Paying for customers, and for the people who find the people who find them

Most products are bought on someone's recommendation: an accountant, a coach, a community owner, a colleague. A referral program pays those people for the customers they send. Most referral programs fail for a simpler reason than bad terms: the people best placed to recommend the product never hear that the program exists.

ViralPayouts adds a second paid role. A partner earns a commission on each approved customer they bring. A recruiter earns an override, paid on top, on the approved sales of the partners they brought in. Nobody is paid for signing anyone up, and there are two levels, never more. The brand pays only when a sale is approved, ViralPayouts pays every partner for it, and every amount can be traced from the click to the payout.

The aim is bigger than a tool. ViralPayouts is meant to become the place where a brand puts a price on its next customer, and a two-level network of people who already know the buyers goes out and finds them. This paper sets out why that should work, how the money moves, what it costs, how it grows into a network, and how we will find out whether the second level really brings a brand more customers. That last question is open. Nothing here is a result: no brand, partner or payout is live on ViralPayouts yet.

The affiliate model is not broken. It pays for the wrong last step. Today a program pays the person whose link the buyer clicked. It pays nothing to the person who brought that affiliate in, even when finding the right affiliate was the hard part. So nobody is paid to go and find the creators, advisers and community owners who move the needle, and most programs wait for them to turn up.

Take a company selling tax software to small businesses. Its best partner is not a deal site. It is an accounting firm that uses the software for all of its clients, and every client of that firm becomes a customer. Whoever introduces the software to that firm has done more for the brand than any single referral. Under a normal program they earn nothing for it. Under ViralPayouts they earn an override on every approved sale the firm brings, for as long as the brand's terms say.

These people are gatekeepers: they sit between a brand and many buyers who already trust them. A second paid level turns finding gatekeepers into paid work, done by people who know a niche, on approved sales only. The brand gets the sales team it would otherwise have had to hire to cold-call them.

A brand launches a partner program. It sets a commission, writes a landing page and waits. The people who could send it customers are busy, scattered and unaware of the program. Finding them one by one is the cold outreach the program was meant to replace. This is the cold start.

The second problem is measurement. A partner often cannot see which click earned a commission, and a brand often cannot see which partner brought which customer, or why a commission was reversed. When nobody can check the numbers, good partners leave and bad traffic stays.

• Cold start: the program exists, but the people who could use it do not know.
• Measurement gap: the brand and the partner cannot both see the same record of what happened and why.
• Payout work: most affiliate software tracks sales but leaves the brand to pay every partner itself, with the transfer fees, tax forms and chasing that come with it.

One person can be both a partner and a recruiter. Each partner gets two links, kept separate: a customer link and a recruit link. A sale never quietly attaches a recruiter, and a recruit link never counts as a sale. Recruiting is per program: a recruiter earns on what their recruits sell for that brand, not for every brand the recruit later joins.

The brand chooses how the override is worked out: a share of the partner's commission, a share of the sale, or a fixed amount, from 1% to 95%. The guided setup starts it at 10% of the partner's commission, because the second level is the point; a brand can change it or turn it off. Commission can be 1% to 95% of a sale, or a flat fee of at least $1 per approved customer. The override is paid on top, so a recruited partner never earns less, and partner and recruiter together can never be paid more than the whole sale.

Why this is not multi-level marketing: nobody pays to join, nobody is paid for recruiting as such, and there are only two levels. A recruiter earns only when a partner they brought in produces an approved sale to a real customer, and that override is reversed whenever the sale is.

An illustration of the intended flow, not a customer story. No brand has run it yet.

A founder selling accounting software opens ViralPayouts, or asks their AI agent to. In a few minutes there is a program: $400 for every firm that signs up and pays, a 10% override for whoever brings in the accountant who referred it, and a clear rule for what counts as a valid customer. The program gets its own public page on viralpayouts.com.

It does not sit there waiting. Recruiters on the network see a new bounty in a niche they know. One of them messages forty accountants that week. Some join from a page that leads with the brand. A few mention the software to clients they already advise. The brand approves the customers that are real, the bounties are paid, and the recruiter's override lands in the same ledger row. Nobody bought an ad. Nobody was paid for signing people up. Everyone can see where the money came from.

Affiliate software gives each brand its own island: its own sign-up page, its own partner list, its own payout set-up. A partner who joins one program starts from nothing at the next. ViralPayouts is built the other way round.

• One account per person across every program, with one public page (viralpayouts.com/u/<handle>) and one payout connection.
• A public directory of programs, searchable by niche, so a partner who joins for one brand finds the next.
• A share of our profit for growing the network itself: 20% to whoever brings a brand, 5% to whoever recruited them, and 5% to whoever first brought a partner to ViralPayouts, on that partner's payouts in every program they join.
• A community where partners learn each program and recruiters find each other.

