Blueballs · Business case · 25 August 2026 · Bumble

The software is free. The placement is the product.

Blueballs gives away the hardest part of starting a bank. What it gets back is the one thing every provider in this market is starving for and cannot buy: a founder at the exact moment they choose their sponsor bank, their card issuer, their KYC vendor and their liquidity. Six ways to charge for that, ranked, priced, and with the cold outreach written.

REPO Josh-Gi3r/blueballs · MIT LIVE blueballs.tech · 181/181 ops ORIGIN/MAIN 8bd5747 NO CODE TOUCHED
01

Start from the true numbers

Everything after this is priced off distribution. So here is the distribution, measured today, not assumed. Two of these numbers are embarrassing and they are the most useful ones on the page — they decide the order you do things in.

GitHub stars
1
0 forks, 0 watchers. Repo is 19 days old — created 6 Aug 2026. And stars are the wrong metric anyway: roughly 1–3% of them are ever buyers. Quote docs traffic and active sandbox builders instead.
Transfers on live stack
0
blueballs.tech/v2/site/stats. 33 accounts, 15 customers, nobody has moved money.
Operations live
181
181 catalogued, 181 implemented, contract-checked against the router.
Release gate
43
Build, types, every suite, Foundry contracts, OpenAPI drift, live tenant-isolation probes.

The one conclusion that follows from those four numbers

You have a finished product and no audience. That rules out three of the six revenue lines for now — you cannot sell ad placement, directory position or managed hosting to an audience of one. But it makes two lines urgent rather than premature: the lifetime revenue share and the origination fee both cost a provider nothing until they work, so the price of that paper is the lowest it will ever be, today, while you have one star. Sign the contracts before the traffic arrives. That is the whole strategy in a sentence.

02

What Blueballs actually sells

It is not a SaaS and it should never be priced like one. Read your own SANDBOX.md again — the Builder journey ends at a step you deliberately left un-automated: "Launch — connect deployment-owned identity, sponsor-bank, card, wallet, payment and compliance providers." That gap is not a missing feature. It is the cash register.

clone / fork free forever brief the Builder markets · currencies · rails provision + settle real double-entry ledger THE MOMENT Launch step choose every provider at once, in one screen sponsor bank / BaaS card issuing identity / KYC / AML stablecoin / on–off ramp FX liquidity / custody money flows back along here
Source: SANDBOX.md product journey (steps 1–5) and docs/ADAPTER-STANDARD.md capability model, origin/main 8bd5747.

Blueballs manufactures intent

A team that has run the Builder has already declared its markets, currencies, capabilities, rails and audience in a structured form. That is a qualified brief, not a lead.

That intent is expensive to buy

Fintech SaaS acquisition cost runs $1,461 SMB, $4,903 mid-market, $14,772 enterprise — the priciest vertical there is, because of compliance diligence.

So you are not asking for a favour

You are undercutting their own funnel. That is the only posture in which a cold email from a 1-star repo gets answered.


03

How everyone else charges for free code

Verified from vendor pricing pages, not blog posts. Two patterns dominate, and neither of them is what Blueballs should copy wholesale — the reason is in the last column.

