call for a web funnel. Steve's Facebook ads to an older audience get strong hook rates, hold rates and cost per install, but installs do not pay. He suspects older users avoid in-app purchases.
He plans to test paying on the web first through RevenueCat's funnels. This is untested.
Meta
Launched at Harvard on Feb 4, 2004, with 650 sign-ups in 24 hours and over half of undergraduates active within two weeks. It grew school by school, hit 230 million daily page views by November 2005, went public in 2012 at $38 and pivoted to mobile feed ads.
Explain trade-offs publicly for onlookers, not to convert critics. Adam Mosseri led the 2009 News Feed redesign, which drew vicious criticism he likens to rearranging someone's desk without warning. He began debating openly, laying out what was done, why and the trade-offs.
Quote: "I'm just trying to enumerate all of the different puts and takes for the rest of the people watching the conversation."
Before You Spend $1 on Meta Ads, Watch This
Meta interrupts people, so its traffic converts worse than search channels, and it optimizes for whatever event it is told to chase. It needs about 50 events per week, so a $1,000 test at $2 per install yields only 35 trials and about $95 per subscriber.
Marketplace came from people already using groups to buy and sell. 40% of Facebook group posts were buying and selling, so Facebook built buy and sell groups and then Marketplace. Dating began from data that 60% of profile views were between non-friends of the opposite gender.
Boris Cherny calls it the clearest example of latent demand and used the same signal to build Cowork in about 10 days. Marketplace now generates over $100 billion in GMV a year.
A seller-count launch gate nearly blocked a region that was working. Marketplace gated region expansion partly on number of sellers. The first region had few sellers, but power sellers meant buyers found what they needed, so the metric was changed.
Quote: "make sure that you're not just having blinders on that's blindly following a metric that used to make sense"
A trip to Chile found the growth blocker no dashboard explained. Marketplace underperformed in Chile. Fiona Fung went with a few Android phones and saw the feed barely loaded on slow LTE.
Her rule: use the product daily, and visit customers when you cannot.
Are You Ignoring These RevenueCat Features? (Stop Losing Money)
Steve's Facebook ads to an older audience get strong hook rates, hold rates and cost per install, but the installs don't convert. He hypothesizes older users won't buy in-app and plans to test web funnels.
Copy the Harvard-only Facebook, not the Facebook of today. Facebook began by letting people open a profile and connect with classmates in two or three steps; its polish comes from about 20 years of stacked features. Founders who copy the finished product overbuild.
Pick a comically narrow first customer, as Facebook did with college students and Amazon with books. Mark Pincus makes the same case from Zynga, which began as a single Facebook poker game. The founders were still working on side projects like webcam gestural controls while it grew.
The rise of taste, human authenticity and judgment in an AI world | Adam Mosseri (Head of IG)
Mosseri designed and led the early version and ran the ranking team. The 2009 redesign drew vicious criticism, which he compares to rearranging someone's desk without warning.
Copy the Harvard-only Facebook, not the Facebook of today. Facebook began by letting people open a profile and connect with classmates in two or three steps; its polish comes from about 20 years of stacked features. Founders who copy the finished product overbuild.
Pick a comically narrow first customer, as Facebook did with college students and Amazon with books. Mark Pincus makes the same case from Zynga, which began as a single Facebook poker game. The founders were still working on side projects like webcam gestural controls while it grew.
What happens after coding is solved? | Fiona Fung (Claude Code & Cowork)
Fiona took it from idea to launch, and it now generates over $100 billion in GMV a year. An early seller-count gate for region launches hid power sellers who were making the first region work, and a trip to Chile showed the feed barely loaded on slow LTE.
Facebook's data advantage pushed AppLovin to buy game studios. Advertisers shared spend data with Facebook but not a small startup, so Adam Foroughi bought 14 or 15 studios to train Axon, held them about five years, then sold them to Tripledot.
The hidden pattern behind successful products | Mark Pincus (FarmVille, Words with Friends, & more)
It began from an embarrassingly small start and gave Zynga its social graph. Net promoter scores went from plus 35 to minus 35 when people quit it and Instagram, and it drifted toward wasting time.
Facebook's Poke clone failed to dent Snapchat. Around 2012 and 2013 Facebook copied Snapchat as Poke and put a download banner atop every Facebook app. On Christmas Day Snapchat was still number one in the App Store.
Evan Spiegel took away that software has no moat and invested in what is hard to copy: a messaging network, an AR platform and a creator ecosystem.
