The foursome thesis

Why golf is the largest network sport on earth without a network underneath it.

There are 108 million people who play golf with some regularity in the world today. There are roughly 50 million members of LinkedIn who would describe themselves as "frequent" users, and roughly 8 million people who pay for Strava's premium tier. By any reasonable definition of the term, golf is the largest active-participant network on earth that has, until now, had no network layer underneath it.

This is a fact. It is also strange. It deserves an explanation.

The conventional explanation is that golf is a niche, expensive, slow-growing sport whose participants are old, wealthy, and Republican, and whose habits are too entrenched to be reorganized by software. This explanation has the virtue of being partially true. It also has the defect of obscuring a more interesting fact: that golf, as a behavior, has more in common with social-network behaviors than almost any other physical activity humans do voluntarily. A round of golf is, in most cases, four people. A round of golf is, in most cases, four-to-five hours of conversation interrupted by brief moments of athletic activity. A round of golf is, structurally, a small dinner party held on a long walk.

If the same activity took place at a restaurant — four people, five hours, regular conversation, occasional small physical performances — it would have spawned several billion-dollar networks by now. OpenTable would not be enough. There would be Resy, Tock, an app for finding dinner companions, an app for booking the wine pairing, an app for the chef's notes, an app for splitting the check. There would be a 500-person engineering team in San Francisco working on the rich-text editor for the diner's reservation note to the server.

For golf, none of this exists. There is GolfNow, which sells tee times the way Hotels.com sells rooms — a marketplace for inventory, not a network for people. There is Whoop or Garmin for tracking the round, but the data is internal to the wearer. There is a small set of round-tracking apps like 18Birdies and TheGrint, none of which has reached the kind of scale a real network needs. There are, of course, the various private group texts among friends, organized through iMessage and WhatsApp, where the foursomes that actually happen are coordinated.

There is no network. The behavior is enormous and the network is zero.

Golf is the only billion-dollar sport without a network layer underneath it.

The structural reason

The reason for the absence is structural, not technological. Private clubs are deliberately closed systems. They were designed to be closed — by people who valued the closure, and who continue to value it. Every previous attempt to build a "network for golfers" has run aground on this fact within twelve months. The networks that have tried to include private-club members have been politely shown the door. The networks that have tried to route around private-club members have been built on the wrong side of the demand curve — they've optimized for golfers who play public courses occasionally, when the unmet demand is concentrated among golfers who play private courses often.

The unmet demand among private-club members is not for access. They have access — that is, by definition, what their membership is. The unmet demand is for other members of other private clubs to play with.

This is a subtle distinction. A member of Friar's Head, on Long Island, has more than enough access to Friar's Head. He has, on average, played the course at least sixty times in the last year. What he lacks is a structured way to find the member of Seminole — also a private club, with similar standards, similar wait-lists, similar membership ethics — who is in Long Island for a week in August and who would happily join him for a Tuesday-morning round, in exchange for a Tuesday morning at Seminole in February.

This kind of trade has always existed informally. It is, in fact, how a significant fraction of top-100 rounds actually happen. The trade is routed through introductions: a member knows a member who knows a member. It works imperfectly. It excludes everyone who is not adjacent to the right introduction chain. It depends on private group texts and on six-month wait times between the first call and the actual round. It is, in the language of network economics, a marketplace that already exists but operates at extreme inefficiency.

Inefficient marketplaces are the most productive places to build software companies. The interesting fact about the golf foursome marketplace is not that it doesn't exist — it does, and it has for fifty years. The interesting fact is that it operates at maybe 8% of its theoretical efficiency, because the matching mechanism is fifty members at fifty clubs holding fifty group-text threads in their phones.

The math

Consider a single top-100 private club with 600 members. Each member, on average, hosts roughly 12 guests per year — friends, business contacts, members of other clubs. That's 7,200 guest-rounds per club per year. Across the top 100 clubs in the United States, that's 720,000 guest-rounds annually.

Of those 720,000 rounds, roughly half are between members of two different private clubs. The other half are between a member and a friend or business contact who is not a private club member.

So: 360,000 inter-club private-member rounds per year, in the United States alone, routed almost entirely through informal text-message coordination. The latency between a member thinking "I should play with X" and the round actually happening is, on average, more than four months. The dropout rate — the percentage of intended rounds that never happen because the coordination collapsed — is, by our estimate from interviews with twenty-three founding members, north of 60%.

The latency is the disease. The dropout is the symptom.

A network that reduced the latency from four months to three weeks would roughly double the number of inter-club rounds per year. A network that reduced the dropout rate from 60% to 30% would roughly double again. The total addressable market, in pure round-count terms, is between 4 and 6 times the current actual.

This calculation does not include the second-order effects. It does not include the rounds that members would play with strangers from other clubs if a trusted matching mechanism existed — rounds that, right now, simply don't happen because there is no way for the matching to occur. It does not include the international layer — the member of San Francisco Golf Club who plays Loch Lomond twice in his life and would love to play it twenty times, but does not know a single member there. It does not include the guests — the friends and business contacts of members, who fill the other half of the 720,000 rounds and who are largely invisible to the current matching infrastructure.

The unmet demand is enormous. The supply exists. The matching layer does not.

Why this is finally tractable

Three things have changed in the last five years that make the matching layer build-able now in a way it was not in 2010 or 2015.

First, the membership at top-100 clubs has gotten younger and more digital. The median age of a new member at most top-100 American clubs is now in the late 30s, not the late 50s. These members are not waiting for the pro shop to call them back. They are accustomed to coordinating travel, dining, and social activity through apps.

Second, the post-pandemic explosion in golf participation has changed the math at the clubs themselves. Initiation fees have risen significantly. Wait-lists have lengthened. The cost of a private club membership has become high enough that members are increasingly motivated to extract maximum value from it — including by leveraging it for access to other clubs through reciprocity.

Third — and this is the structural change that matters most — a small number of private clubs have begun to recognize that their members benefit from a curated inter-club network, and that they themselves benefit from having a hand in shaping it rather than ceding the field entirely to informal channels. The first three GMs we worked with on the Tee Time concept were not, twelve months ago, willing to meet. Today, they are design partners. This change has happened in a window of a year and a half. It will not reverse.

The network is finally build-able because the buyers, the sellers, and the matching technology are aligned, for the first time, in the same eighteen-month window.

What we are building

Tee Time is the matching layer. We connect verified members of private clubs to each other, across clubs and across cities, in a way that respects the privacy of the underlying clubs and that produces real foursomes that real members actually play in. We layer on top a verified caddie network — supply-side infrastructure for the same audience — and a small set of partner-club relationships for the longer-term reciprocity layer.

What we explicitly do not do: we do not sell tee times. We do not take a cut from the clubs. We do not publicize private club rosters. We do not, ever, make private golf less private.

We launched two weeks ago, on July 8th. We are open in three cities. Founding members are joining at a rate that suggests we have substantially underestimated the unmet demand. The next eighteen months will tell us how badly.

Find your foursome. Anywhere in the world.

Garrett is the founder of Tee Time.

The Journal sends one post a week. Subscribe.