Each new brand gives every partner more to sell. Each new partner gives every brand more reach. That is the flywheel. It only turns if everyone can check the numbers, which is why the ledger comes first.

Every amount owed is a row in a ledger that only ever adds rows. A refund or a rejected sale writes reversing rows for both levels and for the fee. Nothing is edited or deleted, so every balance can be rebuilt from its entries. A set of rules about the books (for example: every fee matches the published schedule, nobody is paid more than they were credited) is checked after every test run and every night.

The rates in this example are chosen for illustration. Each brand sets its own.

A commission becomes payable 15 days after the sale by default (never more than 30), after the refund window. Once it is payable the brand cannot reject it except for fraud. Brands are invoiced on the 1st and 16th; partners are paid as soon as the brand pays, at the latest on the 13th or 28th. The brand funds payouts by bank transfer, never by card, and ViralPayouts sends each partner their money, so the brand makes one payment instead of one per partner.

If the customer is refunded in full, the commission, the override and the fee are all reversed; a partial refund reverses each in proportion. A refund that arrives after a commission was already paid is not clawed back from anyone's bank: the partner's balance in that program goes below zero, and the shortfall comes out of their next commissions there.

A brand can reverse a commission before it is payable only for a listed reason, and the partner sees which: the customer was refunded; the payment was charged back or disputed; the customer did not meet the valid customer rule; the sale was a duplicate or a test order; the customer was the partner, someone in their household, or the brand's employee; the sale came from a banned source or banned conduct; the coupon code was posted on a leak or deal site without permission; the sale was fraudulent.

A partner's link sends the visitor to the brand's site with a referral code and a click id. The sale is then reported to ViralPayouts, whichever way the brand sells: a signed request from its own server, Stripe, Shopify or Whop, each connected without the brand writing code. A partner's coupon code counts too, and renewals of a subscription are credited to the partner who brought the first payment. A report sent twice is recorded once.

• Cookie window: 7 to 365 days, 30 by default. A sale after the window is not credited.
• Who is credited: the last click by default, or the first click if the brand chooses. A partner's code the buyer typed wins over a link.
• A sale that matches no partner is listed for the brand to look at. It is never guessed.
• A sale dated before the click it is matched to is refused, because the click cannot have led to it.
• Commission is worked out on what the buyer paid for the product after discounts, without tax.

Most affiliate software charges a brand a monthly fee whether or not its program works, then leaves it to pay each partner itself. ViralPayouts charges nothing until a partner is paid. Our income only grows when a brand's program brings it customers, so we are on the brand's side by construction.

The fee is a share of what a program pays its partners in a calendar month, on top of that amount. It starts at 10% up to $1,000 paid to partners and falls steadily to 3% at $10,000. The rate applies to the whole month, and the total fee never goes down when payouts go up. Paying the partners is included: the brand does not pay transfer fees on top. Programs paying partners $100,000 a month or more can ask for a custom rate.

• Bots, link previews and repeat clicks are filtered before a click is counted.
• A sale is flagged when the customer appears to be the partner, or when the click came from an internet address the brand's console or the partner's portal was used from.
• A sale is flagged when the brand's own checkout also paid an affiliate commission on it, so the brand does not pay twice.
• Joins with throwaway email addresses, the brand's own domain, or many joins from one address wait for the brand to approve.
• A recruiter is paid only on approved sales by their recruits, so inventing partners earns nothing.
• The brand sees the whole tree: every partner, every recruiter and who recruited whom.

Fraud protection is not finished. Refund abuse and organised self-referral need more work before money scales, and we say so on the site.

Every program runs on one standard agreement, and a partner reads and accepts it before getting a link. A brand chooses its numbers inside fixed limits, and may add short rules of its own, which ViralPayouts approves before partners see them.

• Partners must say they may be paid, before the reader clicks.
• A brand can raise what partners earn at once. A cut takes effect at least 3 business days after it is announced. Every partner is emailed either way, and each sale is paid on the terms in force when it was made.
• Paid search on the brand's name, cashback and paid-to-sign-up rewards, and browser extensions are banned by default.
• Customer emails are not stored in plain text. Partners see what they earned, not who the customers are.

The agreement is a working draft and has not yet been reviewed by a lawyer. Holding and paying out brands' money, and collecting tax forms, need a legal review before the first real payout.