Why Most Apps Are Failing in 2026: 7 Mistakes Killing Your App Success
It looks polished because of years of stacked features, and its first version was stripped down. Founders who copy its finished look tend to overbuild.
Copy the Harvard-only Facebook, not the Facebook of today. Facebook began by letting people open a profile and connect with classmates in two or three steps; its polish comes from about 20 years of stacked features. Founders who copy the finished product overbuild.
Pick a comically narrow first customer, as Facebook did with college students and Amazon with books. Mark Pincus makes the same case from Zynga, which began as a single Facebook poker game. The founders were still working on side projects like webcam gestural controls while it grew.
Bootstrapping a Business to $5 Billion in Free Cash Flow | AppLovin’s Adam Foroughi
Its machine learning made ads feel like content, and its social graph and pixels gave it data that AppLovin's rules-based algorithm lacked. That pushed AppLovin to buy studios for data.
Marketplace came from people already using groups to buy and sell. 40% of Facebook group posts were buying and selling, so Facebook built buy and sell groups and then Marketplace. Dating began from data that 60% of profile views were between non-friends of the opposite gender.
Boris Cherny calls it the clearest example of latent demand and used the same signal to build Cowork in about 10 days. Marketplace now generates over $100 billion in GMV a year.
Evan Spiegel, Snapchat: Building a Multi-Billion Dollar Company
Facebook cloned Snapchat as Poke around 2012 or 2013 with a download banner on every Facebook app, yet Snapchat was still number one on Christmas Day. The episode taught Snap to build what is hard to copy.
Platforms open to win content, then close; plan your exit on entry. Facebook opened in 2007 at about a quarter of MySpace's size, then about five years later took revenue cuts, suppressed organic reach and absorbed uses like events and photos. Brian Balfour sees the same four steps at Google, iOS, LinkedIn and Udemy.
Zynga's strategy was tuned to Facebook's graph and channels. Build defensibility right away: own part of the workflow, gather data the platform lacks, or build micro network effects. Startups should bet on one platform, not several.
Facebook beat bigger rivals on retention, not distribution. In 2007 Facebook was smaller than MySpace and Friendster but had better retention. Balfour applies the same test to ChatGPT, whose retention curves level off higher than rivals'.
Quote: "it was never the person who had the biggest distribution at the moment of time. It was the one that had the best retention and engagement."
How to Sell Your App or SaaS Business: Exits, Acquisitions & App Marketing Secrets
Cited as the model for minimum viable revenue: it started small for Harvard, while many founders try to build today's Facebook from day one.
Feed designs had to be presented with real, messy data. Mockups looked perfect in Photoshop but ugly once real photos, post lengths and comments filled them. Designers were required to present with real feed data.
Quote: "you need to actually design a system that produces both in content and visual design, like a delightful experience given input you don't control."
Ask each morning what the most impactful thing is today. After blunt feedback from Sheryl Sandberg, Bret Taylor stopped shaping his job around work he enjoyed and improved at running Facebook's mobile and platform org of over 1,000 people.
Head of Claude Code: What happens after coding is solved | Boris Cherny
It came from the finding that 40% of group posts were buying and selling, and began as buy and sell groups before the full product. Cherny cites it as the clearest example of latent demand. It started from data showing 60% of profile views were between non-friends of the opposite gender.
A carefully designed sharing loop kept News Feed stable for years. Peter Deng, Facebook's 4th PM, says News Feed has changed little in about 12 years because the sharing loop was designed carefully. News Feed and photo tagging came from watching what people wanted.
His broader point: Facebook won by polishing a simple database of human connections, not by lab breakthroughs.
Assign one person to grow the product and another to guard craft. Peter Deng gives two people opposing charges and adjudicates the tension himself, rather than relying on one person to balance both.
He set this up at Facebook, Instagram, Airtable and ChatGPT.
Make the final interview test only growth mindset. After a bad Facebook hire, Deng's last interview asks for the candidate's most painful mistake and how it changed their work, after sharing one of his own.
His bar: within six months, if he is still telling you what to do, he hired the wrong person.
Choose and leave roles by learning and impact. Deng joined Facebook to learn the most and moved teams roughly every two and a half years over nine and a half years. Bangaly Kaba scores manager, resources, scope, team, pay and culture yearly.
Too much scope and several manager changes made Kaba leave a great Facebook team.
