A fumigation experiment in the Florida Keys explains more about marketplace dynamics than most business books.
In the summer of 1966, a Harvard graduate student named Daniel Simberloff pulled on a pair of waders, loaded a fumigation rig onto a small boat, and motored out to six tiny mangrove islands scattered across the Florida Keys. Each island was barely larger than a living room — eleven to eighteen meters across — and each was home to between twenty and fifty species of arthropods: spiders, beetles, ants, moths, mites. Simberloff tented each island in plastic sheeting and pumped in methyl bromide until every arthropod was dead. Then he and his advisor, the legendary ecologist Edward O. Wilson, sat back and watched what happened when an empty island had to fill up again (Simberloff & Wilson, Ecology, 1969).
What happened next became one of the most cited experiments in the history of ecology. It also happens to be the best mental model available for understanding why most marketplaces die — and why the ones that survive look nothing like what their founders originally planned.
The experiment Simberloff and Wilson ran was designed to test a theory Wilson had published the year before with the mathematician Robert MacArthur. Their 1967 book, The Theory of Island Biogeography, proposed something elegant and counterintuitive: the number of species on any island isn’t random. It’s the result of a dynamic equilibrium between two rates — the rate at which new species arrive (immigration) and the rate at which existing populations wink out (extinction). The formula they proposed, S = cAz, relates species count (S) to island area (A), with a scaling exponent z that typically falls between 0.20 and 0.35 for oceanic islands (MacArthur & Wilson, 1967).
Every empty island faces the same problem: the first colonizers arrive to a hostile world. There are no mutualistic relationships, no soil enriched by prior inhabitants, no existing food webs to slot into. Most early arrivals die. The ones that survive do so because enough of them landed at roughly the same time to form a minimally viable population — a cluster large enough to reproduce, find food, and weather random catastrophe.
Anyone who has tried to launch a two-sided marketplace will recognize this instantly.
Andrew Chen, general partner at a16z, named it the cold start problem: a marketplace has zero value until both sides show up in sufficient numbers (Chen, The Cold Start Problem, 2021). Buyers won’t browse a platform with no sellers. Sellers won’t list on a platform with no buyers. The first users arrive to a barren landscape — no reviews, no trust signals, no evidence that anything works. Most of them leave.
The specific numbers are striking. Airbnb’s internal data showed that a city needed roughly 300 listings, with at least 100 that had been reviewed, before booking growth became self-sustaining. Reaching that threshold took two to three years per major market (Lenny Rachitsky, ex-Airbnb, Lenny’s Newsletter, 2022). Uber needed approximately 30 drivers per market with wait times under 15 minutes to match taxi reliability. DoorDash bootstrapped with 70 restaurant partnerships in the Bay Area before expanding, eventually reaching 590,000 partnerships (DemandSage, 2026).
Compare those timelines to Simberloff’s mangrove islands. Five of six fumigated islands returned to approximately their pre-defaunation species counts within 250 days. The process was the same: a trickle of arrivals, most of which failed, until a critical mass established itself and the system stabilized. The difference was in what stabilized. The refaunated islands held roughly the same number of species as before — but they were different species. The equilibrium was about rates, not identity (Simberloff & Wilson, Ecology, 1970).
Marketplaces that rebuild after disruption show the same pattern. Airbnb post-COVID returned to similar transaction volumes, but the mix of hosts and guests had shifted fundamentally — more rural listings, more domestic travel, different buyer demographics. The system found equilibrium at a similar scale, but with different inhabitants.
MacArthur and Wilson’s model identified two master variables: size and isolation. Larger islands intercept more colonizers and support bigger populations that resist stochastic extinction. More isolated islands receive fewer immigrants, dragging down the equilibrium species count.
The marketplace parallels map cleanly. A larger addressable market supports more sellers — the target-area effect. Higher transaction friction — payment complexity, regulatory hurdles, trust deficits — reduces participation, exactly as oceanic distance reduces immigration to remote islands.
But here’s where it gets interesting. Wilson and MacArthur found that the z-value — the scaling exponent that governs how steeply species counts rise with island area — is higher on true oceanic islands (0.20–0.35) than on mainland habitat islands (0.15–0.25). The species-area effect is stronger where barriers are greater.
Marketplace economics shows a similar pattern. Platforms operating in high-friction environments — healthcare, financial services, government procurement — often exhibit steeper scaling curves once they crack the trust problem. The barrier itself creates lock-in. A hospital procurement marketplace that navigates HIPAA compliance, credentialing verification, and insurance billing creates switching costs that a general marketplace never achieves. The friction that nearly killed the platform at launch becomes its moat at scale.
This brings us to the most counterintuitive finding in island biogeography, and the one most useful for marketplace builders.
Isolation doesn’t just impoverish islands. It also creates entirely new forms of life.
Darwin’s finches — approximately 15 species, all descended from a single colonizing ancestor that reached the Galápagos — are the textbook example. But the Hawaiian honeycreepers are more dramatic: roughly 50 species radiated from a single founding population, diversifying into forms that eat seeds, nectar, insects, and snails, with beak shapes ranging from parrot-like crushers to long curved probes (Frontiers in Ecology and Evolution, 2021). This explosion happened precisely because of isolation. The colonizers arrived on islands with “few competitors and predators and many available resources” — empty niches that could never have been occupied on the competitive mainland.
The pattern holds quantitatively. Percentage endemism — the proportion of species found nowhere else — is positively correlated with isolation. The more remote the island, the more unique species it produces, because immigration of mainland generalists dilutes endemism in well-connected islands (Chen, Ecology, 2009).
