Justin McLeod: AI is ‘really going to change the game’ for dating apps

Justin McLeod is the founder and chief executive of dating app Hinge. Initially designed as a Facebook app in 2011 — while McLeod was in his final year at Harvard Business School — it launched as a mobile app in 2013 and was acquired by dating giant Match Group in 2018, making it part of the same portfolio as Tinder and OKCupid.

Hinge, which says it has set up over 30mn dates, is now available in 25 countries, and last year generated revenues of $284mn. It positions itself as “the dating app designed to be deleted” — a service that will yield a lasting relationship instead of an endless stream of dates.

In this interview, McLeod talks about the genesis of his company, the opportunities presented by artificial intelligence, Hinge’s expansion into Europe — and what makes for an effective profile.

Cristina Criddle: I used to use Hinge when it first came out, and I loved the links with social networks, so you could see who prospective dates were friends with, and then work out from there whom you had in common. Why did you drop that approach?

Justin McLeod: For a few reasons. One is that we ultimately found it limiting. We found that, over time, as we got more information on people, we were able to make better matches, and help people get outside their friend group. And the stigma reduced around meeting new people — the idea of “friends of friends” helped people feel OK in the weird new world of online dating back in 2012, ’13 and ’14, but the need for that faded.

Tech Exchange 

The FT’s top reporters and commentators hold monthly conversations with the world’s most thought-provoking technology leaders, innovators and academics, to discuss the future of the digital world and the role of Big Tech companies in shaping it. The dialogues are in-depth and detailed, focusing on the way technology groups, consumers and authorities will interact to solve global problems and provide new services. Read them all here

More practically, Facebook eliminated the ability to map through friends of friends. But mainly, we just felt we could do better than limiting ourselves to friends of friends in terms of helping people to match up.

CC: What was the thinking behind setting up Hinge?

JM: It was originally just that I wanted to create an app for people like me. At the time, dating apps weren’t even a thing — it was dating websites. The idea was just to create something very simple and fun and easy. And because it was connected through friends, that’s where the name Hinge comes from.

And then obviously some other apps came along around that time that did far better than we did in terms of being fast and fun and easy. But around 2015, 2016, I looked at the market and realised that there was a really big opportunity — that I really hadn’t built the company that I wanted to build, which was about long-term relationships and meaningful connection.

And so I rebooted the company, tore it down, and built a new app from scratch that is the Hinge that we know today — designed to be deleted.

CC: You subsequently sold it to Match. How did the company change?

JM: That was shortly after the reboot, around 2018, and we were still just getting started at that point. We were doing $1mn in revenue per year. We’re going to do about $400mn this year. And we had, I don’t know, 30 employees, and now we have 300. So we were a very different company at that time.

Match Group has been very intertwined with our growth. They were able to bring their investment and wisdom to bear, and we were able to bring our differentiated approach and our focus on long-term relationships. That was a magical combination.

CC: How was it a differentiated approach?

JM: When we did the reboot, we moved away from engagement and retention, and instead built something that was really about quality over quantity, helping people really show their personality. We invented Prompts and Voice Prompts and Your Turn and We Met [features designed to elicit a sense of users’ personalities, discourage “ghosting”, and gather feedback on dates], and the whole idea of liking a piece of content instead of just swiping on people.

[We built] a whole suite of features that were really about measuring, ultimately, whether this does get you out on more good dates or not. In fact, we were the only app to even measure whether people were going on dates or not and whether they were good. So that was what we optimised for.

I made 10,000 mistakes along the way, and, at the same time, those mistakes were learnings that allowed me to build the company that I have now

CC: What was it about your background that led you to do this?

JM: There are a number of elements. For one, I have a history of addiction, and I really felt like apps were becoming very addictive and leading to loneliness, and I wanted to create something [different]. That whole idea of anti-digital addiction, of getting people to spend more time out in person than on their phones was really important.

And my personality of a hopeless romantic made me want to create something that would help people find love and find relationships and wasn’t just about hook-ups. Those were probably the most impactful things for me in terms of shaping the company.

CC: And is there anything you’d do differently? Anything you would change if you could go back in time?

JM: A million things and nothing. Of course I made 10,000 mistakes along the way, and, at the same time, those mistakes were learnings that allowed me to build the company that I have now.