Why ChatGPT will be the next big growth channel (and how to capitalize on it) | Brian Balfour
Opened its platform in 2007 at about a quarter of MySpace's size, with the friend graph as its moat and social gaming booming. About five years later it closed up by taking revenue cuts, suppressing organic reach and absorbing uses like events and photos.
Even Facebook never sustained a viral factor above one. In its heyday Facebook's viral factor was about 0.7, for perhaps a year. LinkedIn's address book import had about 0.4 lifetime.
Rahul Vohra's conclusion is to aim for word of mouth, the virality no feature mechanic captures.
He saved OpenAI, invented the “Like” button, and built Google Maps: Bret Taylor (Sierra)
Newsfeed mock-ups looked perfect in Photoshop but ugly with real data, so designers had to present with real, messy data. Sheryl Sandberg's blunt feedback taught Taylor to ask what is most impactful each day.
Prove strangers will pay through search before paying for Meta. ASO and Apple Ads are pull traffic and cheaper, while Meta interrupts people and converts worse. Validate there first, unless your category cannot work on search, like a new social network.
Ladder learned this the hard way: Facebook ads brought the right persona with no context about its coaches, the funnel could not activate them, and spend stopped after about three months.
Fix one funnel step at a time. On Facebook, Ladder tried to fix clicks, first workouts and paid conversion at once and went deep on none. On TikTok it mapped the funnel and fixed creative, then clicks, the homepage, then the quiz.
Run paid ads yourself rather than through an agency. Ladder hired the wrong agency during iOS 14, and the math failed. Greg Stewart now runs TikTok Ads Manager himself every day.
Quote: "you have to be black belt at growth and product"
Winning TikTok ads transfer to Instagram Reels almost unchanged. A campaign getting 10,000 users at a $1 CAC on TikTok performed about the same run as Reels through Facebook ads. Ladder also saw Meta perform well during a TikTok outage.
Superhuman's secret to success | Rahul Vohra (CEO and founder)
Even in its heyday it had a viral factor of about 0.7, and that lasted perhaps a year.
Platforms open to win content, then close; plan your exit on entry. Facebook opened in 2007 at about a quarter of MySpace's size, then about five years later took revenue cuts, suppressed organic reach and absorbed uses like events and photos. Brian Balfour sees the same four steps at Google, iOS, LinkedIn and Udemy.
Zynga's strategy was tuned to Facebook's graph and channels. Build defensibility right away: own part of the workflow, gather data the platform lacks, or build micro network effects. Startups should bet on one platform, not several.
How Ladder Nailed Product/Channel Fit on TikTok and Meta to Grow 2,000%
Ladder's spend after its 2021 seed round failed within about three months because of the $60 price, a funnel not built for low-context users, and a poorly chosen agency. Meta later performed well on organic and paid during a TikTok outage.
Copy the Harvard-only Facebook, not the Facebook of today. Facebook began by letting people open a profile and connect with classmates in two or three steps; its polish comes from about 20 years of stacked features. Founders who copy the finished product overbuild.
Pick a comically narrow first customer, as Facebook did with college students and Amazon with books. Mark Pincus makes the same case from Zynga, which began as a single Facebook poker game. The founders were still working on side projects like webcam gestural controls while it grew.
An operator’s guide to product strategy | Chandra Janakiraman (CPO at VRChat, ex-Meta, Headspace)
Zynga's strategy was tuned to Facebook's social graph and channels. Meta's growth work also added sponsored ads in the feed and a section for other products in the ecosystem.
Facebook treated growth as a product and engineering job, not marketing. When growth stalled in 2008, Chamath Palihapitiya, Alex Schultz, Naomi Gleit and Javier Olivan formed a team to grow Facebook using Facebook itself. Users went from about 100 million in August 2008 to almost 150 million by year end and 350 million in 2009.
They worked on registration, invites and new-user onboarding. Quote: "the product is the biggest lever to drive growth"
Growth accounting showed retention, not acquisition, was the biggest lever. Net growth is new users, minus stale users (no login for 30 days), plus resurrected users. At Facebook the churn and resurrection lines dwarfed new registrations.
The framework came from Danny Ferrante. It pushed the team toward engagement and getting people to the magic moment fast so they would not churn.
One shared activation goal mattered more than its exact number. Friending correlated most with retention, so the team rallied around seven friends in 10 days. Colleagues later remembered it as 10 friends in 14 days, and it still worked.
