This is the first entry in a new series I’m calling Remastered. I take my earliest posts, the ones that started as Twitter threads, and rebuild them into real pieces: revisit the context, translate them out of tweet shorthand, and update the players, the economics, and the thinking with current data. Five years of writing later, some of these early ones have earned a second pass.
We’re starting with a classic: how to build a neobank from scratch. I posted it in 2021, back when almost nobody knew who I was. It was awesome to see the positive engagement and it pushed me into the stratosphere of finfluencer territory, and it was one of my first real moments of being proud of the work and watching it prove useful to the ecosystem.
The thesis hasn’t moved: a neobank is not a bank. It’s a stack of vendors you assemble on top of someone else’s charter, and the craft is picking the right partner at each layer and negotiating the economics. What has changed since 2021 is the cast. For each layer I cover what it does, the players in 2021, the economics (which mostly still hold), and what changed.
The original thread stopped at thirteen layers. The full 2026 stack runs to twenty-six, one for every letter of the alphabet, so this is the A to Z of it, literally. Everything is linked, including Eric’s post, still the cleanest single map of all the pieces. A few of these earn their own deep dives (bank sponsors especially), and I’ll link those as they go up.
A. Acquiring processor
What it does: Lets you act as a merchant / acquirer on a network rather than an issuer, which is how you pull money in and push it out. Pull payments are instant-funding AFTs (account funding transactions); push payments are instant payouts (OCTs).
Players (2021): Finix, Stripe, and TabaPay.
Economics: A fee per transaction plus an interchange fee.
What’s changed: Visa Direct and Mastercard Send moved from underliers to direct relationships with issuers, so you can go straight to Visa for Visa Direct instead of routing through a platform.
B. Bank Sponsor
What it does: Your bank sponsor rents you its charter. It interacts directly with the FDIC and owns compliance, the regulatory obligations, and the ongoing FDIC reporting. You rent the charter, which lets you hold member deposits on its platform.
Players (2021): The Bancorp, Sutton Bank, and WebBank.
Economics: Basis points of purchase volume plus a fee per transaction, plus a reserve / collateral account. That structure has largely held.
What’s changed: The Bancorp is still the common traditional BaaS platform, but the action moved to fintech-minded banks built for this from the ground up. Lead Bank, run by Jackie Reses, who led Square’s banking efforts. Column, founded by former Plaid co-founder and CTO William Hockey as a dedicated fintech bank. Erebor, backed by Palmer Luckey and increasingly used by fintechs as a sponsor. Increase, now a chartered bank in its own right. I’d also watch what CapOne does with Brex - there could be an interesting BaaS platform at play here.
Everything above assumes you go direct: negotiate and integrate each layer yourself, for maximum control and the best economics at scale. The other path is a banking-as-a-service platform that bundles the regulated plumbing behind one contract. Unit, Synctera and TreasuryPrime are the main providers and offer at least 50% of the 26 layers in house with no additional vendors.
C. Core banking
What it does: The platform that tracks balances, enables products, manages accounts, and maintains the ledger.
Players (2021): Fiserv, FIS, and Q2. Galileo can also serve as your core.
Economics: Fees per account, per transaction, and per API call.
What’s changed: The traditional players have stayed more or less the same other than the entry of Nymbus, but now modern sponsor banks like Column and Increase have entered the fray.
D. Direct-deposit switching
What it does: Makes switching a member’s direct deposit instant, which is how you win and engage high-spend members. The member logs in with employer credentials and the switch happens automatically, no HR ticket or Workday hunt.
Players (2021): Atomic, Pinwheel, and ClickSwitch (acquired by Q2).
Economics: API calls per direct-deposit switch, plus monthly or annual minimums.
What’s changed: Atomic and Pinwheel are the two that matter now; ClickSwitch and Q2 faded. Knot does a version of this, and InstaSwitch handles the business side.
E. Enriched Transaction Data
Not in the original thread.
What it does: Turns raw, cryptic card and ACH descriptors into clean merchant names, logos, and categories. Powers search, budgeting, and the account feed, and feeds cleaner signals into underwriting and fraud. Frontier five years ago, table stakes now.
Players: Ntropy, Spade, and Heron Data are the specialists; MX and Plaid also enrich as part of their data products.
Economics: Per-transaction or per-API-call, tiered by volume.
F. Fraud, risk, and disputes
Not in the original thread.
What it does: Watches transactions after onboarding for fraud and risk, runs ongoing AML and sanctions monitoring, and handles the dispute and chargeback (Reg E) workflow when something goes wrong. This is the keep-the-lights-on-and-stay-out-of-trouble layer, distinct from the one-time identity check in KYC.
