Chess.com
An online chess platform for playing, learning and reviewing games with engine analysis.
Chess.com sells a feature-specific upsell once free users hit the one-game-review-a-day limit, and it converts users months or years after sign-up at roughly $80 to $100 a year using contextual upsells and lifecycle messaging. Its network effects deter rivals, and subscribers can stay for decades. Lesson: gate one valued feature contextually, keep nurturing free users, or pick a niche utility.
- 1Free user reviews a gameOne free game review a day
- 2Asks for a second reviewHits the daily free limit
- 3Sees feature-specific paywallGame review only, not general upgrade
- 4Subscribes laterMonths or years after sign-up
- 5Pays yearlyRoughly $80 to $100 a year
From Chess.com page notes
What was said about Chess.com
1How Chess.com caught the 2020 chess boom
Years of chess businesses came before the platform itself. Co-founders Erik and Jay met in college, where Jay ran the chess club. Erik had already run a chess teaching business for about a hundred elementary schools and Wholesale Chess, an SEO-driven online shop for chess equipment.
They bought the chess.com domain in 2005 at a bankruptcy auction for roughly $50,000. Older sites like the Internet Chess Club existed but lacked a strong user experience and community. Asked for one word to explain the success, Erik said "passion".
A decade of product and infrastructure work let it absorb the boom. When the pandemic and The Queen's Gambit (October 2020) hit, new registrations more than quadrupled. Much of the engineering roadmap went into scaling infrastructure so the service stayed reliable.
The wave peaked in late 2020 and early 2021. Usage then fell back but settled at a much higher level, and the user mix shifted toward more women. Chess.com now runs north of 20 million games a day, and Cheng says the infrastructure can scale another 10x.
The next phase is steady product-led growth instead of viral waves. With 200 million registered users, 40 to 50 million monthly actives, about 10 million daily actives and about 1.5 million subscribers, Chess.com now focuses on experiments that lift retention and conversion every day.
Cheng sees his job as making growth "less lumpy". The target is 100 million daily players, which means attracting people who have never played chess, through a better first experience and a personal AI coach.
2Making losing feel better: the game review flip
Game review was built for losses, but 80% of reviews follow wins. Chess.com flipped the post-loss experience to show brilliant moves and an encouraging coach message instead of blunders. Game reviews grew 25% and subscriptions 20%.
Retention also improved "by a lot". The finding shows why intuition alone leaves gains unclaimed: typical experiment win rates are only 30 to 50%.
One positivity win was spread across the whole product. After the flip, other teams audited their own cold experiences, such as red buttons, blunder callouts and harsh game-over screens. Positivity tests in copy, design and logic have repeatedly improved retention over the past year.
The puzzles team tuned success ratings so they are not too low across skill levels, plus copy and button colors. There is a dedicated Slack channel for finding cold experiences. The PM who ran the original test only had to state the hypothesis and finding clearly; the growth leader got others to swarm on it.
Beginners lose early, and a loss cuts retention 10%. Over 75% of new users say they are complete beginners and fewer than a third win their first game. Retention is 10% worse after a loss than after a win.
Chess.com is testing a guided learn-to-play experience instead of a live first game, and hiding ratings for the first five games so people don't watch them plummet. Staff who play a lot can have a blind spot for how frustrating chess is for newcomers.
3Game review as the reason to pay
The core loop of play, review, play again drives upgrades. Users play free, review a game to see strengths and blunders, then play again. They upgrade because they want unlimited reviews.
This loop underpins about 1.5 million subscribers out of 200 million registered users. Quote: "Then you upgrade because you want unlimited access to reviews, right?"
Asking for a second daily review triggers a review-only paywall. Free users get one game review a day. Requesting another shows an upsell about game review specifically, not a general upgrade pitch.
Phil Carter cites it as a contextual paywall working at scale: use the general pitch at onboarding and app launch, and a templated, feature-specific pitch when a user taps a locked feature.