Both are points on the same retention curve. Quote: "the majority of the value is in the latter is just having extreme Clarity around the goal"
Onboarding was built to hit the friend goal. Open registration in 2006 brought people like Naomi's dad and grandma who did not know how to use it. The flow became: upload a profile picture, then find your friends via contact import and People You May Know.
The early college site needed no onboarding because students learned it from the person next to them. Find-your-friends became one of the biggest drivers of activation.
The team paused its roadmap to instrument every signup step. In January 2009 the growth team built nothing but data instrumentation on registration and onboarding, which showed where users dropped off.
The loop was understand, identify, execute. Alex Schultz's shirt read "why guess when you can know".
Counting a notification click as email confirmation removed a 20% drop-off. Instrumentation showed 20% of new users never confirmed their email. An SMS test did not fix it, so unconfirmed accounts kept getting notifications and any click counted as confirmation.
A click proved ownership of the email. Quote: "it's just removing a micro barrier of having to like go find the confirmation email"
Removing macro barriers widened the audience step by step. Eligibility went from college students to high schoolers, work networks, then anyone with an email. The team also built for feature phones, worked on internet access and crowdsourced translation into over 100 languages.
Translation happened inline with ranking of the best versions, like Wikipedia. The team kept one graph rather than a separate Facebook High.
Naomi Gleit doubts how much growth the team really caused. She sees most growth work as at the margin of demand that product-market fit already created, with the clearest impact in macro barrier work like translation.
Colleges were on waiting lists to join in 2005. Quote: "I'm actually not sure how much credit we deserve"
Build extreme clarity with one canonical doc per project. Naomi Gleit keeps one doc listing workstreams, a single-threaded owner for each, meetings and channels. Clarity means shared facts, not agreement.
She sends pre-reads 24 hours ahead with three options and a recommendation, scored in a red, yellow and green table.
Identify your bullseye customer in one day | Michael Margolis (UX Research Partner at GV)
Cited as an example of a narrow start: Facebook began with college student profiles before widening.
Facebook seeded one dense Harvard network so every newcomer found it alive. It launched at Harvard on Feb 4, 2004. 650 people joined in the first 24 hours and over half of undergraduates were active within two weeks.
70% of everyone who had ever signed up was active on a given day, and that held for years as schools were added. Friendster was open to everyone and sent newcomers to an empty site. Quote: "almost nobody can use the Facebook but for the people who can it's an unbelievably great experience"
Real names and user-entered content made trust the foundation from day one. Sign-up required a harvard.edu email, so profiles were verified classmates. After the Facemash scraping trouble, Zuckerberg had users enter all their own content, including AIM screen names and phone numbers.
MySpace and Friendster were open to anyone and grew faster at first. Quote: "an authentic identity is a part of the company from its first moment"
Zuckerberg attacked the campus with a rival first, then walled off neighbors. Three weeks after Harvard, Facebook opened at Columbia, where a competitor already existed, to learn right away whether it could displace an incumbent. It reached over 100 schools in about four months.
Each school was a separate network on its own servers, which kept trust high and contained server load. A school opened only when demand was strong, with waitlists building in between.
Every early feature fed the same engagement loop. Zuckerberg's summer 2005 roadmap listed photos, events, a personalized newspaper, local business and games. Events led to photos, tagging and feed stories, which pulled people to the next event and back onto Facebook.
Photos was written by one engineer in two weeks. By November 2005 Facebook served 230 million page views a day, more than Google, with under 10 million users.
Facebook treated growth as a product and engineering job, not marketing. When growth stalled in 2008, Chamath Palihapitiya, Alex Schultz, Naomi Gleit and Javier Olivan formed a team to grow Facebook using Facebook itself. Users went from about 100 million in August 2008 to almost 150 million by year end and 350 million in 2009.
They worked on registration, invites and new-user onboarding. Quote: "the product is the biggest lever to drive growth"
One shared activation goal mattered more than its exact number. Friending correlated most with retention, so the team rallied around seven friends in 10 days. Colleagues later remembered it as 10 friends in 14 days, and it still worked.
Both are points on the same retention curve. Quote: "the majority of the value is in the latter is just having extreme Clarity around the goal"
Onboarding was built to hit the friend goal. Open registration in 2006 brought people like Naomi's dad and grandma who did not know how to use it. The flow became: upload a profile picture, then find your friends via contact import and People You May Know.
The early college site needed no onboarding because students learned it from the person next to them. Find-your-friends became one of the biggest drivers of activation.