WhoWhat is freeWhat they charge, exactly ModelRead-across for Blueballs
thirdwebSDKs, contracts, docs $99 Growth / $499 Scale / $1,499+ Pro per month, then $0.015 per wallet MAU, $1 /1k sponsored txns, $8 /M RPC — plus a 2.5% mainnet gas markup Metered infra + a take rate on money moved Closest to the model Josh named. The 2.5% is the important part — the subscription is the floor, the percentage is the business. Note the infra is not self-hostable; the take rate depends on that.
Hyperswitch (Juspay)Whole orchestrator, Apache-2.0, self-host Community free. Enterprise and Cloud both "Contact Us". Recovery, cost observability and reconciliation are paid add-ons Open core + managed cloud The direct structural twin — open source sitting above closed processors. And they publish no price. Nobody in this market has a rate card.
FormanceCore ledger, MIT Enterprise annual subscription, price not published: connectors, wallets, reconciliation, console, designated engineer, 24/7 Open core + support Confirms the ceiling of the pure open-core play: you sell a person and an SLA, not a percentage.
Lerian / MidazSource-available ledger Premium plugins and ancillary services; raised a $5.6M seed on it Free core, paid modules Same conclusion, with a funding round attached. Investors buy this story in this category right now.
SupabaseFull stack, Apache-2.0 / MIT, Docker self-host $25 Pro, $599 Team → $170M ARR by May 2026, up from ~$101M Self-host → cloud conversion Proof the conversion model works at scale — and proof of what it takes: years and enormous adoption first.
n8nSelf-host, fair-code $24 / $60 / $800 per month by executions Usage-metered cloud Shows the pricing metric matters more than the number. Theirs is executions. Yours would have to be settled volume.

And the table that actually matters — platforms that charge the provider side

This is the shelf Blueballs is really on, and none of these companies sell software to the people using them.

WhoWhat they charge the providerWhat Blueballs takes from it
Shopify App Store Developers keep 100% of the first $1M lifetime app revenue, 15% above it. Merchant-referral partners earn up to 20% recurring The threshold structure. Nobody pays until they are winning — which is exactly the shape that makes L2 signable at one star.
G2 Vendors pay a subscription for lead generation and buyer-intent data. Cost per lead, no cut of the transaction The closest analogue to L4, and the reason the first registry deal should be priced per qualified builder reached rather than as a revenue share.
Airbyte connector marketplace Third-party connector maintainers get a share when their connector runs in paid cloud, and take on SLA duty for it. Percentage undisclosed The structural twin of L3 and L6: maintenance obligation traded for a commercial position. It is also proof that "who maintains the adapter" is the negotiable asset.
Terraform Registry Nothing. A Partner Premier badge buys distribution and legitimacy; HashiCorp monetises its own cloud instead The cautionary one. A badge with no price attached is a giveaway — if L3 does not carry a fee and an expiry, it is Terraform's model by accident.
Lithic Runs a named Integration Partner Program — co-built certified integrations, shared Slack channels, mutual referral terms. Commercial terms undisclosed The best single first target in this whole document. A named programme and a named Head of Partnerships means the door already exists; you are not inventing the category for them.
Developer advertising Carbon / EthicalAds $2–10 CPM, developer newsletters $60–150 CPM, conference sponsorship $5k–50k+ The floor on what these providers already spend to reach developers — and they are spending it on impressions, not on someone standing at the selection moment.

Why copying open core alone would be a mistake here

Every company above monetises the person running the software. That works when your users are venture-funded and numerous. Blueballs' users are pre-launch fintech founders — the least able to pay of any developer segment, and there are not many of them. Charging them is a small, slow, hard business.

The people with money in this picture are standing on the other side of the Launch step, and they spend $1,461–$14,772 to meet exactly the person you already have. Charge them instead.

The one thing thirdweb proves and the others don't

A percentage of money moved beats a subscription, and it is collectable even from users who would never pay a monthly fee. thirdweb's 2.5% gas markup rides on infrastructure they host, which is why their SDK is open and their infra is not.

Blueballs cannot copy that directly — MIT means anyone can fork, self-host and never pay. So the take rate has to be collected from the provider, not the deployer. Same percentage logic, opposite counterparty. That inversion is the whole business case.


04

Six revenue lines

Ordered by how fast they produce a first invoice, not by size. L1 and L6 pay the rent. L2 is the actual business and it is the one that costs a counterparty nothing to sign today. L5 is last on purpose.

L1 · SIGNABLE NOW

Origination fee — the provider pays per programme you send them

One-time cash when a deployment picks a provider adapter and goes live on real credentials. No volume required, no traffic required, no product change beyond attribution.