Most conversions happen months or years after sign-up. Chess.com stays premium at roughly $80 to $100 a year by converting free users long after registration through contextual upsells and lifecycle messaging.
The lesson is to tailor the offer to what you learned about a user's usage rather than repeat the onboarding paywall. Subscribers can stay for decades.
4Building an experimentation engine toward 1,000 tests a year
Fix tracking and metrics before asking teams to experiment. Chess.com ran almost no experiments before 2023 and began in earnest in 2024. Cheng's first work was auditing the platform, the metrics and which teams were comfortable testing.
He set up event tracking, an experimentation system, a growth model and key metrics so people could trust results. At another company retention had been configured backwards, so every positive result was really negative.
A 1,000-experiment target forces investment in the system. Volume went from near zero to about 50, then about 250 a year, with a target of 1,000. The goal's value is the conversation about what must be true to hit it.
That means no-code configurable screens like home and pricing, lifecycle marketing testing constantly, experiments on app store screenshots and from content teams, engineers launching their own tests, fast analysis and good observability.
Expect 30 to 50% of tests to win, and take bigger swings. Cheng sees a 30 to 50% hit rate, with about 30% clear ship decisions. Fresh areas like a new monetization team win more; retention work wins less.
Chasing 60 to 80% hit rates pushes teams into micro-optimizations. Bigger swings either surprise with a big win or teach something. "A lot of companies are too afraid to take bigger swings and fail."
Rising flat results signal it's time to explore again. Cheng alternates between exploring for insights and exploiting them across the product. When more tests stop reaching significance, he brings the team back to brainstorm divergently.
An experiment explorer tool lets him look across all running tests for links between hypotheses and learnings. Too much exploring is scattershot; too much exploiting becomes local maximizing.
Visible wins and losses build the testing culture. Experiments are posted in Slack and all-hands as show-and-tell: hypothesis, result, impact and learning, tied with real data to revenue or active users.
Founder buy-in matters so the growth leader is not at odds with the founders. A hugely losing experiment is as valuable to share as a winner. "People have to see it working."
Slack bots answer data questions so more get asked. Chess.com is training text-to-SQL Slack bots that give first-pass answers to ad hoc questions, doing the analysis and not only the query.
Requests like subscribers in South Africa used to wait on an analyst. People skip questions that feel embarrassing or bothersome, so an easy tool brought a large explosion of questions.
AI prototypes of core screens make ideas testable faster. The team built prototypes of onboarding, the home screen and the chessboard with v0 and Lovable, and others layer ideas on top.
PMs use v0, designers Figma Make, engineers Cursor, Claude Code and GitHub Copilot. Moving from tinkering to workflow is still clumsy, so functions still hand work off.
5Retaining current users and bringing back dormant ones
DAU is the north star, with current-user retention as the driver. Cheng is bringing Duolingo's growth model to Chess.com to focus a passionate team that "tends to say yes to practically everything" on the biggest pockets of growth.
Retention is what lets a consumer subscription avoid pushing for payment on day one. A day-one retention of 30 to 40% is solid for a consumer app.
Resurrected users are as big a group as new users. About 80% of daily or weekly actives are existing users, and new and resurrected users are about the same size. Dormant users can number in the hundreds of millions.
Cheng argues the returning experience deserves real investment. At Duolingo this meant social pushes when a synced friend started and a new placement test on return.
6Growing without paid ads: creators, crossover events and social clips
Creators and non-chess celebrities bring beginners in. Danny Rensch's team built ties with Hikaru, Magnus, GothamChess and the Botez sisters. PogChamps has non-chess personalities like Mr. Beast and Rainn Wilson play and stream, producing meaningful user growth.
YouTube is a top three channel in the how-did-you-hear-about-us survey. Creators drive acquisition, engagement and the return of lapsed users.
Short-form clips carry chess into mainstream culture. Chess.com posts viral clips on TikTok, Instagram, YouTube Shorts and X, with social projected at about three billion impressions this year. Short-form is the stickiest.