Removing macro barriers widened the audience step by step. Eligibility went from college students to high schoolers, work networks, then anyone with an email. The team also built for feature phones, worked on internet access and crowdsourced translation into over 100 languages.
Translation happened inline with ranking of the best versions, like Wikipedia. The team kept one graph rather than a separate Facebook High.
Zuckerberg shipped News Feed to everyone and trusted engagement over the revolt. News Feed launched September 5, 2006 with no opt in. Users sent 30,000 angry emails and 10% joined a protest group, but engagement soared, so Zuckerberg posted "calm down, breathe, we hear you" and added privacy controls.
A hack that highlighted profile changes in yellow had already shown demand. News Feed cut page views, hurting revenue. Within two weeks users adapted, though the lesson that users always adapt did not always hold later.
Facebook Platform thrived on the web but could not survive mobile. At f8 in May 2007 Facebook opened its API and hit 5,000 developers in two days. On mobile no app can run inside another, credits could not carry a second 30% fee, and Steve Jobs declined Zuckerberg's 2008 pitch to embed the platform in iOS.
Facebook later cut organic app distribution, and what remained was selling distribution as app install ads, a large share of early mobile ad revenue.
Permanent founder control let Zuckerberg refuse every exit. Sean Parker, ousted at Plaxo, steered Facebook to investors who accepted Zuckerberg's board control. He turned down Yahoo's $1 billion in 2006 and a Microsoft offer reported at $24 billion in 2007.
Microsoft instead took 1.6% for $240 million at a $15 billion valuation, plus exclusive international banner ad rights until 2011 to feed its ad marketplace. The stake returned 7x by the 2012 IPO.
Facebook sacrificed desktop ad revenue to rebuild around mobile feed ads. After the May 2012 IPO at $38, with no meaningful mobile revenue, the stock fell 53.5%. Sandberg moved engineering to mobile and accepted missed quarters; by Q4 2013 mobile was 53% of ad revenue and total ad revenue grew 76%.
Feed ads fill the phone screen, so they had to be good, which made a more valuable unit than desktop ever had. Sandberg: "well Mark nobody can fire you and only you can fire me so if you're in I'm in"
Meta buys or copies rival mechanics and runs them on its own AI. Zuckerberg's 2012 email framed buying Instagram as buying time to deploy rivals' mechanics at scale. The AI lab started in 2013 for feed and ad ranking later let Reels answer TikTok quickly.
For live video it reportedly put 300 people on the copy, built live into the core product and paid streamers like Kevin Hart. Live worked as a layer on an asynchronous product, while live-only Periscope failed.
The 5 Biggest Mistakes First-Time App Founders Make
Facebook began by letting ordinary people open a profile page and connect in two or three steps. Its complexity today comes from about 20 years of growth.
Copy the Harvard-only Facebook, not the Facebook of today. Facebook began by letting people open a profile and connect with classmates in two or three steps; its polish comes from about 20 years of stacked features. Founders who copy the finished product overbuild.
Pick a comically narrow first customer, as Facebook did with college students and Amazon with books. Mark Pincus makes the same case from Zynga, which began as a single Facebook poker game. The founders were still working on side projects like webcam gestural controls while it grew.
The original growth hacker reveals his secrets | Sean Ellis (author of “Hacking Growth”)
Switched its North Star from monthly to daily active users after reaching 1 billion monthly users, which gave the team reason to bring people back every day and made it more addictive.
Switching the North Star to daily users pushed daily return. After reaching 1 billion monthly users, Facebook moved from monthly to daily active users, giving the team a reason to bring people back every day.
Sean Ellis uses it as an example of a North Star that counts delivered value, not revenue.
How to consistently go viral: Nikita Bier’s playbook for winning at consumer apps
Put 80 people on the TBH deal against Nikita alone, and works like an academic lab for social networks. Large-company incentives reward copying proven apps, and incumbents take 12 to 24 months to respond to a competitor.
Big-company incentives favor copying proven apps, slowly. Nikita Bier says incumbents take 12 to 24 months to respond, because staff protecting reviews pitch copies of a number one app they do not own. Facebook put 80 people on the TBH deal against him alone.
That delay is the opening for startups with novel ideas.
Microsoft Volume II: The Complete History and Strategy of the Ballmer Years
Microsoft paid $240 million for 1.6% in October 2007 and got exclusive international banner ad rights until 2011. The stake returned 7x by the 2012 IPO and 14x by 2014.