Players: Sardine (which also shows up in KYC, layer K), Unit21, Sift, and Feedzai on real-time monitoring; Hummingbird and ComplyAdvantage for AML, sanctions, and case management; Quavo and Chargebacks911 for disputes, plus the network tools Verifi (Visa) and Ethoca (Mastercard).
Economics: Per-transaction or per-decision screening fees plus a platform fee; disputes usually priced per case or per seat.
G. Global Stablecoin Infrastructure
The one genuinely new category since 2021.
What it does: Stands up a stablecoin payments platform that reaches close to full banking: accounts, cards, and money movement denominated in stablecoins, plus on-ramps and off-ramps, virtual cards, payments, and wallets.
Players: Rain, Bridge & Wirex receive universal acclaim and have the full stack of software tooling, but there’s an ever-growing list of stablecoin-native issuers.
Economics: Usage-based on volume and settlement, and typically cheaper than legacy rails because settlement happens on-chain.
H. Help desk
What it does: Everything on the member-support side, the people and the tooling. Outsourced member services plus dispute and fraud agents that scale on and offshore, and the CRM and IVR (interactive voice response) that route phone and chat, including self-service.
Players (2021): For staffing, Teleperformance, TaskUs, and Sutherland. For CRM and IVR, Talkdesk, Twilio Flex, and Zendesk.
Economics: Fees per ticket or per hour on the staffing side; API calls plus a fee per seat on the CRM and IVR side.
What’s changed: AI is the shift. Decagon handles member conversations and LiveKit powers the voice layer, while Intercom’s Fin now sits in the CRM slot. Human outsourcers still exist, but the center of gravity moved.
I. Issuing processor
What it does: The system of record for every money-movement transaction. It’s your gateway to the payment networks, and it evaluates, authorizes, and settles transactions on behalf of your bank to power card issuance and account maintenance.
Players (2021): Galileo, i2c, Marqeta, and Stripe (Stripe did card issuance rather than full issuing processing).
Economics: Fees per transaction and per API call. (Fee-per-account shows up here sometimes, but it belongs to core banking, layer C.)
What’s changed: Visa DPS took real share. Visa builds and runs it, and it sits under the core of many fintech-driven processors, Chime’s included. Galileo still runs, now owned by SoFi. Marqeta is still here. Lithic grew into a major processor. And as of this week, Column just entered the fray with their own issuer processor.
J. Joined Banking Data
What it does: Connects member accounts and transactions from other sources into your app, links to other financial institutions, and powers account funding and money movement, including ACH transfers. Where outside financial data joins your app.
Players (2021): Finicity, MX, Plaid, and Yodlee.
Economics: API calls plus a fee per connected account, usually with steep monthly or annual minimums and usage-based tiering on top.
What’s changed: Stripe shipped its own Financial Connections product, now a serious option.
K. KYC
What it does: Handles know-your-customer compliance, a regulatory requirement. It verifies and validates applications and runs ongoing AML compliance.
Players (2021): Alloy, Persona, and Socure.
Economics: API calls and/or a fee per account, with the per-check cost scaling with what you run (basic SSN verification versus a selfie or document check).
What’s changed: Onfido and Vouched round out the roster, along with Prove for phone-centric identity, Sardine for fraud plus compliance, and Middesk for business KYB alongside consumer KYC.
The frontier here is zero-knowledge KYC. The user proves a fact (real, unique human; over 18; not sanctioned; in-jurisdiction) without handing you the underlying documents, so you get the compliance signal without the PII liability. Watch zkMe, Self, and ZKPassport. It’s still crypto-native, and today it complements rather than replaces the regulated KYC you’re legally required to keep, but it gets far more interesting as stablecoin rails (layer G) enter the stack.
L. Lending (credit)
Not in the original thread.
What it does: Stand up the credit product itself, co-brand or embedded credit cards, installment and BNPL, and lines of credit, plus the program management and loan servicing to run it. This is the “build me a card or a loan product” bucket, separate from deciding who qualifies (that’s underwriting, layer U).
Players: On credit-card-as-a-service, Imprint runs the whole program end to end, and Synchrony anchors co-brand and private-label at scale. Cardless, Deserve, and Highnote are the other program platforms, and Marqeta and Stripe issue the credit and charge cards. For installment and BNPL, ChargeAfter runs a white-label, multi-lender POS platform banks build on (single-lender or waterfall, BNPL through personal loans), and Wisetack embeds pay-over-time for in-person services. On loan servicing, the modern API-first platforms are LoanPro (it runs Chime, SoFi, and Dave), Peach Finance, and Canopy; Vervent and Nortridge cover outsourced servicing if you don’t want to run it in-house.