Magnus Carlsen pounding the table against Gukesh at Norway Chess became a viral meme, making a game that can feel intimidating or dry feel relatable.
A how-did-you-hear survey shows which moments move registrations. A sampled onboarding question is directional, but after viral posts a channel can climb from about 4% to 10% of responses in a week.
During the world championship, registrations in India passed registrations in the US.
Generic paid ads failed the LTV to CAC test. Small Facebook and Google tests did not make the economics work, so paid is not in the mix. Owning the domain already ranks it high in search and app stores.
The aim is to put chess itself in the culture; the Google Trends curve for "chess" is a proxy for its demand. Chess.com may later retarget lapsed subscribers with offers, but will never rely mainly on performance marketing.
After the boom, events were cut to fewer, better formats. Chess.com scaled back expensive tournaments and broadcasts, keeping the Speed Chess Championship, the weekly Titled Tuesday and two or three big tournaments a year.
Speed Chess uses short games, strong announcers and game review to help viewers follow. The broadcast effort had mixed success. "We've dialed it down, but we've sharpened it."
7AI coaches and the network-effect moat
LLMs handle coaching conversation while chess engines do analysis. Chess.com plans virtual coaches with different personalities in game review, puzzles and lessons, because language models are not good at playing chess. It uses ElevenLabs to power a voice chess tutor.
AI is also used heavily in fair play detection, judging how likely a human at a given skill level is to repeatedly find the engine's best move.
Its network effects make head-on competition a poor bet. Andrew Maguire calls Chess.com nearly impossible to disrupt, which has kept VCs from funding rivals.
He sees niche chess apps that solve a specific problem as a better lane, able to reach $10M a year with a small bootstrapped team.
Finding hidden growth opportunities in your product | Albert Cheng (Duolingo, Grammarly, Chess.com)
Went from almost no experiments before 2023 to about 50, then about 250, with a target of 1,000. Showing best moves after a loss grew game reviews 25% and subscriptions 20%, and over 75% of new users are beginners, with retention 10% worse after a loss.
Asking for a second daily review triggers a review-only paywall. Free users get one game review a day. Requesting another shows an upsell about game review specifically, not a general upgrade pitch.
Phil Carter cites it as a contextual paywall working at scale: use the general pitch at onboarding and app launch, and a templated, feature-specific pitch when a user taps a locked feature.
How Chess.com Became the World’s Top Chess App
Bought the domain in 2005 for roughly $50,000 and now runs over 20 million games a day. It has 200 million registered users, 10 million daily actives and about 1.5 million subscribers. Experimentation only began in earnest in 2024, with a goal of 1,000 experiments a year and 100 million daily players.
Most conversions happen months or years after sign-up. Chess.com stays premium at roughly $80 to $100 a year by converting free users long after registration through contextual upsells and lifecycle messaging.
The lesson is to tailor the offer to what you learned about a user's usage rather than repeat the onboarding paywall. Subscribers can stay for decades.
Its network effects make head-on competition a poor bet. Andrew Maguire calls Chess.com nearly impossible to disrupt, which has kept VCs from funding rivals.
He sees niche chess apps that solve a specific problem as a better lane, able to reach $10M a year with a small bootstrapped team.
How Tolan 3X'd Subscriber Conversion after Shifting to Freemium
Free users get one game review a day, and asking for another triggers an upsell about game review only. This shows a contextual paywall working at scale.
LLMs handle coaching conversation while chess engines do analysis. Chess.com plans virtual coaches with different personalities in game review, puzzles and lessons, because language models are not good at playing chess. It uses ElevenLabs to power a voice chess tutor.
AI is also used heavily in fair play detection, judging how likely a human at a given skill level is to repeatedly find the engine's best move.
The Bootstrapper's Path to $10M ARR Without Venture Capital – Andrew Maguire, Volo Ventures
Its network effects make it nearly impossible to disrupt and kept VCs from funding rivals. Niche chess apps could still reach $10M a year by solving a specific problem, and subscribers can stay for decades.
