Manifold Love - Retrospective
What we learnt from creating a wild dating app
Written by @goblinodds
Manifold.Love is going into maintenance mode while we focus on our core product. We hope to return with improvements once we have more bandwidth; we’re still stoked on the idea of a prediction market-based dating app! In the meantime, some reflections on how things went:
The biggest difficulty was always going to be getting a critical mass of users. A dating app is only as good as the dating pool it provides. We had the advantage of a built-in community of prediction market enthusiasts, rationalists, Astral Codex Ten readers, Effective Altruists, etc. to get us started, but the network effect didn’t take off quite as hard as it needed to. We did some advertising in the hopes of broadening that pool but found that this tended to attract people who were less likely to be a good culture fit given our user base. The fact that so many users knew each other already was, I think, a big part of the hype—it’s fun to open up a dating app and see a bunch of your friends!
But the fact that a lot of people already knew each other was a double-edged sword: most people aren’t on dating apps looking for people they’re already acquainted with! I think we could turn this to our advantage given more time/funding, though.
To my mind, Manifold.Love’s value proposition is threefold:
It’s a niche dating app, in the same way that Jdate is a niche dating app but moreso; any two people on the platform are unusually likely to get along.
A big issue with existing dating apps is the fact that you’re dealing with complete strangers. You have to do all the vetting yourself and there aren’t any real incentives for people to treat each other well. It’s a weirdly adversarial environment given the degree of vulnerability people need to cultivate if they want to date successfully!
But Manifold.Love’s social graph is tightly knit. We took advantage of this by including endorsements:
People. Hate. Swiping. Or scrolling through page after page of eligible bachelor/ettes they don’t want to date. In theory, if you’re looking for a partner you can create a Manifold.Love profile, answer a bunch of compatibility questions, and then walk away and let someone else deal with the problem of finding someone for you.
I think there are a few ways we could take better advantage of these strengths in future:
We could lean harder into the matchmaking angle. A lot of the work we did was geared towards making the platform more usable for people looking to date; “likes” and this sort of thing, but it’s possible our focus should have been on improving matchmaking tools. Part of the difficulty here is that the Venn diagram of “people who are excited to bet on markets” and “people who are excited to matchmake”… doesn’t overlap as much as we’d want. This was part of why we shifted from markets for all users to markets as a paid feature plus “shipping” as a simplified way to matchmake. Market enthusiasts could focus their energies on a smaller number of users and matchmakers weren’t obligated to understand how mana works. (Betting with mana is dead simple! But it’s difficult to communicate that simplicity; this is something we need to work on for our main product, Manifold.Markets, as well.)
The other thing we could improve, with respect to matchmaking, is the ability to browse for another user. Initially, we restricted the geographic range in which you could browse matches, but our user base is small and (if you aren’t in San Francisco or NYC) scattered to the four winds. So we expanded it! But that forced matchmakers to browse sometimes hundreds of matches.
If/when we reboot, we’ll need to include the ability to filter browsed matches, so people looking to matchmake aren’t stuck relying on users having already liked each other.
Speaking of expanding geographic range: most people aren’t willing to date outside of easy travel distance, which is a big issue when you have a small, distributed user base. But a surprising number of Manifold.Love users are willing to date long-distance, or live pretty nomadic lifestyles that make it worthwhile to meet far-away matches. It would be great to include a way to flag willingness to travel (or maximum acceptable radius) on user’s profiles.
I also think we could take more advantage of our tight social graph by implementing “friending.” Being able to see a user’s friends would be great for people seeking love, for the reputational advantages mentioned above, and for people looking to match others, so they don’t waste time re-introducing people to their friends. (And maybe there’s value in making these reintroductions! But it wouldn’t hurt to know that it is, in fact, a re-introduction.) We could also try building more viral features for inviting friends, like the ability to make a stub profile on a friend’s behalf, with the aim of achieving sustainable organic growth.
The markets themselves also proved to be a bit of a challenge; for instance, there’s not much incentive to bet “NO” on the probability of something as unlikely as “going on three dates” with the same person. We’ve got some ideas for how to improve this in future:
building in a mechanic for leverage (e.g. automatic loans on very likely bets)
incorporating shorter term markets, like “Will x and y like each other,” or “Who will x like? [independent multimarket]”
One last issue: Manifold.Love profiles are publicly accessible, but many of our potential users are anons or have privacy concerns. In future, we’d like to give users the option to be private by default and revealed only to sufficiently compatible matches.
Now for what went well:
Bet on Love was spectacular, a truly absurd experiment and a fantastic time. You might be tired of hearing about it but we’re not tired of talking about it.
We expected our wider community’s gender ratio to be a big issue, but as of the time of writing, our gender breakdown wasn’t too different from Tinder’s: as of 2021, Tinder was 75.8% male and 24.2% female, while as of the time of writing this post, Manifold.Love is 73% male (including 0.4% transmasc), 21.9% female (including 3.2% transfemme), and 5.1% nonbinary. (Side note, probably this is value-neutral but I personally think it’s kind of magic that our users represent such a broad palette of sexualities, genders, relationship styles, and political/philosophical views.)
The Manifold.Love twitter account went unexpectedly well; it took us something like a month to hit 1k followers, thanks in large part to this thread blowing up.
Lastly, people have been clamoring for a return to the golden era of OKCupid for years, now, and we actually did it, in no small part thanks to Eli Tyre, who submitted the vast majority of our roughly 270 compatibility questions.
I think we could have taken better advantage of this by following up with our earliest users, many of whom didn’t revisit their profiles to answer any compatibility questions after we added the feature.
We still have a lot of faith in this particular application of prediction markets. Incentives rarely align when it comes to dating apps, and I think we made some headway on cracking this. It’s tricky to get an accurate sense of how many dates have come out of Manifold.Love so far, but there’ve been at least 141 dates that we know of based on market resolutions (some people were too shy to resolve their markets “YES,” and not every date had a corresponding market). Additionally, 31% or so of users exchanged mutual messages on the platform!
The site’s not going anywhere; we’re just not developing new features. Whether or not we start developing the platform again depends on how the rest of this year goes. Thanks for coming along on this ride with us!
To be continued…?
Something you didn't mention but that I think could be another benefit of friending. If I (A) see someone's profile who I don't know (B), look through their friend list and see that we have a mutual friend (C), I can now message C, who can tell me if they think it would be a good fit and then play matchmaker for A and B. With an interconnected user base, there are probably lots of people who are two degrees of separation away, but you'd never think to reach out to C because you don't know that B and C know each other.
another user and I are creating AI emulations of ourselves and putting them together to chat - if each user had a sufficiently advanced AI emulation, the manifold.love experience will be indistinguishable from magic - reply if you're interested in joining!