Two-level commissions are not new. FirstPromoter documentation: multi-tiered commissions shows how FirstPromoter supports an override as a share of the commission or of the sale. Among the real programs we documented (checked 10 October 2026), 4 of 8 pay a recruiter: Paraform, HighLevel, Undetectable AI, LiveChat (Text). In each, it is a line in the terms rather than the centre of the product.

ViralPayouts differs in three ways at once: it charges only when partners are paid, it pays them for the brand, and it puts the recruiter level and the shared network at the centre. The comparison pages on viralpayouts.com set out each tool's own published terms, with links and dates.

None of the companies named here uses ViralPayouts or endorses it. Their terms are advertised offers, not results.

• Phase 1, prove it with one brand. A closed pilot with one brand in one niche, run as a two-level program next to a comparable single-level one, so the result shows whether the second level causes growth or just comes with it. The first pilot is with a seller on Whop, where the founder already knows both the sellers and the community owners who act as gatekeepers. Business software, where one adviser brings many customers, is the main market and the second pilot.
• Phase 2, the network. Every brand becomes supply for every partner: the program directory, partner pages, the partner community, and our own profit-share program, which makes every partner a salesperson for ViralPayouts itself.
• Phase 3, agent-native. A founder tells their AI agent to set up a program, and it does: drafts it, connects tracking, sends a test sale and returns one link to review. The REST API and the MCP server for this are built; a person still funds, publishes and approves.
• Phase 4, the public bounty board. Large public bounties are marketing in their own right. At scale ViralPayouts is where they live: real programs, real payout histories from the ledger, and the right people hearing about programs they would never have found.

The claim to test is narrow: for the same brand and the same commission, adding a paid recruiter level brings more approved customers per month than a single level, at a cost per customer the brand accepts. The number that decides it is approved customers per active recruiter per month, against the single-level control. Supporting measures: how many recruited partners become active, approved customers per active partner, and days from a program going live to its first approved sale.

• Proven wrong if recruited partners rarely bring a customer, so overrides cost money without adding customers.
• Proven wrong if the two-level program does no better than a comparable single-level one over the same period.
• Proven wrong if recruiting mainly produces low-quality sign-ups or reversed sales.
• Risk: being read as multi-level marketing. The answer is the rules in section 4, said plainly everywhere.
• Risk: two cold starts at once, brands and partners. The answer is to start where the founder already knows both sides.
• Risk: one payout provider, and holding brands' money. Both need a real test transfer and a legal review before money scales.

This is a hypothesis. We will publish the pilot result, whichever way it goes.

Built and tested: the two links per partner and the override rules; sale reporting from a brand's server, Stripe, Shopify and Whop, with coupon codes and renewals; the add-only ledger with reversals; twice-monthly brand invoices and payouts through Whop or recorded bank transfers; brand and partner accounts with one account across programs; the standard program terms; the program directory; a REST API and an MCP server; and a test network that runs the same software with simulated money and a time machine.

Not done yet: a real payout and a real brand funding have only been tested against stand-ins, fraud protection needs more work (section 10), and the agreement needs a lawyer's review.

Not live yet: no brand, partner or payout is live on ViralPayouts. The first step is the pilot in section 13.

• ViralPayouts fee schedule and own program: src/lib/fees.ts in the ViralPayouts code, shown on viralpayouts.com/pricing.
• Program limits, reversal reasons and the standard agreement: src/lib/program-terms.ts, shown on viralpayouts.com/program-terms.
• Payout calendar: src/lib/payout-calendar.ts.
• FirstPromoter, FirstPromoter documentation: multi-tiered commissions: https://help.firstpromoter.com/en/articles/8971329-how-to-set-up-multi-tiered-commissions-in-firstpromoter
• Micro1, micro1.ai/company-referral: https://www.micro1.ai/company-referral (checked 10 October 2026)
• Paraform, paraform.com help: Refer a recruiter: https://www.paraform.com/help/article/recruiter-referrals (checked 10 October 2026)
• HighLevel, gohighlevel.com/affiliate-policy: https://www.gohighlevel.com/affiliate-policy (checked 10 October 2026)
• Undetectable AI, undetectable.ai/earn: https://undetectable.ai/earn (checked 10 October 2026)
• LiveChat (Text), partners.livechat.com/help/partner-referral: https://partners.livechat.com/help/partner-referral/ (checked 10 October 2026)
• Shopify, help.shopify.com: Affiliate earnings: https://help.shopify.com/en/affiliates/earnings (checked 10 October 2026)
• HubSpot, hubspot.com/partners/affiliates: https://www.hubspot.com/partners/affiliates (checked 10 October 2026)
• Dropbox, help.dropbox.com: Refer friends: https://help.dropbox.com/storage-space/earn-space-referring-friends (checked 10 October 2026)