Most conversions happen months or years after sign-up. Chess.com stays premium at roughly $80 to $100 a year by converting free users long after registration through contextual upsells and lifecycle messaging.
The lesson is to tailor the offer to what you learned about a user's usage rather than repeat the onboarding paywall. Subscribers can stay for decades.
How ElevenLabs is Building Consumer Apps on its $11B Voice AI Platform
It uses ElevenLabs to power a voice chess tutor.
Game review was built for losses, but 80% of reviews follow wins. Chess.com flipped the post-loss experience to show brilliant moves and an encouraging coach message instead of blunders. Game reviews grew 25% and subscriptions 20%.
Retention also improved "by a lot". The finding shows why intuition alone leaves gains unclaimed: typical experiment win rates are only 30 to 50%.
One positivity win was spread across the whole product. After the flip, other teams audited their own cold experiences, such as red buttons, blunder callouts and harsh game-over screens. Positivity tests in copy, design and logic have repeatedly improved retention over the past year.
The puzzles team tuned success ratings so they are not too low across skill levels, plus copy and button colors. There is a dedicated Slack channel for finding cold experiences. The PM who ran the original test only had to state the hypothesis and finding clearly; the growth leader got others to swarm on it.
Beginners lose early, and a loss cuts retention 10%. Over 75% of new users say they are complete beginners and fewer than a third win their first game. Retention is 10% worse after a loss than after a win.
Chess.com is testing a guided learn-to-play experience instead of a live first game, and hiding ratings for the first five games so people don't watch them plummet. Staff who play a lot can have a blind spot for how frustrating chess is for newcomers.
Fix tracking and metrics before asking teams to experiment. Chess.com ran almost no experiments before 2023 and began in earnest in 2024. Cheng's first work was auditing the platform, the metrics and which teams were comfortable testing.
He set up event tracking, an experimentation system, a growth model and key metrics so people could trust results. At another company retention had been configured backwards, so every positive result was really negative.
A 1,000-experiment target forces investment in the system. Volume went from near zero to about 50, then about 250 a year, with a target of 1,000. The goal's value is the conversation about what must be true to hit it.
That means no-code configurable screens like home and pricing, lifecycle marketing testing constantly, experiments on app store screenshots and from content teams, engineers launching their own tests, fast analysis and good observability.
Rising flat results signal it's time to explore again. Cheng alternates between exploring for insights and exploiting them across the product. When more tests stop reaching significance, he brings the team back to brainstorm divergently.
An experiment explorer tool lets him look across all running tests for links between hypotheses and learnings. Too much exploring is scattershot; too much exploiting becomes local maximizing.
Visible wins and losses build the testing culture. Experiments are posted in Slack and all-hands as show-and-tell: hypothesis, result, impact and learning, tied with real data to revenue or active users.
Founder buy-in matters so the growth leader is not at odds with the founders. A hugely losing experiment is as valuable to share as a winner. "People have to see it working."
Slack bots answer data questions so more get asked. Chess.com is training text-to-SQL Slack bots that give first-pass answers to ad hoc questions, doing the analysis and not only the query.
Requests like subscribers in South Africa used to wait on an analyst. People skip questions that feel embarrassing or bothersome, so an easy tool brought a large explosion of questions.
AI prototypes of core screens make ideas testable faster. The team built prototypes of onboarding, the home screen and the chessboard with v0 and Lovable, and others layer ideas on top.
PMs use v0, designers Figma Make, engineers Cursor, Claude Code and GitHub Copilot. Moving from tinkering to workflow is still clumsy, so functions still hand work off.
DAU is the north star, with current-user retention as the driver. Cheng is bringing Duolingo's growth model to Chess.com to focus a passionate team that "tends to say yes to practically everything" on the biggest pockets of growth.
Retention is what lets a consumer subscription avoid pushing for payment on day one. A day-one retention of 30 to 40% is solid for a consumer app.
