PastBehavior Research
What history says about when an earner starts behaving like a business
Companies have spent decades deciding when a casual seller, freelancer, creator, or customer becomes something commercially different. We looked at the historical record to see which signals companies used, what happened when they acted on them, and where the evidence still stops.
The economy creates people who sit between consumer and business long before they formally choose one category. The interesting historical question is how companies recognized that transition, and what happened when they acted on it.
10 min read · 8 company cases · 5 academic studies · 1 policy natural experiment
Evidence labels: EVIDENCED (documented in filings, policy records, or peer-reviewed research), DIRECTIONAL (a recurring pattern without causal isolation), UNRESOLVED (the record cannot answer it).
Executive verdict
The transition shows up in behavior before identity. The evidence behind any single signal is narrower.
Being raised or working around entrepreneurship raises the probability of later entrepreneurship. Swedish adoption data shows the transmission is largely learned rather than inherited: post-birth environment accounts for roughly twice as much as pre-birth factors, with inheritance of the business and family capital largely ruled out as the channel.
Evidenced
The best-conditioned contrast available compares owners who worked in a family business with owners who merely had a self-employed relative. The first group closes less, earns more, and sells roughly 40 percent more. The second group shows almost nothing. But that sample is conditioned on already owning a business, and no study isolates which responsibility carries the effect.
Directional
eBay tiered sellers on trailing sales volume and transaction count. Amazon's Professional plan triggers at 40 items a month. Square, Shopify, and PayPal underwrote lending on processing history, not incorporation documents. The label followed the behavior in nearly every documented case.
Evidenced
Transaction frequency, sustained volume over a trailing window, multiple unrelated payers, repeat customers, and operating complexity recur across the case set. They were observable in real time, and they are what companies actually built triggers on. Whether each is independently predictive is not established.
Evidenced as used · Directional as predictive
The 1099-K threshold cut to $600 gross with no transaction minimum swept in casual sellers disposing of personal items at a loss. The IRS estimated 44 million forms, delayed the rule three times, and Congress reversed it in July 2025.
Evidenced
Sellers who bring their own customers look more independent. They were also the strongest sellers before anyone measured them. Reverse causality runs through every available comparison, and no dated policy shock breaks it.
Unresolved
01 / The historical problem
A person can receive money outside payroll while relying almost completely on someone else for demand, pricing, and infrastructure. Another person can look identical in the data and already behave like an operator: finding customers, setting prices, absorbing the loss when something goes wrong.
Institutions have had to decide where that boundary sits for a century. Telephone carriers classified lines as residential or business and priced them differently; in October 1995 a single business line averaged $41.77 a month, and the marker was use pattern and line count, not a filing. Merchant acquirers policed commercial use on consumer accounts by volume, chargeback exposure, and transaction type. The boundary was behavioral long before it was digital.
A prior question sits underneath the commercial one. Does early exposure to market activity change later economic behavior at all? The strongest research designs say environment transmits entrepreneurship, and that the transmission is largely learned.
Nearly 4,000 Swedish adoptees. Having an entrepreneurial parent raises the probability of the child's own entrepreneurship by about 60 percent. Post-birth environment accounts for roughly twice as much as pre-birth factors, and the design largely rules out inheriting the business or accessing family capital as the channel. Role modeling is the favored reading.
What it cannot say: Which experiences inside the environment matter. Nothing is measured at the level of market functions.
Evidenced
Roughly 270,000 Danish workers hired in 1990 and tracked to 1997. Workers with formerly entrepreneurial coworkers were more likely to become entrepreneurs, with the strongest effects for people with little other exposure.
What it cannot say: Passive proximity already predicts entry. The design cannot say whether active responsibility would predict it more.
Evidenced
Among US business owners, having worked in a self-employed relative's business predicts lower closure, higher profits, higher employer status, and roughly 40 percent higher sales. Merely having the self-employed relative predicts almost none of these.
What it cannot say: The sample contains only existing owners, so it cannot say active exposure causes entry. Selection into working in the family business is severe.
Directional
Gig workers are more likely to become entrepreneurs, and their firms start larger: about 23 percent higher revenue and 39 percent higher employment at inception. They are also about 3 percentage points less likely to survive one to three years.
What it cannot say: Which function the gig role taught. The mechanism reads as on-the-job learning, not proven market-making skill.
Directional
~60%
increase in the probability of later entrepreneurship when a parent is an entrepreneur (Lindquist, Sol and Van Praag 2015)
2×
post-birth environment accounts for twice as much as pre-birth factors in the adoption decomposition
~40%
higher sales among owners who worked in a family business; merely having the relative showed almost no effect (Fairlie and Robb 2007)
Read together: exposure matters, and active responsibility is the better bet over passive proximity. But no design isolates customer acquisition, pricing authority, or risk-bearing as the causal ingredient. The question of which market function matters sits exactly where the identification is weakest.
One asymmetry recurs: exposure raises entry more reliably than it raises survival.