Price
$1,000 – $5,000 per qualified integration — deliberately set at 25–50% of their own acquisition cost so the answer is arithmetic, not goodwill
Anchored to
Fintech SaaS CAC $1,461 SMB / $4,903 mid-market; blended B2B fintech cost per lead ~$452 for a cold one
Payback
The line that closes it: a builder doing $500k/month of card volume earns the BaaS stack $1,500–$3,600 a month. A $1,000–$3,000 fee pays back in one to three months — say that number out loud in the first call
Hook in tree
Adapter descriptor id / capability / maturity in docs/ADAPTER-STANDARD.md
Blocker
Attribution. See section 05 — without it this is unenforceable
Time to first £
Ceiling
L2 · SIGN NOW, EARNS LATER

Lifetime share of the provider's take — "referral for life", done properly

Not per signup. A slice of what the provider earns from every programme you originated, for as long as that programme runs. This is the compounding line and the one that makes the company worth something.

Ask
10–20% of the provider's own revenue on originated programmes, perpetual, no cap, survives your acquisition
Re: Kast
Kast's "referral for life" is marketing language, not a perpetual share. It is a milestone ladder — $20 at $100 spent, +$30 at $500, +$50 at $1,000, +$150 at $5,000 — with a hard ceiling of $250 per referred user, and terms they can change at will. True perpetual comps do exist and are worth citing instead: Ahrvo publishes 10–30% of net revenue per referred client, monthly and ongoing. You are not copying Kast. You are asking for the thing Kast only implies.
In cards
Blended interchange runs 100–240 bps; the standard split leaves the fintech ~70–170 bps and the BaaS ~30–70 bps. Your 10–20% of their slice ≈ 3–14 bps of programme volume
Named split
Synctera publishes its own: the bank receives 100% from the network and passes 70% to the fintech; Synctera takes about half of the bank's share — roughly fintech 70 / bank 15 / Synctera 15. Use a published split, not an industry average, when you name your number to them
Worked
One $50M/yr card programme → $15k–$70k a year, recurring, from a single origination, at zero marginal cost
Biggest
Stablecoin issuance. Bridge's Open Issuance advertises 3–4% APY reserve yield to issuers — a share of that on originated float is the largest single number on this page
Time to first £
Ceiling
L3 · SIGNABLE NOW

Certified Adapter — the provider pays to be marked production

Your adapter standard already defines four maturity levels and an evidence bar: conformance suite, failure mapping, reconciliation, observability, runbook. That bar is a product. You are selling verification, not position.

Price
$10k – $50k per capability, per year — includes the conformance suite kept green against their live API and a dated badge that expires
Why it sells
Nobody else can independently certify a fintech API from inside a working 181-operation stack. It is the only claim here a competitor cannot copy by writing a blog post
Needs no traffic
Correct — this is engineering credibility, not audience. It works at 1 star
Hook in tree
maturity: reference | experimental | production | deprecated and production_ready is already evidence-based, never inferred
Time to first £
Ceiling
L6 · FASTEST FIRST INVOICE

Sponsored capability build — they pay for their adapter to exist

A provider with no open-source footprint pays Blueballs to build and maintain the reference adapter against their API. Engineering-as-marketing for them, funded R&D for the repo, and it feeds straight into L3.

Price
$25k – $75k per capability, plus an annual maintenance retainer
Why first
It converts a cold conversation into a scoped statement of work in one step. It is the only line here that a provider's existing budget already has a name for
Second effect
Every sponsored build makes the stack more useful, which grows the very distribution the other five lines need. It is the only line that pays you to fix your own weakness
Watch
Never let a sponsor's money buy a claim of production readiness — that is L3's gate and it must stay independent, or you have sold the only thing you have
Time to first £
Ceiling
L4 · NEEDS AUDIENCE

The provider registry — your "card reviews", with a rule attached

You already write these and you already write them honestly: docs/partners/BRIDGE.md records that Bridge's sandbox fires no payment webhooks and can return dummy data. Nobody else audits provider sandboxes from inside a running stack. That honesty is the asset.