Resurrected users are as big a group as new users. About 80% of daily or weekly actives are existing users, and new and resurrected users are about the same size. Dormant users can number in the hundreds of millions.
Cheng argues the returning experience deserves real investment. At Duolingo this meant social pushes when a synced friend started and a new placement test on return.
Insights from RevenueCat’s 2026 State of Subscription Apps Report
Stays premium by converting users months or years after sign-up, at roughly $80 to $100 a year, using contextual upsells and lifecycle messaging.
Years of chess businesses came before the platform itself. Co-founders Erik and Jay met in college, where Jay ran the chess club. Erik had already run a chess teaching business for about a hundred elementary schools and Wholesale Chess, an SEO-driven online shop for chess equipment.
They bought the chess.com domain in 2005 at a bankruptcy auction for roughly $50,000. Older sites like the Internet Chess Club existed but lacked a strong user experience and community. Asked for one word to explain the success, Erik said "passion".
A decade of product and infrastructure work let it absorb the boom. When the pandemic and The Queen's Gambit (October 2020) hit, new registrations more than quadrupled. Much of the engineering roadmap went into scaling infrastructure so the service stayed reliable.
The wave peaked in late 2020 and early 2021. Usage then fell back but settled at a much higher level, and the user mix shifted toward more women. Chess.com now runs north of 20 million games a day, and Cheng says the infrastructure can scale another 10x.
The next phase is steady product-led growth instead of viral waves. With 200 million registered users, 40 to 50 million monthly actives, about 10 million daily actives and about 1.5 million subscribers, Chess.com now focuses on experiments that lift retention and conversion every day.
Cheng sees his job as making growth "less lumpy". The target is 100 million daily players, which means attracting people who have never played chess, through a better first experience and a personal AI coach.
One positivity win was spread across the whole product. After the flip, other teams audited their own cold experiences, such as red buttons, blunder callouts and harsh game-over screens. Positivity tests in copy, design and logic have repeatedly improved retention over the past year.
The puzzles team tuned success ratings so they are not too low across skill levels, plus copy and button colors. There is a dedicated Slack channel for finding cold experiences. The PM who ran the original test only had to state the hypothesis and finding clearly; the growth leader got others to swarm on it.
Beginners lose early, and a loss cuts retention 10%. Over 75% of new users say they are complete beginners and fewer than a third win their first game. Retention is 10% worse after a loss than after a win.
Chess.com is testing a guided learn-to-play experience instead of a live first game, and hiding ratings for the first five games so people don't watch them plummet. Staff who play a lot can have a blind spot for how frustrating chess is for newcomers.
The core loop of play, review, play again drives upgrades. Users play free, review a game to see strengths and blunders, then play again. They upgrade because they want unlimited reviews.
This loop underpins about 1.5 million subscribers out of 200 million registered users. Quote: "Then you upgrade because you want unlimited access to reviews, right?"
Fix tracking and metrics before asking teams to experiment. Chess.com ran almost no experiments before 2023 and began in earnest in 2024. Cheng's first work was auditing the platform, the metrics and which teams were comfortable testing.
He set up event tracking, an experimentation system, a growth model and key metrics so people could trust results. At another company retention had been configured backwards, so every positive result was really negative.
A 1,000-experiment target forces investment in the system. Volume went from near zero to about 50, then about 250 a year, with a target of 1,000. The goal's value is the conversation about what must be true to hit it.
That means no-code configurable screens like home and pricing, lifecycle marketing testing constantly, experiments on app store screenshots and from content teams, engineers launching their own tests, fast analysis and good observability.
Expect 30 to 50% of tests to win, and take bigger swings. Cheng sees a 30 to 50% hit rate, with about 30% clear ship decisions. Fresh areas like a new monetization team win more; retention work wins less.
Chasing 60 to 80% hit rates pushes teams into micro-optimizations. Bigger swings either surprise with a big win or teach something. "A lot of companies are too afraid to take bigger swings and fail."