02 / The signals
Companies could not wait for self-declared identity. They built triggers on what they could observe. Six signals recur across the historical record.
Repeated activity rather than a one-off receipt.
Amazon draws its plan line at 40 items a month. eBay's PowerSeller tiers required at least 100 transactions.
Evidenced as used
Activity that persists over a trailing window rather than spiking once.
eBay keyed tiers to trailing-12-month volume. Square underwrites on processing history and account tenure.
Evidenced as used
More than one unrelated payer or customer.
Intuit built QuickBooks Self-Employed around comingled expenses and multiple payers, not an income line.
Directional
Expenses, invoicing, mileage, and bookkeeping outgrowing a consumer product.
Intuit's signal was Mint users forcing business expenses into a personal finance tool.
Directional
Buyers who come back.
eBay's 2008 changes explicitly rewarded repeat transactions and buyer satisfaction.
Directional
The user starts walling off business flows from personal ones.
Mint users categorized business expenses as “other” to keep them apart. Venmo let a personal login spin up a business profile.
Directional
These six recur as the signals companies built on. The research does not establish that each is independently predictive, and no case isolates one signal from the others. Income, the most obvious candidate, is the weakest of all. Section 05 covers why.
03 / Historical cases
Eight cases, chosen because each shows a different way of drawing the boundary. The format is fixed: what changed, what the company could see, what it did, what happened, and what the evidence actually supports.
622,000
customers upgraded from personal to business accounts in the nine months ended September 30, 2001
What we can actually conclude
Behavior forced the category. The trigger was the act of accepting commercial payment, not the user's declared identity. Descriptive, but one of the few hard before-and-after counts in the record.
Descriptive
What we can actually conclude
The graduation signal was operating complexity, not revenue or legal structure. The behavior preceded any formal business identity. Directional.
Directional
$5.7B
originated in 2024, the largest online business lending total deBanked tracks
What we can actually conclude
Processing behavior alone underwrote credit, well below any incorporation event. Square's claim that borrowers “typically experience growth” is selection, not a proven treatment effect. Directional.
Directional
$4.2B
cumulative funding in 2025, up from roughly $3 billion in 2024 (Funder Intel)
What we can actually conclude
The same pattern as Square. Sales history is the underwriting substrate, not incorporation. Descriptive.
Descriptive
40
items per month marks the explicit graduation line between selling plans
What we can actually conclude
The graduation line is transaction count, not revenue. Amazon prices the boundary and lets sellers cross it themselves. Descriptive.
Descriptive
−5%
decline in Marketplaces net transaction revenue in 2009 versus 2008 (FY2009 10-K)
What we can actually conclude
A dated imposition followed by weak metrics is a sequence, not an effect. The 2008 to 2009 financial crisis is a large confound, and eBay survived to renarrate the episode as a quality push. The caution stands with that caveat: repricing a marketplace to graduate its base toward professional economics can suppress the casual supply that fed it.
Descriptive to directional
12%
mandatory ad fee imposed on sellers at $10,000 or more in trailing-365-day revenue
What we can actually conclude
The threshold selects exactly the sellers most able to leave. This is the value-versus-leavability tension made into policy. Directional, and confounded.
Directional
34
developers out of roughly 136,000 who applied for Apple's link-out entitlement
What we can actually conclude
A narrow claim, used narrowly: when a platform controls customer access, the friction of the escape hatch governs whether anyone uses it, not only the fee. Quasi-causal for that claim. This page does not generalize it into an App Store case study.
Quasi-causal (friction claim only)
04 / What worked
Four patterns recur across the cases that ended well, or at least without visible damage. All are directional: the record shows responses and sequences, not controlled experiments.
Business-like behavior was visible months or years before incorporation or a tax entity would have revealed it. PayPal, Intuit, Square, and Shopify all acted on signals visible before any registration.
Directional
Processing volume, payment frequency, and account tenure are observable in real time and refresh with every transaction. Square, Shopify, PayPal, and Amazon's item-count trigger all ran on data the company already had.
Evidenced as used
PayPal's upgrade-any-time and Venmo's business profile spun from a personal login let users cross the boundary themselves. The imposed versions, eBay's repricing and Etsy's mandatory ads, produced the resistance.
Directional
Comingled expenses, multiple payers, and consumer tools stretched past their design surfaced product need before any business registration. Intuit is the clearest case.
Directional
05 / What did not work
The failures are more instructive than the successes, and better documented. Three ways of drawing the line produced visible damage.
The American Rescue Plan Act of 2021 replaced a reporting threshold of more than $20,000 and more than 200 transactions with a single $600 gross threshold and no transaction minimum.
Gross receipts, not profit or business activity, drove the form. The IRS's own worked example is a car sold online for $21,000 at a loss, which nets to zero taxable income and still exceeds even the old threshold. The IRS conceded the rule “could also include casual sellers who sold personal stuff... that they paid more than they sold it for.”
The Coalition for 1099-K Fairness, co-founded by eBay, Etsy, Mercari, OfferUp, Poshmark, and Reverb with PayPal, Airbnb, and Block among its members, argued the threshold swept in people “not running a business.”