Model
Cost-per-qualified-signup on the outbound click. Consumer analogue: $50–$150+ per approved card application, NerdWallet up to $100; Kast pays up to $250 per referred user with tier bonuses and a sub-affiliate share. B2B developer signups price higher than either
Hard rule
The ranking is never for sale. Sell the click, never the order — and say so publicly, dated, in the repo. Your provider inclusion policy already separates "relationship" from "technical maturity". Keep that wall or L4 destroys L1–L3
Already built
Most of this exists in the tree already. src/ecosystem/providers.ts carries 56 providers across ~37 categories — sponsor banks, issuing, KYC, stablecoin, custody — each with access, sandbox, categories and modules, and each defaulting to relationshipStatus: "No relationship claimed" / technicalStatus: "Link only". Those two fields are the paid-versus-earned wall, already modelled. L4 is a commercial decision, not a build
Gate
Meaningless below roughly 2,000 monthly developer sessions. Do not pitch it before then
Time to first £
Ceiling
L5 · LAST, AND NARROW

Managed core — but only the two things nobody can self-host

Do not sell "hosted Blueballs". Sell precisely the two limitations your own README already admits: single-writer SQLite persistence, and a Builder Agent with no cost ceiling. Those are the only two things a serious team genuinely cannot run themselves.

Product
Clustered multi-region ledger runtime, backups and point-in-time restore, a metered agent with a spend cap, and the 43-check release gate run against their fork on every commit
Price
$499 – $2,500 / month. Benchmarks: Supabase $25/$599, n8n $24/$60/$800, Hyperswitch and Formance both "contact us"
Why last
MIT means the fork is always free. This line only converts at real adoption scale and it competes for the same engineering hours as L3 and L6, which pay sooner
Time to first £
Ceiling

Where the L2 basis points actually come from

The card case, drawn to scale. You are not taking a bite out of the founder's margin — you are taking a slice of the BaaS's slice, which is why the founder never feels it and never objects.

BLENDED INTERCHANGE ON PROGRAMME VOLUME — 100 TO 240 BPS 100 – 240 bps collected from the merchant side STANDARD SPLIT — FINTECH 70% / BAAS 30% the fintech keeps ~70 – 170 bps BaaS / sponsor ~30 – 70 bps BLUEBALLS ASKS FOR 10–20% OF THIS BAR ONLY ≈ 3 – 14 bps of programme volume, perpetual
Interchange band and the 70/30 convention: Synctera, "The FinTech's Guide to Interchange Revenue". Bars drawn proportionally.

05

The one thing that has to exist first

There is currently no way to prove you originated anything

I read the tree. Nothing in the adapter descriptor, the Builder blueprint or the outbound provider links carries an origination token. Which means L1, L2 and L4 are unenforceable today — you would be asking a provider to self-report revenue owed to you, which no provider has ever done.

The fix is small and it is a product decision, not mine to make: a referral identifier in the capability descriptor, a Referred-By header on adapter calls, and tracked outbound links from the registry. Do not sign a revenue-share contract before this exists — you would be locking in a number you cannot invoice.

Owner: Josh + whoever holds the code lane. Not me — I have not touched code and will not.

And the strategic consequence of the MIT licence

Anyone can fork Blueballs, strip every reference to you and ship. So the revenue lines cannot depend on the repository — the repository is the free part and always will be.

What they depend on is the registry and the conformance gate, both of which are services you host and control. A fork gets the code. It does not get the certification, the badge, the directory listing or the conformance suite running green against a live provider API.

That is the moat. Build it deliberately. Every hour spent making the certification more rigorous is an hour spent on the only defensible asset in the plan — and it happens to be the thing the repo is already best at.


06

Who to hit, in what order

Ranked by three tests, each of which you can check in five minutes without talking to anyone: does it have a self-serve sandbox (proves developer-first), does it have a public partner or affiliate page (proves it already pays for distribution), and is it already named in your repo (proves you have a reason to write). Highlighted names score on all three.