And in fact that’s been my whole ethos as a founder and CEO: being thoughtful about continuous improvement — documenting what I feel like I know now, my principles for operating, and then constantly updating those as I learn more, and being transparent with the company about those learnings. That is essentially what allowed me to create an app that, frankly, wasn’t successful at all in the beginning but just to keep trying and trying again.

CC: Are there any specific mistakes that you’d be happy to share?

JM: Where do I begin? I’d say the biggest was losing sight of our customers and focusing on the competition. We just lost sight of what people came to us for. We were really dialled on engagement and retention and we were constantly looking at who was bigger than us and moving faster than us. And they’d release a feature, and we’d be like, “we have to have that feature”. 

So we did the reboot and we completely changed our orientation — we said we’re going to play a different game. With everything we build, does it deliver more great dates or not? That actually led to us building a very different product.

And when we launched the new product in 2016, there were a lot of mistakes. We thought that we should make everyone pay in order to use it — we thought that people would pay $9 a month if we just charged everyone. And it turned out that people are not willing to pay for dating, and all of the core product definitely had to be free.

I made mistakes by cluttering the app too much in the early days, and I learned how much design minimalism and thoughtfulness matter for creating a product that has longevity. That’s why our colours are black and white. We really tried to pare down and become minimalist in our approach.

CC: One of the things we’re seeing now is a lot of new players in the market, especially focused on different niches: Raya, Thursday, Feeld. What do you feel that you offer that’s different? And how do you try and stay relevant?

JM: It’s not new that there are new things popping up and going away. I’ve been doing this for 12-and-a-half years, and there are so many competitors that had a moment — Coffee Meets Bagel, say. I think the strategy, when you’re really niche, and you have a very specific identity that you’re trying to attract, is just a bit limiting. 

The better strategy is to have an app with more people on it that gives you the tools and preferences and filters that you need to find your person. Because if you’re facing apps for different identities — like enjoying a particular sport, say, or being a parent — which one do you join? Because you may have all those identities.  

CC: There’s definitely a feeling of burnout from using lots of different apps. Are you hoping to be, ideally, the only app people have, and then they delete it when they meet that person? 

The opportunity of AI is that it allows us to function more and more like a matchmaker that’s making really, really targeted introductions

JM: Yes. We’re the app that is very squarely trying to address burnout. And the key to that is, most simply, quality over quantity. People want to spend more time out on dates, and less time on the app.

When people feel burned out, it’s really one of two reasons. One is that you are overwhelmed: there’s so much activity, and so many people, and everyone starts to look the same, and conversations are dying.

At the other end of the spectrum, a lot of users get very, very little activity. They burn out because they’re trying to get that match, and they send a lot of likes, but then they’re not even getting enough [reciprocal] activity to go on one date. And so there’s a really big opportunity, and Hinge is already pretty good at that, relative to other apps. 

The opportunity of AI is that it allows us to function more and more like a matchmaker that’s making really, really targeted introductions, and helping people put their best foot forward. Then it can really become, over time, about more and more quality, and less and less quantity — less and less people you have to go through in order to find the right one. 

CC: I was going to ask about AI. What do you make of the new advances with generative AI and large language models? How are you looking at that at Hinge? 

JM: What matters is the deep learning aspect of the LLM, and, over time, continually dialling up the sophistication of our matching algorithms, so that it’s learning. Every single time someone goes on a great date, we can look at all the information that led up to it — everything from their profile information to their behaviour on the app, to how many matches they had at the time that they met. 

It is just a lot of information that we’re learning to use better and better, so that we can more predictably make matches that we think are going to be high fidelity, where people are going to like each other, and go on a date and have a good time. 

CC: And what’s the balance between image analysis and analysis of the text people put on their profiles? 

JM: We want as much signal as we can get, whether it’s images or the way people respond to prompts, or people’s behaviour in terms of liking versus passing on certain people. All of that is good information that can help us learn over time who’s going to lead to a good match, because we’re optimising. We ask people if it was a good date, so we actually have the information that we need to figure out: what are the components that lead to a good date? And how do we look for those, when we’re trying to match people up? 

We really wanted to make sure we had the product in a really good place in our core markets, in the US, UK, Canada, Australia, before we started to expand

We have the most detailed profiles, and we have the most selective and thoughtful [information on] liking and passing behaviour. And we have amazing coverage on that data. Being able to run the optimisation function on that, to figure out what leads to a great date and what doesn’t, [puts us in an] amazing place. Over the next few years, you’re going to see the app move through a giant evolution, I think as big as we saw with dating apps back in 2012, 2013, 2014.