Economics: A mix of per-account fees and a share of the credit economics (interchange, interest, or fees), depending on who holds the receivable.
M. Money movement
Not in the original thread.
What it does: Moves money in and out across rails, ACH, RTP, wires, and card push/pull, plus the orchestration and ledgering around it.
Players: Astra for instant card-rail transfers, Dwolla for ACH and RTP, Modern Treasury for payment operations and ledgering, and Orum for instant payouts across rails.
Economics: Per-transaction pricing, usually with a monthly platform minimum.
N. Networks (front-of-card)
What it does: The network that issues your spending card. Partner with Visa and you get a Visa-branded card with Visa as front-of-card, so most transactions (EMV chip, signature, tap-to-pay) run on Visa. The network handles signature and PIN transactions and ATM withdrawals, sets interchange, and pays incentives to drive volume.
Players (2021): Visa and Mastercard.
Economics: Fees per transaction plus basis-point purchase-volume incentives, with interchange front and center.
What’s changed: Discover is now a real option, especially since Capital One acquired it. And we’re starting to get to a place where Block may be a contender in a few years, with the launch and consolidation around Neighborhoods and the Cash App Pay network.
O. Offline Cash Access
What it does: The physical-cash layer, money out and money in. Fee-free ATM networks power withdrawals, which matter most for paycheck-driven neobanks, and cash-deposit networks let members load cash at retail POS.
Players (2021): For withdrawals, Allpoint (roughly 55,000 ATMs at major retailers) and MoneyPass (Fiserv-owned, around 32,000 locations). For deposits, Green Dot and InComm Payments.
Economics: A fee per ATM withdrawal (sometimes a monthly platform fee), and a fee per cash deposit with a revenue share split between the retailer and the fintech.
What’s changed: Little on the ATM side, the same two own the networks. On deposits, Blackhawk Network and PayNearMe now handle cash-in and allocations too.
P. Plastic (card fulfillment)
Not in the original thread.
What it does: Produces the physical plastic or metal that ships with your card program.
Players: CPI Card Group and Idemia manufacture direct. If you’d rather not go direct, Stripe and Lithic will print and fulfill for you.
Economics: A fee per card, plus an upfront inventory buy.
Q. Quick Mobile Check Deposit
Not in the original thread.
What it does: Lets members deposit a paper check by photographing it in-app (mobile remote deposit capture, or mRDC): image capture, OCR of the MICR line and amount, duplicate and fraud checks, and a funds-availability decision.
Players (2021): Mitek owned the underlying OCR and image-capture technology (MiSnap) and licensed it to fintechs and banks, who ran their own distribution and paid Mitek a royalty. Urban FT was the other main turnkey capture option.
Economics: A fee per check deposit.
What’s changed: The layer splits into who builds the check-reading engine and who just wires it in. Mitek still leads capture and OCR (still independent and public, NASDAQ: MITK) and now sells its Mobile Deposit product direct rather than only licensing the engine; the other pure-play recognition APIs are Veryfi and Parascript.
Most digital-banking platforms don’t build their own OCR: Alkami, Q2, and Nymbus integrate background vendors like Vertifi for capture and Advanced Fraud Solutions for check-fraud screening, while legacy cores Fiserv and Finastra ship native RDC with their own imaging and CAR/LAR (courtesy/legal amount recognition) engines.
The bigger shift is risk and funding: Ingo Money (now Ingo Payments) added real-time deposit underwriting and guaranteed funding, so a fintech can accept checks and release money instantly without eating the fraud. UrbanFT is still around and expanded through its CFC Technology acquisition, and BaaS cores like Synctera and Green Dot now expose mRDC as a native endpoint.
R. Rewards Infrastructure
Not in the original thread.
What it does: A merchant-funded (or self-funded) rewards layer on top of your card program. It’s one of the most common neobank hooks, so it earns its own line.
Economics: A revenue share on the merchant-funded rewards.
S. Subscriptions & Bill Pay Management
Not in the original thread.
What it does: Pay billers, manage and track recurring subscriptions, and offer embedded subscription products like phone plans. The recurring-payments-and-services layer.
Players: Fiserv (CheckFree), Mastercard Bill Pay Exchange, and Payrailz run the bill-pay networks; Knot manages subscriptions and the card on file across merchants; Gigs powers embedded phone plans (the OS behind mobile plans at Cash App, Nubank, and Klarna).
Economics: A fee per payment, sometimes a monthly platform minimum; embedded products like phone plans add a revenue share or margin on the plan.