A · Stablecoin & crypto infrastructure

Fastest movers, developer-first, self-serve sandboxes, and structurally hungry for exactly the distribution you have. Two of them are already sitting in docs/partners/. Start here.

FIRST
DakotaBridgeRainBrale Zero HashBVNKConduitReapCrossmintIronNoah

D · FX, rails & liquidity

Where Blueballs is genuinely unique — nobody else ships an open, policy-gated, reserve-before-firm FX runtime with a Solidity kernel. Highest differentiation, so the highest-quality first conversation.

SECOND
Modern TreasuryNiumCurrencycloudThunes iBanFirstTransFiWise Platform

C · Identity, KYC & compliance

Easiest to sign and lowest per-deal value — which makes them the right place to prove the mechanism works before you spend a swing on a big card deal. Sumsub already runs a public three-track partner programme with referral and reseller margins.

SECOND
SumsubPersonaAlloyVeriff ComplyAdvantageFootprint

B · Card issuing & BaaS

The biggest lifetime value by far — interchange compounds — and the slowest, most compliance-gated door, with one exception. Lithic runs a named Integration Partner Program with a named Head of Partnerships — certified integrations, shared channels, mutual referral terms. That door already exists, so it belongs in your first batch even though the segment as a whole comes third.

Leverage worth knowing before any of these calls: roughly 45% of BaaS programmes sit with banks under formal enforcement action. These providers do not need more leads — they need cleaner, better-qualified ones. A platform that pre-qualifies a builder before they arrive is selling the thing they are actually short of.

THIRD · except Lithic
LithicHighnoteMarqetaGalileo SyncteraUnitColumnIncreaseQoloAlviere

E · Adjacent open source and platforms

Not a revenue line — a distribution line, and the cheapest fix available for the 1-star problem. Blueballs runs on Cloudflare Workers with Durable Object SQLite; that is a case study Cloudflare's developer marketing wants and pays for in reach rather than cash. TigerBeetle and Formance are category-adjacent rather than competitive: co-marketing, not conflict.

ALWAYS ON
CloudflareTigerBeetleFormanceLerian / Midaz SupabaseHyperswitchApache Fineract / Mifos

07

The cold outreach

This is the part that is genuinely new, and the standard playbook fails here — a partnership email from an unknown 1-star repo goes in the bin, correctly. So do not send a partnership email.

1 · Ship the artefact before you write

For every target, publish the honest capability descriptor first — exactly like docs/partners/DAKOTA.md already does. Then you are not asking for anything. You are a fact in their world that arrived without permission, and you did unpaid work about their product.

2 · The reader is DevRel, not sales

Order: Developer Relations / Developer Experience → Partnerships / Ecosystem → Growth. DevRel replies because you shipped something about their API. Sales does not reply to unknown open-source projects, and partnerships will not move without an internal advocate. Title patterns to search: Head of Developer Relations, Ecosystem Partnerships, Partner Engineering.

3 · The subject line carries a fact, not an offer

Not "Partnership opportunity with Blueballs". Something more like "We mapped Bridge into an open-source neobank stack — 4 things your sandbox doesn't do". One is a request. The other is information they cannot get anywhere else, about themselves.

publish descriptor public, dated, honest ask in their Discord a real sandbox question DM the DevRel lead ≤ 60 words, link only email partnerships cc the DevRel reply ONLY NOW the money ask L6 scope or L2 paper
The money never appears in the first message. It appears once they have replied about their own product.
Cold DM — DevRel lead, X or LinkedIn≤ 60 WORDS · NO ASK
Hi {name} — we maintain Blueballs, an MIT open-source neobank stack
(181 banking operations, double-entry ledger, FX engine, self-hostable).

We mapped {provider} into it and published what we found, including
that {specific honest finding — e.g. "your sandbox doesn't fire
payments webhooks"}.