Vertical marketplaces are marketplace founders doing adaptive radiation.
Faire (wholesale), Veeva (pharma), Procore (construction) — these platforms thrive precisely because horizontal giants like Amazon can’t serve their specialized needs. The “isolation” of a narrow vertical creates space for endemic features: compliance workflows built into every transaction, industry-specific matching algorithms, domain expertise embedded in the platform’s architecture. A general marketplace won’t build HIPAA-compliant messaging. A construction marketplace won’t bother with retail checkout optimization. Each vertical develops unique capabilities that have no analogue on the generalist mainland.
And just as biological endemism increases with isolation, marketplace “endemism” — the proportion of features, workflows, and integrations unique to that platform — increases with vertical specialization. The more focused the niche, the more irreplaceable the platform becomes.
Not every island is equally vulnerable. In 1977, James Brown and Astrid Kodric-Brown identified what they called the rescue effect: islands closer to the mainland experience lower extinction rates because ongoing immigration reinforces declining populations before they die out (Brown & Kodric-Brown, Ecology, 1977). It’s not charity — it’s mathematics. When the immigration rate exceeds the extinction rate for any given population, that population persists even if it couldn’t sustain itself in isolation.
For marketplaces, the rescue effect is external demand. Platforms with strong inbound traffic — from SEO, content marketing, API integrations, or partnerships — experience lower seller churn because new buyers constantly reinforce underperforming listings. The classic example is Airbnb’s early Craigslist integration: by cross-posting listings to an already-liquid platform, Airbnb created a literal rescue effect, driving immigration from the mainland (Craigslist’s massive user base) to its own sparsely populated island.
The implication is practical: marketplace founders who treat external traffic as a “nice to have” are ignoring the mechanism that separates islands that thrive from islands that empty out. The rescue effect isn’t a growth hack. It’s the mathematical difference between crossing critical mass and dying in the cold start valley.
The darkest concept in island biogeography is extinction debt. When an island loses habitat — through deforestation, sea-level rise, or human development — species don’t vanish immediately. Populations shrink below viable thresholds but persist for years or decades as “living dead,” organisms that appear healthy but belong to populations too small to sustain themselves. In the Macaronesian archipelagos (the Canaries, Azores, Madeira), researchers calculated that island ecosystems carry enormous extinction debts — their current species counts are artificially inflated and will decline over coming decades even without further habitat destruction (Triantis et al., Ecography, 2010).
The species are still there. The census looks fine. But the extinction is already locked in.
Platform leaders should find this terrifying.
Marketplaces that stop investing in supply quality, trust infrastructure, or demand generation accumulate platform debt — the commercial equivalent of extinction debt. Gross merchandise value holds steady. Transaction counts look healthy. But the ecosystem is hollowing out: top sellers are quietly multi-homing to competitors, buyer satisfaction is trending downward, and the trust mechanisms that once differentiated the platform are falling behind industry standards.
A Harvard Business School study of more than 250 platforms identified four primary failure modes: mispricing on one side, failure to build trust, prematurely dismissing competition, and entering too late (Zhu & Iansiti, Harvard Business Review, 2019). What’s notable is that three of the four are slow-moving, invisible failures — the kind that accumulate as extinction debt long before they show up in quarterly metrics. MySpace, Groupon, and the early daily-deal clones all exhibited this pattern: the numbers looked stable until a cascade made them suddenly, irreversibly not.
There’s one more finding from island biogeography that marketplace builders should know about, and it comes from the edges of the theory.
Among very small islands, the species-area relationship breaks down. Below a threshold size, species richness varies independently of area — a phenomenon called the small-island effect. On these tiny islands, niche availability governs diversity rather than colonization-extinction dynamics. The power law that governs larger islands simply stops applying (Lomolino & Weiser, Journal of Biogeography, 2001).
Different rules for different scales.
Micro-marketplaces — platforms for rare collectibles, hyperlocal services, professional communities — operate below the threshold where traditional platform economics apply. Network effects, liquidity metrics, and take-rate optimization matter less than niche depth, community trust, and curation quality. The hyperlocal services market, valued at $1.4 billion in 2022 and projected to reach $4.6 billion by 2028, competes through local knowledge and speed rather than scale (Aalpha, 2026). These platforms don’t need to become islands the size of Borneo. They need to become islands where every niche is occupied by exactly the right species.
After two years of censusing his fumigated mangrove islands every eighteen days, Simberloff reported a finding that didn’t make the textbook summaries but matters more than the ones that did. The most distant island — the one farthest from the mainland, the one that received the fewest colonizers, the one that recovered slowest — eventually reached equilibrium too. It just took longer, and when it did, its species composition was the most distinctive of all. The island that looked like it was failing the longest ended up being the most original.
Every marketplace founder staring at an empty platform and a burn rate should take a moment to consider the mangrove islands of the Florida Keys. The dynamics governing their success are not metaphorical. They are structural. Colonization is a patience game governed by arrival rate and survival rate. Friction is both barrier and moat. Isolation creates what competition never could. And the metrics that tell you everything is fine might be carrying a debt you haven’t noticed yet.
The islands are still out there, filling up according to rules discovered sixty years ago. The only question is whether you’re building on one that’s large enough, connected enough, and patient enough to reach equilibrium — or whether you’re pumping resources into an island that’s already carrying its extinction debt.
Trust infrastructure is what separates marketplaces that reach equilibrium from ones carrying extinction debt
The Agent Rating Protocol provides portable, verifiable trust scores for agent marketplaces — the kind of trust signals this essay argues every platform needs to invest in before the metrics start lying to you.