Rising flat results signal it's time to explore again. Cheng alternates between exploring for insights and exploiting them across the product. When more tests stop reaching significance, he brings the team back to brainstorm divergently.
An experiment explorer tool lets him look across all running tests for links between hypotheses and learnings. Too much exploring is scattershot; too much exploiting becomes local maximizing.
Visible wins and losses build the testing culture. Experiments are posted in Slack and all-hands as show-and-tell: hypothesis, result, impact and learning, tied with real data to revenue or active users.
Founder buy-in matters so the growth leader is not at odds with the founders. A hugely losing experiment is as valuable to share as a winner. "People have to see it working."
DAU is the north star, with current-user retention as the driver. Cheng is bringing Duolingo's growth model to Chess.com to focus a passionate team that "tends to say yes to practically everything" on the biggest pockets of growth.
Retention is what lets a consumer subscription avoid pushing for payment on day one. A day-one retention of 30 to 40% is solid for a consumer app.
Creators and non-chess celebrities bring beginners in. Danny Rensch's team built ties with Hikaru, Magnus, GothamChess and the Botez sisters. PogChamps has non-chess personalities like Mr. Beast and Rainn Wilson play and stream, producing meaningful user growth.
YouTube is a top three channel in the how-did-you-hear-about-us survey. Creators drive acquisition, engagement and the return of lapsed users.
Short-form clips carry chess into mainstream culture. Chess.com posts viral clips on TikTok, Instagram, YouTube Shorts and X, with social projected at about three billion impressions this year. Short-form is the stickiest.
Magnus Carlsen pounding the table against Gukesh at Norway Chess became a viral meme, making a game that can feel intimidating or dry feel relatable.
A how-did-you-hear survey shows which moments move registrations. A sampled onboarding question is directional, but after viral posts a channel can climb from about 4% to 10% of responses in a week.
During the world championship, registrations in India passed registrations in the US.
Generic paid ads failed the LTV to CAC test. Small Facebook and Google tests did not make the economics work, so paid is not in the mix. Owning the domain already ranks it high in search and app stores.
The aim is to put chess itself in the culture; the Google Trends curve for "chess" is a proxy for its demand. Chess.com may later retarget lapsed subscribers with offers, but will never rely mainly on performance marketing.
After the boom, events were cut to fewer, better formats. Chess.com scaled back expensive tournaments and broadcasts, keeping the Speed Chess Championship, the weekly Titled Tuesday and two or three big tournaments a year.
Speed Chess uses short games, strong announcers and game review to help viewers follow. The broadcast effort had mixed success. "We've dialed it down, but we've sharpened it."
LLMs handle coaching conversation while chess engines do analysis. Chess.com plans virtual coaches with different personalities in game review, puzzles and lessons, because language models are not good at playing chess. It uses ElevenLabs to power a voice chess tutor.
AI is also used heavily in fair play detection, judging how likely a human at a given skill level is to repeatedly find the engine's best move.
More in Games
Paintball Rush
A mobile paintball game used as the worked example of a feature pitch.
7 sources · 14 tips
GamesColor Switch
A mobile arcade game where players pass a ball through obstacles by matching its color.
6 sources · 19 tips
GamesCrossy Road
An endless arcade hopper mixing Frogger and Flappy Bird with collectible voxel characters.
9 sources · 21 tips
GamesWordle
A news organization selling a bundled digital subscription of journalism, games, cooking, sports, audio and shopping.
2 sources · 19 tips
GamesZynga
A social games company behind FarmVille, CityVille, Words with Friends and Zynga Poker.
5 sources · 12 tips
GamesSubway Surfers
An endless runner mobile game and the most downloaded mobile game ever.
2 sources · 11 tips
GamesBlack Box
A puzzle game of unconventional brain teasers that uses the device's features, sold as a paid app with in-app purchases.
1 source · 11 tipsFly Bye Birdie
A casual one-tap bird game with flat cartoon art, built solo by a graphic designer.
1 source · 8 tips