CC: How do you get around bias there? Although naturally users will have particular preferences in the types of people that they’re choosing. 

JM: There are so many different layers to that question. Obviously it’s not just AI, there’s bias in human behaviour, period. That’s something we continually are conscious of, and were conscious of before AI.

We try to make sure that our algorithms, of course, are mitigating, or certainly not exacerbating, bias. But people also have preferences, so we have to pay attention to those preferences as they’re looking for their person. Our goal — and we do a lot of internal audits and analysis of this — is making sure that what we’re doing isn’t taking a majoritarian approach.

CC: What has been your strategy when you’re launching in non-English language countries? 

JM: We really wanted to make sure we had the product in a really good place in our core markets, in the US, UK, Canada, Australia, before we started to expand. Once we felt we had that, and had a product that was ready to go, we started in Europe. And we started with one country at a time.

So, we did Germany last year, and really focused on Germany and that region for the first three, six months — making sure that we got the translations right, and making sure that we were thinking about how we adapt Vitals and Dating Preferences, how we adapt Prompts. [“Vitals”, one of Hinge’s three categories of personal information, comprise details such as height and location; the other categories are “virtues”, which include job and education, and “vices”, such as smoking and drinking.]

Once we had our learnings there, we went to the Nordics, and then to France, and then started expanding, again, basically, one country at a time, with Spain and Italy, I think, being our most recent openings. And we’ve had good success in those markets — we’re in the top five in the countries where we have launched. 

In terms of growth in continental Europe, we’re expecting revenue, I think, to be about three times this year relative to last year, and downloads have increased about four times relative to last year.

CC: One concern that’s been raised about dating apps generally is how realistic some of the photos generated by AIs are now, and similarly with text, which means the number of bots on these platforms is going to increase. How are you approaching that concern? 

JM: We have a huge team on trust and safety, who are very much thinking about that. There’s a whole host of behaviours and components that bots have, beyond just photos and prompts — everything from IP address, phone number, carrier, device type, and all of that. So we have a lot of tools at our disposal to catch people who are bad actors and bots. But we’re also considering our capabilities to be able to identify AI-generated photos, and things like that. 

CC: Is it something that you’ve noticed a lot this year? 

JM: No, but we are keeping an eye on it. It’s something that we’re concerned about and monitoring for. 

CC: Another concern that some users have is Standouts, and the fact that you have to pay to access it. [Standouts is a Hinge feature that notifies users about likely matches who are also receiving a lot of attention.] Pay blocks can be intrusive for people who want the free experience, but obviously you’re a company and you have to make money. How do you strike a balance? 

JM: One of our key principles is that the free experience is sacred, and we want it to feel really good and functional for everyone. Anything that we can give to our users for free, we do.  

That said, there are opportunities for people to get a leg up, and Standouts is an example. I’m going to a wedding in two weeks of someone who sent a Rose [an ardent “like” icon] to a Standout, and now they’re getting married. So the tools are very effective for those who want to pay, but if they don’t, then the free experience is also really good. 

We view churn as good: people find their person and then it ends. That’s great news for us, because they’re going to tell their friends about it

CC: And how are you finding the take-up of paid services? 

JM: We’ve been pretty pleased with the launch of Hinge+ and Hinge X [Subscription services with greater functionality than the free account]. I still think there’s actually more opportunity for us, in terms of monetisation, when you look at our penetration rates, relative to other dating apps. But it’s about in line with where we expect. 

CC: Given your ethos that Hinge is meant to be deleted, and, more generally, the very high level of churn with dating apps, how are you attracting new users? And what trends are you seeing there? 

JM: The best way to attract new users is to be really good at what you do, so that people like it and tell their friends. That’s going to be cheaper than any marketing that you can do, which is why, for us, we view churn as good: people find their person and then it ends. That’s great news for us, because they’re going to tell their friends about it. And that’s been our growth strategy so far. 

Every time that we’ve seen an improvement in our ability to get users out on great dates, our growth tends to accelerate. So that is our primary method of growing, and the introduction of AI is going to open up a whole new realm of us being able to get people out on great dates, and reduce burnout. 

CC: Have you considered more explainability with the AI — why it is you’re giving certain matches or not? 

JM: Yes. That’s definitely something we’re looking into. From our experience, saying someone is “most compatible” is highly effective at getting people to go out on a date. And we don’t say anything other than that, so the ability for us to add more, to “prove” the algorithm, and to give people context, has a ton of potential for us to reduce burnout. It helps increase focus and really determine who people connect with. 

