What really happens inside a dating app - My blog

• Women like exactly ~5% of profiles no matter who you show them—their behavior is quota-based, not quality-based, meaning your chances as a guy depend on who she saw before you
• The feed algorithm is 70% strategic filler for women (new users, paid users, ugly guys to boost subsequent likes) and only 20% people they'll actually like—it's attention allocation, not matchmaking
• Retention drivers are completely different by gender: women stay if they send 5+ likes in 24 hours; for men, almost nothing matters except getting notifications
• The "perceived attractiveness gap" ruins dating apps: men look worse in photos than real life (need professional photographers), women look better (creating inflated standards)
• Match Group isn't evil—they're just optimizing a leaky bucket where the product is entertainment, not dating, and users fundamentally don't want to put in effort to meet people

This is a rare data-driven account from someone who worked inside a dating app, revealing how the entire product is engineered around behavioral economics, not romance. The core insight: women's swiping behavior is quota-based—they'll like roughly 5% of profiles regardless of quality. Show them 100 profiles, remove the 50 ugliest, and they'll still like 5%. This means a guy's chances depend less on his own attractiveness and more on the competition in her recent feed. Men, meanwhile, like 26% of profiles and will like almost anyone if shown long enough.

The feed algorithm exploits this asymmetry. For women, only 20% of feed space goes to guys they might actually like (active, will like back, geographically close). The remaining 70% is strategic filler: new users who need initial likes, paid users, unranked profiles, and even ugly guys (some apps found this increases subsequent like rates). For men, the algorithm prioritizes showing them to new women immediately (so women get their dopamine hit of 100 likes in 24 hours) and ensuring they see any woman who liked them first. The rest doesn't matter—men's retention is barely affected by what they see.

Retention is the only metric that matters, and it's driven by completely different factors by gender. Women stay if they send 5 likes to active guys in the first 24 hours (leading to a match). For men, retention is nearly impossible to predict or optimize—likes received barely help, profile quality doesn't matter, and even limiting swipes has no effect. The author suspects men's retention is mostly driven by their activity on competing apps.

The fundamental problem is the "perceived attractiveness gap": photos make men look worse than real life and women look better than real life. This creates a doom loop where women develop inflated standards (thinking they can get anyone based on their matches) and men need professional photographers to compete. Video doesn't fix this—women refuse to post unflattering videos even if authentic.

His conclusion: dating apps aren't broken, they're working as designed. They're entertainment platforms that happen to occasionally produce dates. Match Group isn't evil—they're just the best at monetizing a leaky bucket business where users churn by nature. The solution isn't better algorithms or AI—it's removing choice entirely. Future dating apps should be paid-only, give users minimal information about matches, and force them to meet strangers based on algorithmic pairing (like Timeleft). Because the real problem isn't the product—it's that users want partners without effort, women won't lower standards, and men won't invest in quality photos.