The Rabbit Hole, Part 5: Who the Next Worlds Get Built For

In 2017, women were 8 percent of Formula 1’s global audience. By 2025 they were 42 percent, which works out to roughly 350 million people, according to Formula 1’s own global fan survey run with Motorsport Network. The total fan base reached 827 million, up 12 percent in a single year.

No regulation change explains that. The cars did not get 34 percentage points more interesting to a group that had supposedly never cared. What happened is duller and much more expensive to have missed: the sport opened entrances into its world through channels it had spent decades treating as beside the point, and an audience it had never counted walked in fast enough to break every planning assumption in the paddock.

That is the whole finale in one number. The enthusiast economy did not discover a new customer. It discovered a filing error.

The series so far, and the one question left

Part 1 argued that nobody falls in love with the object, and named the five properties that make a system swallow people: depth you can keep descending, jargon that doubles as an identity test, lore that rewards the time you put in, legible progress you can point at, and contested taste that gives you something to argue about. Part 2 showed how brands learned to build those five properties on purpose and sell the world rather than the widget. Part 3 showed the internet pouring rocket fuel on the whole thing by collapsing the cost of finding your people and learning the vocabulary. Part 4 showed the payoff nobody budgeted for, an unpaid research and development layer made of hobbyists who solve problems the industry has not gotten around to yet.

Four parts on how the machine works. One question left, and it is the one with money on it.

Who has the industry been building these worlds for, and who did it forget to count?

The behavior was never male. The categories were.

Here is the claim I will spend the rest of this piece defending, stated flat.

Deep, systematic, jargon-heavy, progression-driven obsession has always been distributed across everyone. It is a property of how attention works, not a property of who is paying it. But the industry only ever filed that behavior under “hobby” when it showed up in a domain it already recognized as one. Everywhere else, the same five properties ran at enormous scale and got filed under vanity, frivolity, or nothing at all.

Call it misfiled depth. Same structure, same intensity, same documentation habits, sorted into the wrong drawer, and therefore never given a world to buy into.

“

The behavior was never male. The categories were. An industry that mistook a category for a personality spent thirty years under-pricing the largest opportunity in consumer business.

”

This is not an argument about who deserves what. It is an argument about a measurement failure with a price tag. If your model of an enthusiast comes preloaded with an assumed face, you will look for the behavior only where that face already is, find it there, and record the result as proof. Misfiled depth is what that self-confirming loop leaves lying on the floor.

Run the diagnostic, not the demographic

The clean way to test this is to stop arguing about people and start applying the test from Part 1. Take the five properties as a checklist and run them across domains the business press never called hobbies. If the properties are all present, the rabbit hole is present, whatever the category was labeled.

Now run the same test on the rest, quickly, because the pattern repeats and the repetition is the argument.

Fibre crafts and Ravelry. The knitting and crochet world built one of the most sophisticated hobbyist databases anyone has ever assembled, and built it themselves. Patterns, yarns, gauge, fibre content, needle sizes, project logs, modifications, substitutions, finished-object photography, all cross-referenced so you can start from a yarn you own and find every pattern that works with it. Commercial software teams with real funding routinely ship worse data models than that. It got called a craft site.

Fan fiction archives and their tagging systems. The archive world independently arrived at a solution to a hard information-retrieval problem: how do you let millions of works be findable by content, tone, structure, and specific reader preference, without a central authority defining the taxonomy in advance? The answer, roughly, was a controlled vocabulary layered over free tagging, curated by volunteers who reconcile synonyms and hierarchies so that a search actually returns what a reader wants. That is library science, performed at scale, unpaid, for a category that gets described as a pastime.

Competitive fandom analytics. Chart tracking, streaming-pattern analysis, release-window strategy, coordinated purchasing, sales forecasting, and post-mortems on why a campaign under-performed. Communities run this continuously, with spreadsheets, dashboards, and a shared statistical vocabulary. Any trading desk would recognize the behavior instantly.

Historical costuming. Primary-source research, pattern drafting from extant garments, period-correct construction techniques, fabric and fibre authenticity, documented reproduction with citations. It is applied material history with a sewing machine, and the documentation standards in that community are stricter than in plenty of professional fields.

Competitive baking and pastry. Hydration percentages, lamination ratios, fermentation schedules, crumb structure analysis, tempering curves, equipment obsession that runs from thermometers to specific stone compositions. Legible progress every single bake. Taste factions everywhere.

Look at that grid and the uncomfortable part arrives on its own. These are not lighter versions of the rabbit hole. The structure is identical, and in several cases the documentation is better than what you find in the domains that got the magazines, the trade shows, the sponsorship money, and the word “enthusiast” printed on the box.

They were not counted. That is the entire finding.

The money makes it impossible to keep misfiling

Arguments about categorization move slowly. Money moves fast, and the money has arrived.

Take those in order, because each one does separate work.

The Formula 1 shift is the proof of concept. Per the sport’s own survey with Motorsport Network, a segment went from 8 percent to 42 percent of a global audience in eight years. That is one of the fastest audience transformations modern sport has recorded. It did not require changing the product. It required building entrances, and then counting who came through them.

The spending numbers turn the audience into a market. Women are projected to control 75 percent of global discretionary spending by 2030, and already control roughly 80 percent of household spending. Inside sport specifically, 84 percent of women sports fans are key decision-makers in sports-related purchases. Read those together and the “hard to monetize” story that got told about these audiences for years looks less like an observation and more like a failure to try.

