How to Fix Mobile Dating Apps
or, Why All Dating Apps Suck
Every dating app is a garbage fire of development, product, and user experience.
How do we fix it?
Why waste our time giving free advice to companies worth hundreds of millions of dollars? Because we’re really bad at business, that’s why.
Every dating app forces users into the same lonely, lonely funnel:
- evaluate pictures (80%)
- evaluate profiles (40%)
- initiate messages (5%)
- attempt chats (3%)
- move off app into iMessage or Signal (0.5%)
Those steps were fine back in 2010, but why hasn’t anything changed with online dating since then? The only change since 2010 is now people “swipe.” That’s it. The biggest innovation in dating over the last 10 years has been moving your finger a little left or a little right on a screen.
This is pathetic.
Why No Changes?
In 2010, one top of the line GPU had 0.25 teraflops of compute power.
By 2017, one top of the line GPU had 13.8 teraflops of compute power.
Why haven’t dating app companies used our huge explosion in processing power to improve user experience?
Instead of improving user experience, online/app dating now involves:
- every app wants to charge you $12/month for an undegraded (but still degrading) on-device experience.
- every two years a new “my first CRUD site” dating app shows up targeting a new demographic or implementing a new user interaction gimmick.
Nobody has tried to improve the technical aspect of social/mobile dating since a web server scholar pivoted into an increasingly inept quiz provider.
How can we fix online dating?
Companies sit on so much personal data they could basically print you a report of who your best matches are globally, but what do they do instead? swipe right, swipe left, photo grid, infinitely frustrating bespoke messaging platforms.
Platform Exploitation Prevention
Spam and botting infect all dating apps. Only some platforms even attempt to control botting problems. Other platforms don’t even try: one user can send 1000 text graphics an hour to random profiles promoting other sites. It would be unbelievably easy to stop this simple fan-out single-image spam before it hits users, but most platforms don’t even try.
At the very minimum, a modern dating platform must take sophisticated precautions against users being interrupted with blatant advertising and adversarial messaging.
Dating by its very nature discriminatory. A 28 year old, 5-day-a-week gym-doer probably doesn’t want to receive messages from a 65 year old obese shut-in. (elder shaming! body shaming! unhealthy body image issues! yeah, yeah, get over yourself.)
We know users are going to discriminate based on their desired acquisition targets. Apps often help users discriminate based on any profile details like age, height, weight, nose hair length, etc.
How can apps help users pre-filter the global user census in a more intelligent way than just selecting database fields like a spreadsheet from 1985?
Some platforms perform limited auto-discrimination today. If your profile has more inbound activity than average, you get shown other highly active profiles.
Some dating platforms also try to match similar people through profile details or “quiz” platforms, but those have limited effectiveness due to most people:
- not answering enough questions
- not writing enough detail (or any detail) about themselves
- answering ranked questions sarcastically
How can we figure out a person’s attributes for matching if they won’t tell us directly?
Most users don’t really provide enough explicit detail at-scale to generate paired matches, which is also the motivating factor behind modern dating apps restricting everybody to “1-5 pictures and a one sentence description” profile formats.
It’s time the era of “kid’s first CRUD app” dating platforms comes to an end.
Dating apps acquire so much data we can mine:
- user interactions (swipes or browses or engages for profile details)
- what a user says
- location (obvs)
What can we infer about users given passively acquired data points?
Interaction points give us the ability to generate per-user models describing each user’s preferred attractiveness vector.
After a user selects enough pictures as desired or attractive, a dating app should be able to guess future attractive pictures from the already selected pictures. This would be a week long project for a high school intern given the current state of open source image categorizing libraries.
Though, most dating apps still haven’t even figured out how to so something as simple as automatic face detection, so modeling per-user attractiveness vectors based on profile pixel grids may be beyond them.
User text input gives us the ability to recognize how a user communicates.
Does this user ask ten word questions when chatting? Do they respond to 10 word questions with just “yah” or “cool?” Can they even communicate like a human or is it more like a rabbit bouncing on a touch screen?
Pre-filtering non-communicators trying to engage people who can actually talk would save endless frustration. People who constantly send messages with no follow up should be flagged as such and given a lower priority in messaging.
Location gives us the ability to narrow a match radius down to not only current location but routine historical locations.
Does a user always take the same route to work? Does a user always go to the same gym? We can act not only on present location, but also on location ghosts—what if you match with someone at your gym, but they attend your gym at 7pm and you go at 5pm so you’re never there at the same time?
People are begging for better dating platforms. Dating platforms have so much ambient data already being provided to them it’s inexcusable the current state of people matching is just “swipe it if you like it” for years on end with no changes or automated evolutions in sight.
Basic goals for modern dating apps:
- intelligent, sophisticated, ML-driven botting prevention
- auto-bucketing people by basic details they seem to want
- communication helpers to smooth out differences between flakes and actives
Advanced goals for modern dating apps:
- Per-user models describing:
- which pictures/profiles a user interacts with
- enabling user to create automatic people-i-find-attractive priority filters
- communication patterns
- match users who communicate similarly—i.e. don’t give someone who only says “cool, yah, whatevz” to someone else who prefers to chat online for an hour before meeting.
- which pictures/profiles a user interacts with
- Use current location and also historical routine locations for suggesting matches
- Queue “best met” people throughout the day for later review
- e.g. did you pass 12 great matches while on the subway but didn’t stick around long enough to interact with them in-app? View them later while they are hopefully in the same area, but maybe not “user is X feet away” distance.
- difficulty in finding people you successfully chatted with long ago (most platforms just show chats in chronological order in an ever growing avalanche of threads)
- lack of programmability
- why can’t users write tiny little messaging responders to default questions?
- maybe even let user responders chat with each other to see how far they can get automatically.
- lack of global map manipulation
- allow users to draw geographic interest areas/zones for matching
- allow users to draw temporal-geographic areas denoting present, past, or future locations, which would allows users to expand their range based on where they will be or where they have been. (may platforms today don’t allow “location browsing” (without GPS spoofing) in a weak attempt to limit botting/abuse spreading from region to region)
Will these changes ever happen?
I doubt any existing local mobile social dating apps have the humility to re-tool their platfors to enable modern ends. If not existing platforms, then where will these changes happen? You end up with a classic marketplace problem. Create a new platform, but have no users, so why bother? Who wants to use a new dating app with no users? Unless you’re a frat bro asshole how do you “go viral” at a scale needed for new social marketplace type deployments? Can anybody over the age of 22 even create new apps? The world may never know.
Whatever the end result, don’t sell out to the dumb butts at IAC’s Match Group.
Enjoy Being Forever Alone!