CC: So it could be something like, if they have a cat and you also have a cat, then that could explain it?

JM: Yes. This is a debate that we’re having internally in the team right now, because it’s interesting. So there are things that seem a bit superficial — you both have cats — but maybe it’s a good conversation starter. And probably it’s worth surfacing for that reason.

But sometimes in the world of AI, it’s not always fully explainable. We’ll have a really good read that you two should meet each other, but we don’t necessarily know why. So, it’s interesting to talk about the balance of just indicating that we think you are an especially good match, and trying to explain it — because it may not be as simple as you both have cats.

CC: Now that the pandemic has died down, what have you noticed in people’s behaviour? Has there been any change? 

Voice paired with photos can be really effective: it gives you a three-dimensional sense of what someone’s going to be like, yet it’s quite easy to record a voice prompt

JM: Not so much. With the onset of the pandemic, obviously, there were huge changes in behaviour, especially towards the beginning, and a pretty big increase in adoption of online dating, and of tools like voice and video. But since then, things have started to return to — well, I’ll call it the new norm. Those norms around using voice and video, and comfort around that, are not going away. But we’re not seeing any major shifts since then. 

CC: And do you feel like voice and video are more effective mediums for assessing somebody than, say, a still image? 

JM: Most definitely. You get a much better sense of someone, and we’ve found that voice is the most powerful. Even just a quick snippet of someone’s voice can be really, really powerful. People are still a little bit iffy about putting a video up on a dating app, it’s pretty vulnerable to do that. But voice paired with photos can be really effective: it’s a rich medium that gives you a three-dimensional sense of what someone’s going to be like, yet at the same time it’s quite easy to record a voice prompt. When people use Voice Notes [voice messages between matched users], the chance that it leads to a date is more than 40 per cent higher, which is an incredible difference.

CC: I’m sure you get asked this all the time, but what in particular works on a profile? What seems to do well, and what doesn’t? 

JM: Ultimately, there’s what does well in terms of what’s going to get you a lot of likes, and then there is what does well that is going to help you find your person. Those are different questions.

For what’s going to lead to a lot of likes, there’s some basic things that you can do — like your first shot should be a headshot, that has you smiling, and is in focus, and is not with other people, and things like that. 

In terms of what’s actually going to be effective: the more people are [prepared to make themselves] vulnerable, and to reveal more about themselves in their profile, especially things that are not like everyone else. . . that’s not about trying to get as many likes as possible, it’s about getting the right likes. It’s about attracting the person who’s going to like you, and actually turning away the people who aren’t. 

The more you’re willing to share on a profile that really gets to the heart of who you are and what you’re looking for, the more successful you’re going to be on Hinge. That’s the simple advice. 

CC: In terms of what the next iteration might be for dating apps, have you looked at different forms of hardware — things like headsets, for example? 

JM: We are much more focused on AI. With things like headsets and augmented reality, or whatever, I don’t immediately see how that’s going to be a game-changing benefit — not relative to what AI offers. AI is the next disruptive technology after mobile that’s really going to change the game. 

And dating is a bit of a laggard, I would say, when it comes to people using new technologies. Because people have to get to a certain level of adoption before you believe that you could find your person on a dating app. 

If you were limited to people who were using whatever brand headset you’re using right now, there would be a pretty limited dating pool. So until we’re in a world where 70 per cent or 80 per cent of people have one platform, or at least one interoperable platform, for a headset, it’s not the right platform for dating apps. 

CC: A question on dating trends: non-monogamy is becoming a big thing, and I know you can signal that on your Hinge profile as well. How are you incorporating it into your message of, you do want to get off the app at some point?

JM: This has been a big topic at Hinge, thinking about how we both navigate changing culture, and stay true to our value proposition of being designed so that you delete it after you find your person. 

Ultimately, more than it’s about finding your one person forever, the brand is about authenticity and vulnerability, and trying to find real connections. It’s not a game. It’s not a superficial or casual experience. It’s for depth. That is the real core of what we’re about. 

As long as you’re really looking for authentic, deep connections with other people, that can come in a variety of relationship forms, and we want to be very inclusive of all of those. Again, it comes back to the idea of making sure that people have the right preferences, tools and settings, so that they can zero in on their right person. 

The above transcript has been edited for brevity and clarity