44M
estimated Forms 1099-K the IRS expected the $600 threshold to produce
Gross income is a poor and politically fragile proxy for business status. Transaction count plus behavioral markers outperforms it. This is the cleanest natural experiment in the research, and it fails income as a trigger.
Quasi-causal for the narrow claim
eBay's 2008 restructuring treated fee economics as the lever to push the seller base toward professional economics. It was imposed, precisely dated, and followed by seller boycotts, reported declines in items sold and prices, and multi-year active-user stagnation.
Two caveats keep the claim honest. The financial crisis is a large confound. And eBay survived to renarrate the episode as a quality push, which is what survivorship bias looks like in practice.
A marketplace that reprices to graduate its base can suppress the casual supply that feeds the professional tier. Directional, crisis-confounded.
Descriptive to directional
Etsy imposed a lifetime 12 percent ad fee on its highest-revenue sellers. Apple's link-out design attracted 34 of roughly 136,000 developers and ended in a court finding of willful violation.
The modest lesson: companies create resistance when they treat growing independence as something to punish rather than something to serve. Both episodes are confounded or narrow, so this is not a universal rule. It is a pattern worth respecting.
Restrictions in the record took three forms: block (Amazon), tax (Etsy, Upwork), or allow with tools (Airbnb, Venmo). Whether any restriction worked is not resolvable from public data.
Directional
06 / The marketplace tension
The behaviors that make a participant more valuable can also make them more capable of leaving. This is the finding marketplace leaders should sit with.
Value
Leavability
Three forms of restriction appear in the record: block (Amazon), tax (Etsy, Upwork), allow with tools (Airbnb, Venmo). Whether any of them worked is not resolvable from public data. The tension itself is documented at the filing level.
07 / The open question
It remains a useful question. It may reveal how dependent someone is on a platform, and how much of the commercial relationship they control. The exposure research calls customer ownership the sharpest divide between older independent earners and platform earners: owner-operators and independent youth earners often owned the customer relationship outright, while platform earners usually do not.
But the historical record does not establish that owned demand causes better business outcomes. Sellers who bring their own customers were already the strongest before anyone measured them. Reverse causality is the default explanation, not a threat to be dismissed. The dated policy shocks, Amazon's messaging limits, Etsy's Offsite Ads, Apple's anti-steering, show platforms restricting or pricing customer access. None provides a clean before-and-after on whether owned demand caused independence.
“Who brought the customer?” may be a better measurement question. It is not yet a proven causal answer.
Unresolved
The honest position is to keep asking it as measurement while declining to treat it as mechanism.
08 / Operator implications
For fintech product leaders
Signals to watch
Do not wait for incorporation or self-declared business status before surfacing business-relevant products. Square and Shopify underwrote credit on processing history; Intuit built after watching users segregate business expenses inside a personal tool. Treat this as a historically informed product hypothesis, not a guaranteed rule.
What would change it: Internal cohort data showing gross-income thresholds predict adoption and repayment as well as complexity signals. The 1099-K reversal is the cautionary precedent.
Directional
For marketplace leaders
Signals to watch
Track value and independence at the same time. Instrument multihoming directly rather than waiting for a lagging churn number; Etsy could see that half its sellers sell elsewhere. Serving professional users with better tools may be more durable than trying to trap the relationship.
What would change it: A restriction that demonstrably raises retention without churn or off-platform migration in the top decile. If churn or price pass-through rises after an imposed change, reverse quickly.
Evidenced (tension) · Directional (prediction)
For consumer-to-business leaders
Signals to watch
Look for business-like usage patterns before forcing a user into a new category. Many freelancers decline the “business owner” label even while operating as one, often reserving it for the moment they hire an employee. Offer identity portability: Venmo spun a business profile off a personal login, and PayPal allowed upgrade at any time. Optional graduation paths may outperform premature reclassification.
What would change it: Cohort data showing revenue or legal-formation triggers convert and retain better than complexity triggers. The historical weight favors complexity.
Directional
09 / Honest limits
10 / Final takeaway
Companies have been drawing the line between consumer and business for decades. The historical pattern is that the line rarely appears first in a legal filing or a self-declared identity. It appears in behavior.
More transactions. More counterparties. More recurring activity. More operating complexity.
The mistake is treating one threshold as the answer: a revenue line, a registration, a label. History suggests a better place to look is a cluster of behaviors that signals a person's economic role is changing, rather than a single revenue threshold or formal label.
And hold the two honest caveats alongside. Exposure raises entry more reliably than durability. And the behaviors that make a participant valuable also make them portable.
The line between consumer and business appears in behavior before it appears in any filing.
Known weaknesses
PastBehavior studies how similar problems have played out, separates durable patterns from attractive stories, and turns that evidence into product, distribution, and execution decisions.
Discuss a product problemRelated reading
The First Economy
The companion essay on ThenForward.
Read the ThenForward essay ↗