Permanent founder control let Zuckerberg refuse every exit. Sean Parker, ousted at Plaxo, steered Facebook to investors who accepted Zuckerberg's board control. He turned down Yahoo's $1 billion in 2006 and a Microsoft offer reported at $24 billion in 2007.
Microsoft instead took 1.6% for $240 million at a $15 billion valuation, plus exclusive international banner ad rights until 2011 to feed its ad marketplace. The stake returned 7x by the 2012 IPO.
Unorthodox frameworks for growing your product, career, and impact | Bangaly Kaba (YT, IG, FB)
Bangaly owned people recommendations there and found that profile fields failed in India, where friends shared about 7 mutual friends versus 22 in the US. Engineers visited homes in Delhi and Mumbai, and common names like Amit Kumar (250,000 people a month) made search useless.
In India, engineers watched people friend each other at home. People recommendations underperformed in India, where friends shared about 7 mutual friends versus 22 in the US. Bangaly Kaba took engineers to homes in Delhi and Mumbai every three months.
Profile fields were irrelevant, so people checked photos, and common names made search useless: Amit Kumar matched about 250,000 people a month. The team then designed around name frequency.
Choose and leave roles by learning and impact. Deng joined Facebook to learn the most and moved teams roughly every two and a half years over nine and a half years. Bangaly Kaba scores manager, resources, scope, team, pay and culture yearly.
Too much scope and several manager changes made Kaba leave a great Facebook team.
Be fundamentally different, not incrementally better | Jag Duggal (Nubank, Facebook, Google)
Jag led monetization of video and third-party content there. Fidji Simo's shift to "mobile native" in Facebook Ads is his analogy for becoming AI native.
Facebook Ads' shift to mobile native is a template for AI native. Jag Duggal cites Fidji Simo's mobile native push to argue for building AI at the heart of a product, not its corners.
Twitter’s ex-Head of Product on Elon, consumer products, culture, more | Kayvon Beykpour
It copied live video by reportedly putting 300 people on it, built live into the core product, and paid prolific streamers like Kevin Hart to stream exclusively. Kayvon calls its execution brilliant.
Meta buys or copies rival mechanics and runs them on its own AI. Zuckerberg's 2012 email framed buying Instagram as buying time to deploy rivals' mechanics at scale. The AI lab started in 2013 for feed and ad ranking later let Reels answer TikTok quickly.
For live video it reportedly put 300 people on the copy, built live into the core product and paid streamers like Kevin Hart. Live worked as a layer on an asynchronous product, while live-only Periscope failed.
The Anatomy of a High-Performing UGC Ad
Its SDK must be installed so Meta can report results and find people likely to install the app. Its Ads Library lets you keep only active ads and scroll to the oldest to spot proven performers.
Install the SDK and verify Meta's events against App Store Connect. The Facebook SDK reports results so Meta can find likely installers. Send install, paywall view, checkout, trial and subscribe events, then compare counts with App Store Connect for the same dates.
In Steve's check Meta showed about 276 trial starts vs 312, and a little over 4,000 installs vs about 3,000: close enough to trust. Aim for near 100% install-to-paywall-view.
Mine the Meta Ads Library for long-running competitor ads. Filter to active ads and find the oldest; an ad running for months is probably profitable. A Cal AI ad started June 3rd was still running about four months later.
It was a 9-second video with a calorie-tracking hook, in-app demo and App Store end card. Study complementary apps too and save examples with start dates.
The TikTok Ads Strategy for Beginners
Winning TikTok creatives ran as Instagram Reels through Facebook ads and performed almost the same. A campaign getting 10,000 users at a $1 CAC on TikTok transferred.
Winning TikTok ads transfer to Instagram Reels almost unchanged. A campaign getting 10,000 users at a $1 CAC on TikTok performed about the same run as Reels through Facebook ads. Ladder also saw Meta perform well during a TikTok outage.
The Simplest Facebook Ads Strategy for App Installs
It needs about 50 events in a rolling seven days to optimize. Broad targeting plus strong creative scales best, and Advantage Plus beats manual targeting but can over-expand a lookalike. Dynamic placements across Instagram, Messenger and Facebook now work best.
Cheap early results at small budgets are not product-market fit. At a few hundred dollars a day the algorithm finds the low-hanging fruit, so an MVP looks like it fits. Companies raise, scale to about $10,000 a day, and costs blow up.
Optimize as far down the funnel as event volume allows. Meta delivers whatever you optimize for, so install-optimized campaigns get cheap installs that never pay. Facebook needs at least 50 events in a rolling seven days.