T. Trading Infrastructure
Not in the original thread.
What it does: Lets members invest on platform. On the traditional side, equities and ETFs with optional robo-advisory; on the crypto side, buy, sell, and hold, with a licensed partner handling execution, custody, and settlement (distinct from the stablecoin rails in layer G, which are about payments, not investing).
Players: On brokerage, Apex, Alpaca, and DriveWealth. On crypto, Zero Hash and Paxos for brokerage-and-custody, Fireblocks on custody infrastructure, and Bakkt as a turnkey alternative.
Economics: API calls, AUM fees, and per-transaction fees on brokerage; a spread or per-trade fee plus custody fees on crypto.
U. Underwriting
Not in the original thread.
What it does: Decide who gets credit and on what terms. This layer pulls the risk signals, scores the applicant, and sets limits and pricing: the decisioning engine plus the alternative and cash-flow data that feed it. It’s the risk brain behind the credit product above.
Players: On decisioning, Zest AI, Taktile, and Oscilar lead the modern set, with Scienaptic and GDS Link as established credit-decisioning engines and Provenir on the enterprise end. For the data that feeds the models, Nova Credit brings cross-border and cash-flow credit data, Prism Data scores deposit-account cash flow (its CashScore), Plaid adds cash-flow and income signals (LendScore), Argyle verifies income and employment from payroll, and Ocrolus handles income and document verification.
Economics: Usually per-application or per-decision pricing plus a platform fee, and the data feeds carry their own per-pull cost.
V. Variable PIN Networks
What it does: The Durbin-required secondary networks (the variant to your front-of-card) that carry PIN transactions. Durbin mandates at least two unaffiliated networks on a debit card, so you run your primary front-of-card through Visa or Mastercard and a variant network for PIN. They use fees and incentives to drive volume.
Players (2021): Visa’s PIN network is Interlink, Mastercard’s is Maestro, Pulse is Discover’s, and Accel and STAR belong to Fiserv.
Economics: A fee per transaction, interchange revenue, and per-transaction incentives.
What’s changed: Not much, and that is the point. Becoming a variable PIN network is brutally hard: it takes scale, direct bank and processor connectivity, years of certification, and Durbin’s two-unaffiliated-networks rule keeps the incumbents entrenched. The more likely challenge comes from a different rail entirely (pay-by-bank and account-to-account, or stablecoin settlement) rather than a new PIN network. Nobody has displaced these five yet, but that is the seam to watch.
W. Wage Access
Not in the original thread.
What it does: Gives members access to wages early, or as they’re earned, instead of on the paycheck cycle.
Players: Clair, Rain App (the earned-wage company, a different Rain from the stablecoin one in layer G), and DailyPay.
Economics: A fee per push of the money.
X. X-border Transfers
Not in the original thread.
What it does: Send and receive money internationally, and hold or convert multiple currencies. Provides the FX conversion and the correspondent and local rails behind international transfers and multi-currency accounts.
Players: Wise Platform, Nium, and Currencycloud (now Visa) are the infrastructure names; Airwallex for multi-currency accounts and global payouts.
Economics: An FX spread plus a per-transfer fee.
Y. Yield (deposit sweep and FDIC insurance)
Not in the original thread.
What it does: Spreads a member’s balance across a network of banks so deposits stay FDIC-insured well past the $250k single-bank cap, and generates yield (and usually a revenue share) on those balances, with no extra accounts for the member to open. This is how neobanks advertise “millions in coverage” and monetize idle cash.
Players: IntraFi (ICS and CDARS, the incumbent) and R&T Deposit Solutions (Demand Deposit Marketplace, purpose-built for fintechs) lead the established side; ModernFi is the modern, API-first entrant running institution-owned deposit networks; StoneCastle rounds it out.
Economics: A spread on the swept balances (you keep part of the yield), sometimes a platform or per-account fee.
Z. Zoom-In Analytics
Not in the original thread.
What it does: Gives you a full view of what flows through the stack and the economics underneath it.
Players: Margin is the purpose-built option; some teams run Hex.
Economics: Platform or seat-based, sometimes usage-based.
The stack is only half of it
The vendor map never captures the hard part. Assembling these vendors is necessary, but it is not the business. Positioning, distribution, and onboarding turn a stack into a neobank people actually use. Eric’s post is the cleanest single map of all the pieces, so I’ll point to it rather than reinvent it.
Strong opinion, weakly held, as always. If I missed a letter or mispriced a deal, tell me and I’ll fold it into the next remaster.
Disclosure: I advise or invest in a handful of the companies mentioned here, including Margin, Percents, and InstaSwitch.