{link to the descriptor}

Nothing to sell. Just tell me if we got anything wrong.
Cold email — Partnerships / Ecosystem≤ 130 WORDS · ONE ASK
Subject: we mapped {provider} into an open-source bank stack —
{n} gaps in your sandbox

{Name},

Blueballs is MIT-licensed open source for building a neobank: 181
banking operations, a double-entry ledger, a stablecoin FX engine,
running at blueballs.tech. Founders clone it, describe the product
they want, and get a working environment in minutes.

At the end of that flow they choose a sponsor bank, a card issuer,
a KYC vendor and a liquidity provider. Right now that screen is
blank.

We already publish a capability map for {provider}: {link}

Two questions. Do you want the reference adapter built and
maintained, and do you want to be the default in that screen?

Happy to do the first one before we talk about the second.

{Josh}{role}, Blueballs
The follow-up that carries the money — sent only after a replyTHE ASK
Short version of the commercial shape, so you can route it
internally.

Now, no money: we build and maintain the {provider} adapter
against your public sandbox. You review the PR. It ships in the
repo either way — we would rather it were correct.

If you want it certified: we run a conformance suite against your
live API and keep it green. That carries a dated production badge
that expires if the suite goes red. {$X}/year per capability.

The part worth your attention: when a deployment picks you at the
launch step, that programme was originated here. We would like
{10–20}% of your revenue on those programmes, for the life of
the programme.

Two things about that number. It costs you nothing until it
works — there is no floor, no minimum, no fee. And your own
acquisition cost in this segment is four figures at the low end,
which is the arithmetic this is priced against.

We are early and we are asking early. Terms will not be this
cheap once the funnel is running.

Objections you will get, and the honest answer to each

"You have one star." Correct, and that is why the revenue share is free to you and why you should sign it now. You are not buying traffic. You are buying an option on traffic, at the only price it will ever be.
"We don't do open source." You do not have to. We maintain the adapter, you review it. Your name appears on a capability map, not in your repo.
"What's the catch — do we have to pay for ranking?" No, and it is written down publicly with a date on it. The inclusion policy separates whether we claim a relationship from whether the integration is technically mature. We sell the click. We have never sold the order.
"Legal will take six months." Then start with the descriptor, which needs no paper at all, and the adapter, which is an ordinary vendor SOW. The revenue share can follow.
"How do we verify what you originated?" Fair, and today the honest answer is that you cannot — which is why this is the one thing being built before any contract is signed. Do not bluff this one.

Three things that will kill this if you let them

  • Selling ranking. The moment position is purchasable, the registry is an advertorial and the certification is worthless. It converts your only defensible asset into a commodity, for one cheque.
  • Signing revenue share before attribution exists. You will spend a year discovering you cannot invoice it, and you will have burned the one moment when your terms were cheapest.
  • Chasing the founder for money. Pre-launch fintech founders are the poorest buyer in the developer market. Every hour spent building a paywall for them is an hour not spent on the side of the table where the money is.

And one thing that will make it work faster than anything else

  • Publish the honest sandbox audits whether or not anyone pays you. "The open-source project that tells you what fintech APIs actually do in test mode" is a position no incumbent can take, because they all sell one. That reputation is what makes every email in this section get opened.

08

Order of operations

Sequence, not a schedule. Each step unblocks the next one; nothing here waits on a date.

  1. Build attribution

    Referral identifier in the capability descriptor, a Referred-By header on adapter calls, tracked outbound links from the registry. Small, and it gates three of the six lines. Code lane, not mine.

  2. Write the commercial policy into the repo, publicly and dated

    What is for sale (adapters, certification, clicks) and what is not (position, ranking, the word "production"). This is the document you link in every cold email, and it is the reason the answer to "what's the catch" is a URL rather than a promise.

  3. Publish six more capability descriptors before contacting anyone

    Rain, Zero Hash, BVNK, Sumsub, Lithic, Modern Treasury. Same shape as DAKOTA.md and BRIDGE.md, same honesty about what their sandbox does and does not do. Each one is a reason to write to that company that is not a pitch.

