Marriage Pact:
the algorithm that gives you exactly one match.

Here is a fact that should not be true, and yet is: one of the more durable ideas in American college dating started as a problem set. In the fall of 2017, two Stanford undergraduates, Liam McGregor and Sophia Sterling-Angus, needed a final project for Economics 136, a course on market design taught by Paul Milgrom - who would go on to win the 2020 Nobel Prize in economics for work on auctions. Their project was to test whether you could use matching theory, the same math that pairs medical residents to hospitals, to pair people to people.

The mechanism they built is elegant in the way that good market design is elegant, which is to say it removes choices rather than adding them. You do not browse. You do not swipe. You fill out a long questionnaire about your values - roughly 70 statements you rank on a 1-to-7 scale - and then an algorithm "interviews" every other participant on your campus on your behalf and hands you a single name. One person. Your most compatible option, framed cheerfully as a "backup plan" you agree to marry if you are both still single years from now.

The framing is the trick. Calling it a backup plan lowers the stakes to almost nothing, which is precisely why so many people say yes. Within four days of the 2017 launch, 3,400 Stanford students had signed up - about 58% of the undergraduate body. Flyers went up. Screenshots went around. The thing that spreads a product across a campus is not a marketing budget; it is a match result that a person wants to show their roommate. Marriage Pact had that from day one.

What is genuinely unusual, from a business standpoint, is what the company optimizes for. Dating apps make money when you stay. Their incentive is your continued loneliness, dressed up as engagement. Marriage Pact runs once a year and its unofficial line is closer to "we hope you don't need us." A product whose success metric is users leaving satisfied is a strange thing to build a company around. It is also, arguably, the only honest way to build one in this category.

Tinder gave you infinite faces. Hinge gave you prompts and a like-limit. Bumble gave you a clock. Marriage Pact looked at that arms race and walked the other direction. There is no infinite scroll to design, because there is no feed. There is no ranking by attractiveness, because there are no photos. There is no daily habit to engineer, because the whole thing happens once a year.

That restraint is not minimalism for its own sake. It is a bet about what actually predicts a relationship. The questionnaire asks about the unglamorous stuff - how you handle conflict, whether "I love you" is a promise, how much you weight ambition or spontaneity or politically incorrect humor. These are the variables that tend to matter at year three, and almost none of them are visible in a photo.