K-culture shows what a fully built world does to a P&L. Industry estimates put consumer spending attributable to TikTok’s contribution to K-culture at 10.75 trillion Korean won this year, about 7.8 billion US dollars, projected to rise 29 percent to 13.92 trillion won by 2030. In the same research, 55.7 percent of surveyed respondents said they spent money on products or experiences after consuming K-culture content. More than half. Any marketer who saw that conversion rate on a channel they controlled would rebuild the company around it.

BookTok shows it working on an old, supposedly flat category. BookTok-related book sales rose 20 percent in 2024, to roughly 60 million units in the United States alone. Publishing spent decades being told its problem was that people had stopped reading. The actual problem was that nobody had built a world around reading with the five properties switched on.

The blindness, priced

Now the sharpest piece of evidence, and the reason I keep saying this is a measurement failure rather than a moral one.

Fashion and beauty brands have largely excluded BookTok creators from their influencer programmes, despite those creators generating millions of organic views on character-inspired makeup and styling content. Sit with the shape of that. Creators were producing beauty and fashion content, at scale, unpaid, for an audience that buys beauty and fashion, and the beauty and fashion industry did not classify them as beauty and fashion creators. They were filed under books. The category label beat the observed behavior.

Meanwhile, in the other direction, ELEMIS became the official skincare partner of the Aston Martin Formula 1 team. A skincare brand buying its way into motorsport is the market correcting itself in public. Someone finally read the survey, noticed that 42 percent number, and did the obvious arithmetic that nobody had done for years.

“

This is a mis-pricing, not a moral failing. That is exactly why it will correct quickly and completely. Mis-pricings get arbitraged by whoever notices first, and the noticing has started.

”

I want to be precise about the mechanism, because it matters for what happens next. Nobody sat in a meeting and decided to ignore a 350 million person audience. What happened is that categories carry assumed customers, assumed customers shape which behavior gets measured, and unmeasured behavior does not appear in the plan. The loop is self-confirming and it runs quietly for decades. Then one number gets published, and the whole thing unwinds in about two budget cycles.

What changes next (these are predictions, not reports)

The following are my bets on the next five years. Label them as such. Each one follows from the mis-pricing above rather than from any survey.

Prediction one: worlds get built for audiences that were previously treated as incidental. Every one of the world-building moves from Part 2, the lore, the tiers, the collectibles, the documentary series, the ladder of belonging, gets rebuilt and aimed at people who were previously served with a marketing campaign instead of a world. The first movers in each category will look brilliant. They are mostly just early to a number that was already published.

Prediction two: the working definition of “depth” broadens. The industry’s inherited picture of an enthusiast is solitary and technical: one person, one workbench, one specification sheet. That is one shape of depth, not the shape. Fandoms already run coordination, analytics, archival, and quality control at a scale most companies cannot match, and they do it collectively. Expect collective depth to be recognized as depth, and expect products designed for it, which means built for groups by default rather than for individuals who happen to share.

Prediction three: community-built infrastructure gets treated as infrastructure. Community databases, tagging systems, wikis, archives, and authentication registries are load-bearing for entire markets, and they are usually maintained by volunteers on donated hosting. Expect serious money to start arriving in that layer, through acquisition, funding, and partnership. Expect at least one ugly public fight when a brand tries to buy a community’s archive and the community declines to sell.

Prediction four: the resale, authentication, and provenance layer keeps compounding. The sneaker and handbag markets already proved that authentication is a business, that provenance carries a premium, and that condition grading creates a whole second economy on top of the first. That pattern travels to every category where an object carries status and history. Expect it to spread anywhere the five properties are running, because provenance is just lore with a serial number.

Prediction five: late arrivals will have to ask permission. This is the one most brands will handle badly. When you arrive late to a world the community already built without you, you are not entering an empty market. You are a guest in something with its own history, standards, and memory. Communities can smell a brand that showed up because a slide deck told it to, and the entry price is genuine participation over time. Money alone will not buy the door.

The rabbit hole itself is about to change shape

One last turn, because the series should not end on a market opportunity.

Part 3 was about the internet collapsing the cost of entry. Finding your people used to take years and geography. It became instant. Learning the vocabulary used to take a mentor or a decade of magazines. It became a search box.

AI does that again, harder. An expert tutor on tap collapses the learning curve to almost nothing. The vocabulary, the mental model, the whole “why does everyone in this hobby care about that one variable” question that used to gate newcomers for a year, all of it becomes available in an afternoon, at any hour, without having to survive a forum’s mood. I think this is genuinely democratizing. The five properties made rabbit holes seductive, and the same properties made them hostile to beginners, because jargon is an entry test and lore is an accumulated debt that newcomers cannot pay. Anything that lets more people through that door builds more worlds, not fewer.

The honest risk sits right next to the upside. If the learning curve can be collapsed, so can the appearance of depth. Synthetic expertise is now cheap to manufacture. A brand can generate lore, produce a heritage, populate a forum, simulate a founding myth, and ship a world that has all five properties on the surface and no history underneath it.

So here is the bet I will close the series on. As synthetic worlds get cheap and abundant, the scarce asset becomes the authentic community with real history. Not the biggest audience. Not the slickest world. The one with receipts: arguments it survived, standards it enforced on itself, mistakes it corrected in public, and members who were there. That is the component AI cannot generate, which makes it the one that appreciates while everything around it inflates.

Which brings the whole thing back to where Part 1 started, and to you.

Whatever you are building, you are not choosing whether to build a world. Your customers will build one around your product with or without you, and Part 4 showed they will do your research and development for free while they are at it. The only real questions are how deliberately you build it, and who you assumed would show up. Go check that second assumption. Find the domains where all five properties are already running and nobody has bothered to file them correctly. That gap is not a gap in the market. It is a gap in who somebody remembered to count.