Steve goes straight to trial starts if volume allows. Vivek Girotra starts at installs to gather signal, then walks down to trial start and subscription; the event must be recent, frequent and correlated with revenue.
Go broad on targeting and let creative and placements do the work. Interest targeting never delivered scale; broad targeting with strong creative does. Use dynamic placements across Facebook, Instagram, Messenger and Audience Network.
Seed lookalikes with first-party lists and test 1%, 5% and 10%. Keep Advantage Plus expansion within 50% of your estimated audience; 4 million growing to 15 million is too far.
Don't count on measurable retargeting on iOS. With no device ID there is no suppression list, so ads reach people who already have the app. Use lookalikes for prospecting and custom audiences for retargeting, where Android has more options.
Reflections on a movement | Eric Ries (creator of the Lean Startup methodology)
Its founders were working on other projects, such as webcam gestural controls, while it was already growing. Dustin Moskovitz taught himself to program to keep up with demand at new schools, showing a vision discovered along the way.
Copy the Harvard-only Facebook, not the Facebook of today. Facebook began by letting people open a profile and connect with classmates in two or three steps; its polish comes from about 20 years of stacked features. Founders who copy the finished product overbuild.
Pick a comically narrow first customer, as Facebook did with college students and Amazon with books. Mark Pincus makes the same case from Zynga, which began as a single Facebook poker game. The founders were still working on side projects like webcam gestural controls while it grew.
Becoming evidence-guided | Itamar Gilad (Gmail, YouTube, Microsoft)
Its fast growth and the hours people spent on it frightened Google into building Google+. Facebook itself went to a 'code DEFCON one' response, yet neither company's ad revenue was affected.
Facebook's rise scared Google into an all-in Google+ bet. Google built Google+ with about 1,000 people and integrated it into Gmail for years; it shut down in 2019. Facebook went to a "code DEFCON one" response, yet neither company's ad revenue moved.
Itamar Gilad's lesson is not to go all in on an opinion-backed idea, however strategic.
Picking sharp problems, increasing virality, and unique product frameworks | Oji Udezue (Typeform)
Its network effect makes it more useful to passive members as more people join. Its onboarding asks who you know, while Twitter asks who you want to know.
Lessons from scaling Ramp | Sri Batchu (Ramp, Instacart, Opendoor)
Used monthly active users as its North Star and 10 friends in seven days as its activation metric. It ran long-term holdouts per surface area and can reach statistical significance in two hours.
How the Gaming Ecosystem is Approaching Growth in 2023
Has grown expensive in recent years. Optimizing for purchases pushes CPI up and needs more spend to cross SKAN thresholds, so results should be judged with SKAN or the MMP.
The science of product, big bets, and how AI is impacting the future of music | Gustav Söderström
Its single news feed drew anger at launch, then solved the chore of collecting events by hand. Used as an example of a redesign that upset users even though it was better.
The ultimate guide to adding a PLG motion | Hila Qu (Reforge, GitLab)
Made the aha moment famous with its rule of adding 10 friends in 7 days. As a consumer product it is product-led by default because lifetime value cannot pay for a sales team.
Competing with giants: An inside look at how The Browser Company builds product | Josh Miller (CEO)
Josh calls it one of the best-executing organizations, but numbers drove every debate. It asked how often people post each week while Snapchat rose, and it split user research into a service for PMs.
Mastering onboarding | Lauryn Isford (Head of Growth at Airtable)
Lauryn was a product growth lead there, working on the growth of Facebook in India before moving to B2B software.
How to build your product strategy stack | Ravi Mehta (Tinder, Facebook, Tripadvisor, Outpace)
Ravi worked there as a product director before moving to Tinder.
An inside look at how the New York Times builds product | Alex Hardiman (CPO, the New York Times)
News ranking there could only train on engagement, never on the quality of a piece, and the team controlled software and distribution but not content. After the 2016 election it struggled to separate trusted journalism from misleading content, and its focus on scale, engagement and revenue differed from the Times mission.
Building Substack | Sachin Monga (Substack, Facebook)
Its "people you may know" box drove non-trivial growth after 2011, when it had just over 500 million users, but Substack rejected that model because writers wouldn't control it. Its teams were reorganized every three to six months and its trade-offs were often permanent.