  4. Run segment A and C in parallel — twelve first contacts

    Descriptor → Discord question → DevRel DM → partnerships email. Nothing about money in any of them. Measure replies, not sends.

  5. Convert the first reply into a sponsored adapter (L6)

    One paid, named integration does more for lines 1–5 than any amount of further outreach, because it turns "we mapped your API" into "we maintain their integration".

  6. Sign revenue-share paper with everyone who will take it

    Including — especially — the ones who are not paying you anything else. It costs them nothing, which is exactly why you can get it now and not later.

  7. Only then look at the registry and hosting

    L4 and L5 become real at roughly 2,000 monthly developer sessions and a few dozen live deployments. Pitching them earlier teaches providers that your numbers are aspirational, which is expensive to undo.

What this is built on

Repository facts read at origin/main 8bd5747, 25 Aug 2026: README.md, VISION.md, ROADMAP.md, SANDBOX.md, docs/ADAPTER-STANDARD.md, docs/partners/{README,DAKOTA,BRIDGE}.md. Live facts from blueballs.tech — /v2/site/stats returned 181/181 operations, 33 accounts, 15 customers, 0 transfers, source commit 5739476 (the live deploy is three commits behind main). GitHub: 1 star, 0 forks, 0 watchers, created 2026-08-06. No code was modified.

External sources, primary where a primary source exists:

  • thirdweb pricing page — plan prices, per-MAU and per-request overage rates, 2.5% mainnet gas markup
  • Hyperswitch pricing page (Juspay) — Community free, Enterprise and Cloud unpriced, paid add-ons
  • Formance pricing page — MIT core, enterprise subscription unpriced
  • Synctera, "The FinTech's Guide to Interchange Revenue" — 100–240 bps blended interchange, 70/30 fintech/BaaS split, 70–170 bps to the fintech
  • Lerian / Midaz — source-available ledger, premium plugins model, $5.6M seed
  • Supabase — $25 Pro / $599 Team, $170M ARR May 2026 (Sacra); n8n — $24 / $60 / $800 tiers
  • Fintech CAC benchmarks 2026 — $1,461 SMB, $4,903 mid-market, $14,772 enterprise
  • Credit-card affiliate CPA range $50–$150+ per approved application; NerdWallet up to $100; Kast affiliate programme up to $250 per referred user, tier bonuses, sub-affiliate share (Kast affiliate T&Cs delegate all rates to product pages)
  • Bridge Open Issuance — 3–4% APY reserve yield to issuers; stablecoin infrastructure landscape 2026
  • Sumsub partner programme — referral, ecosystem and technology tracks; commission terms not published

Second research pass, folded in after first publication. The Kast milestone ladder and its $250 ceiling, Synctera's published 70/15/15 split, the $500k-TPV payback arithmetic, the ~45% enforcement-action figure, the Shopify / G2 / Airbyte / Terraform / Lithic precedents and the developer-advertising CPMs come from Fizz-Claude-5's three research sweeps posted in the same thread (full versions in .scratch/biz-case/sweep-{1,2,3}-*-FULL.md). The src/ecosystem/providers.ts inventory was flagged by the same agent and reproduced here against origin/main 8bd5747: 56 entries, counted. Also confirmed there: daxota.xyz does not resolve — the company is Dakota, dakota.xyz, already mapped in docs/partners/DAKOTA.md.

One honest note on the sources. Not one provider in this market publishes its referral or revenue-share terms — not Stripe, not Sumsub, not Lithic, not Vercel for its own marketplace. I checked. That is not a gap in the research; it is the finding. Every number in section 04 is a number you propose, not a number you accept, and the first side to name one usually sets the range.

Bumble · 25 August 2026 · research and commercial assessment only · no code touched · attribution dependency in section 05 is unresolved and owned by the code lane