Startup Spotlight: ‘Tinder for friends’ app Patook uses artificial intelligence to weed out flirting
Geosocial apps are having a bit of a moment. Tinder, Bumble, and a host of other services have emerged in recent years, promising to foster connections with real people, nearby. But online dating paved the way for these apps — and even supposedly platonic services like Bumble BFF have struggled to shrug off the romantic connotation. So where does that leave people who aren’t looking for love, but do want to use new technology to make friends?
That question — or “shower thought,” as he puts it — motivated Antoine Daher to create Patook, an app that connects people based on common interests. It uses artificial intelligence to detect flirtatious messages and automatically bans users who send them.
“The goal of Patook is to create an app where people can make new, interest-sharing platonic friends without having to worry about people who are looking for more,” said Daher.
The former Microsoft engineer released a closed alpha version of the friend-making platform in 2014 and then released an alpha app in 2015. This past spring of this year, he moved the app to beta and plans to officially launch in the fall.
We caught up with Daher for this Startup Spotlight, a regular GeekWire feature. Continue reading for his answers to our questionnaire.
Explain what you do so our parents can understand it: “Patook is an app that allows you to make strictly platonic friends. It’s like Tinder for friends. It uses artificial intelligence to identify people who are trying to flirt and bans them immediately, which means the people left are actually all looking for friends.”
Inspiration hit us when: “It was more of a shower thought — there were lots of dating apps on the market and lots of news about them in various outlets, but nothing really about friend-making apps. Could it be people just don’t need such an app? We ran a survey a year ago of about 2,000 random people to see how hard it was for them to make new friends once they hit adulthood. We were surprised to see that only one out of three people found it easy to make friends, and fewer than 20 percent of people found it easy to make new friends within five years of moving to a new city.
We wondered why this need was unanswered and tried out several ‘friend-making apps’ and websites. All of the ones we tried were littered with spam, fake profiles, and people who were clearly not there for actual platonic friendships. There just wasn’t any app out there that was strictly platonic. The ‘need’ was answered but it was only poor answers. So we decided to tackle the problem.”
VC, Angel or Bootstrap: “Bootstrap for now. We will look into angel funding once there is a clear signal as to which growth methods work and which do not.”
Our ‘secret sauce’ is: “Our secret sauce is the flirt detector. It is a machine-learning model trained on hundreds of thousands of flirty messages, pick-up lines, creepy messages, extracted from various sources. It analyzes user behavior and messages and figures out when a user is flirting, with very high accuracy. Flirters get banned automatically, leaving only users who are actually there to make friends.
Patook has a lot of AI happening behind the scenes. Without it, any such app left unchecked seems to almost always invite spammers and people looking for romance/more, which in turn chases away people who are looking for actual friends.”
The smartest move we’ve made so far: “The flirt detector wasn’t there in the original version, but as more and more users joined it was clear that short of moderating every single message, the site wouldn’t remain ‘friends only’ unless something was done.
We went through some subreddits in Reddit that contained creepy PMs, websites that had pick-up lines, and extracted several hundred-thousand such messages by running optical character recognition on screenshots that frustrated users had taken within their apps. Then we were able to build a classifier that would differentiate flirty messages from non-flirty ones and that stabilized website growth by kicking out flirters.”
The biggest mistake we’ve made so far: “There are plenty to choose from, but the main one is that the original app had too many features in it. While they were all useful to the overall goal of making platonic friends, they tended to overwhelm users. Almost nobody really took the time to learn how to use them. People are quick to delete apps they don’t fully understand (or don’t have the patience to). It took us a while to realize it.
As we began cutting and hiding certain features, engagement went up. The current app has fewer than half the features of the original one and we are still debating whether to axe a few more and just leave the absolute bare bones. It seems that if you have a target market as diverse as this, it’s better to focus on having one or two main features and then either hide the additional features away or do away with them entirely.”
Would you rather have Gates, Zuckerberg or Bezos in your corner: For this project, Zuckerberg would probably be the best fit as he could easily help with growth advice and has the most experience in the social field. He could probably increase our users ten-fold just by making a Facebook post about it.
Of course, it goes without saying that we’d also love Gates and Bezos in our corner.”
Our favorite team-building activity is: “Talking to the users is always fun. I actually have had coffee with a few people through the app itself and it helps me really understand what they’re looking for. Also, really love going through the ‘thank yous’ from users both through e-mail and as app store feedback.”
The biggest thing we look for when hiring is: “Skill and perseverance. The candidates have to be good enough that we don’t think ‘Oh, we could have done this ourselves much faster.’ For example, a candidate for a design job (we hired a few for contracts) must have a portfolio that impresses us and goes beyond themes we could have bought out of ThemeForest.
We usually start off with a small contract job to see how perseverant the candidates are. A lot of them will give it 100 percent for the first couple of days, then will barely respond after that. Those that keep up usually get the rest of the work – that’s how we contracted design, translations, etc.”
What’s the one piece of advice you’d give to other entrepreneurs just starting out: “If you’re targeting consumers from a non-tech population, try to only expose a couple of big features. Unless your product is famous, most people don’t want to go through any kind of onboarding or read any help dialog. If there’s more than one call to action on a page, you already lose people who think it’s too complicated. Try to funnel your users from the moment they first visit your page until the moment they have a success state through a single, short and simple path.”
Startups – GeekWire http://www.geekwire.com/2016/patook-llc/