When to invest in new acquisition channels | Adam Grenier (Uber, MasterClass)
It ended the notification feature that gave Zynga and Zoosk early growth. Its first 18 months of mobile installs felt emerging, with weekly product changes and broken tracking, and few companies should chase new channels before getting volume from it and Google.
AdMob Mediation
Along with Google, one of the two partners that rarely cause payment trouble. Developers pay it up front for user acquisition and then wait for ad revenue.
How to Increase AdMob eCPM for Growth Apps (Part 2/3)
Sid calls it the obvious first partner after AdMob, with the largest ad demand and good global eCPMs. In a waterfall it sits as one step between AdMob and AppLovin.
Standard Oil Part II (Extended Cut)
Its roll-up looks most like Standard Oil, with buying Instagram and WhatsApp resembling the Cleveland massacre. Its silence after the Wall Street Journal's Facebook Files parallels Standard's silence after Tarbell.
App Store Optimization for Apps Ready to Scale (Part 3/4)
When Apple Search Ads showed no impressions for a client, the team tested five interest categories on Facebook and checked which converted best. The winners guided keyword choice for ASO.
How to Get Started with Facebook Automated App Ads
The easiest, fastest and likely cheapest way to test a new app. Launch with 10 to 20 creatives, a couple hundred dollars a day per campaign per platform, and start on install optimization before moving to purchase. It hides placement and most demographic data, so geo and creative are the main levers.
Avoid These App Launch Mistakes (Part 4) - App Startup Series
Started only at Harvard, an example of starting small, and it is paid in user data rather than charging users, which small developers cannot copy.
The Facebook Retargeting Ads That Drives App Revenues
Its algorithm shifts budget to winning ads, so ads should run up to seven days before being turned off. Campaigns can optimize for installs, trials or purchases, with events sent through the Facebook SDK.
Do This Before You Advertise Your Mobile App
Lets you search any advertiser and filter by country, impressions, platform and date range up to 90 days back. It only covers Facebook and Instagram, since Meta owns both.
Video Ads Example for Facebook & Mobile Advertising
TikTok-style native videos run well in Facebook feeds, with square crops working for feed placements. The platform supplies its own call-to-action buttons like learn more or play games.
Chad Mureta's App Marketing Strategy (App Empire)
Chad ran zero outside ads until about six months before the episode, then invested heavily with a strong media buyer. Giving Facebook all the data lets it optimize for him, and he expects it to stay around.
Analyzing a Keyboard App Making $300K a Month - Grammarly App Review
Showed Grammarly's short app-preview ads, which Steve used to study its creative style such as the game-show ad asking viewers to find errors.
To Increase App Subscriptions You Need to Look At This Data with Lior Barak
Take impressions and clicks from the ad partner, such as Facebook, instead of blending them with attribution tool clicks, since sources never match. A monthly review example showed more budget should have gone to Facebook because Snapchat soaked up spend.
Storytelling with Facebook Ads
The story spread on Facebook and Instagram, and the Facebook SDK inside the app was the only way to measure whether a campaign worked. Daniela's separate Facebook page reached 140,000 followers in eight months.
How to Market a Paid Niche App & 10X Revenues
Preferred for a niche business app because targeting is specific. Employees of chosen companies can be targeted around year-end performance review season.
The Easiest Way to Drive Downloads & Revenue for a Paid App
Used as the cost comparison: it might take $500 of ad spend to bring in $1,000, while the paid-to-free campaign had no acquisition cost.
App Marketing Strategies: The Most Effective Campaigns
Well optimized campaigns can get cost per install under $1 and scale to millions of downloads if the budget allows. It costs more than the tier one tactics.
More in Social & Dating
Social & DatingA photo and video sharing social app built around creators, Stories, Reels and feeds.
38 sources · 79 tips
Social & DatingTikTok
A short-form video app where creators post content and can earn through a revenue-sharing program.
35 sources · 75 tips
Social & DatingA professional networking platform for jobs, hiring and career content.
15 sources · 55 tips
Social & DatingTinder
A dating app with tiered subscriptions plus boosts, Super Likes and other in-app purchases.
11 sources · 47 tips
Social & DatingSnap
The company behind Snapchat, a camera-first messaging app popular with teens.
6 sources · 29 tips
Social & DatingDiscord
A chat and voice platform for gaming friends and communities.
10 sources · 29 tips
Social & DatingSnapchat
A camera-first messaging and social app built around close friends, Stories and AR lenses.
10 sources · 23 tips
Social & DatingA social network for short public posts and real-time conversation.
11 sources · 29 tips