Credit intermediation and related activities

A core replacement is the single riskiest project a bank undertakes — ledger, deposits, payments and regulatory reporting move at once — so cores are replaced roughly once a generation. The incumbents' position rests on that, not on product quality. The modern challengers (Mambu, Thought Machine, 10x) have each raised hundreds of millions and are still fighting for reference customers a decade in.

The asset in this category is a shared data network, not the software, and the people who own the largest networks are buying the vendors. FICO's FY2025 10-K says its fraud models are trained on transaction data from more than 10,000 institutions in the FICO Falcon Intelligence Network [A]. Riskified's FY2025 20-F reports $155.1B of ecommerce orders reviewed, about 2.5% of global ecommerce by its own calculation [A]. That is what an entrant cannot copy. A new model on day one has seen no fraud, and the buyer is paying for what the model has seen. The exits tell the same story. Visa bought Featurespace in December 2024 for $946M (Visa FY2025 10-K) [A]. Mastercard's FY2025 10-K describes Recorded Future as part of its fraud and cyber offering [A]. Entrust bought Onfido in April 2024, and onfido.com now redirects to entrust.com [B]. Permira took majority control of BioCatch at a $1.3B valuation in September 2024 [B]. Socure bought Effectiv for about $136M in late 2024 [B]. In each case the buyer already owned traffic, a payment network or a bureau. The listed pure-plays are not growing fast. NICE's Financial Crime and Compliance segment, which is mostly Actimize, earned $485.4M in 2025, up 7.0% [A]. FICO's whole Software segment earned $822.3M, up 3%, with software ARR up 4% [A]. FICO does not break out fraud revenue. Riskified earned $344.6M, up 5.2%, with a net loss of $27.6M [A]. Mitek earned $179.7M, up 4% [A]. Meanwhile the bureaus get bigger. RELX's Risk segment, which includes LexisNexis Risk Solutions, earned £3,485M in 2025 with 8% underlying growth. RELX names "digital Fraud & Identity" as one of the drivers [A]. How this differs from the neighbouring records. The regtech and compliance automation record (5223) covers AML transaction monitoring, sanctions and KYC case management, a budget the bank spends because a regulator requires it. This record covers the loss budget: card and payment fraud, account takeover, application fraud, identity proofing at onboarding, and chargebacks. The two overlap at NICE, Feedzai and Sardine, which sell both. The core banking record (5221) is the ledger these tools plug into. The embedded payments record (522390) is the rail whose risk this software scores. The wider sector is mapped on the fintech topic page. The 5221 anchor (depository credit intermediation) is the anchor buyer. The same software is sold to payments companies and processors (522320), consumer lenders (5222), e-commerce retailers and marketplaces (4541, 459), and gig and crypto platforms. Incumbent vulnerability decides it. The large data networks are owned by companies whose main business is elsewhere, and they treat fraud scoring as a feature of the network. That leaves an independent to sell either to a long tail of fintechs that have little data of their own, or to be acquired by one of those companies.

Two systems of record share the bank and credit-union desk, and the decision engine underneath belongs to the score and bureau companies. nCino earned $594.8M in the year to 31 January 2026, up 10%, and reported its first GAAP profit ($5.2M) after fourteen years [A]. It ended the year with over 2,700 customers, about 1,500 of them depository institutions, and 114 paying more than $1M a year [A]. Its strength is commercial and small-business credit at banks; mortgage is a separate $79.6M US line [A]. MeridianLink is the consumer and credit-union side: it earned $316.3M in 2024, $249.3M of it from lending software, across about 2,000 institutions, and charges per application for personal, auto and card loans whether the application is approved or not [A]. Centerbridge took it private at $20.00 a share, about $2.39B of total funding including debt, and the NYSE delisting was filed on 24 October 2025 [A]. MeridianLink's own 10-K says its main competition in non-mortgage consumer origination is the bank's core provider (Fiserv, FIS, Jack Henry, Temenos), then nCino and Q2, then CU Direct's Origence and Sync1 [A]. Q2's 10-K names Abrigo, Baker Hill, nCino, Finastra, Brilliance and Temenos against its lending products, and Q2 has only about 100 digital-lending customer accounts against $794.8M of total revenue [A]. The decision layer is held by FICO, whose 10-K names Experian, Equifax, Moody's, MeridianLink and CGI as its origination competitors [A], with Provenir, Taktile and Zest AI as the funded independents. The AI-underwriting entrant already exists and is well funded. Zest AI took $200M from Insight Partners in December 2024 and says nearly 300 lenders use it [B/C]. Upstart sells the same idea to banks and credit unions as a lending platform: $1.04B revenue in 2025, $11.0B of loans, 91% fully automated, over 100 lending partners [A]. Consolidation is the exit. Moody's absorbed Numerated (numerated.com now redirects to Moody's loan-origination page), Alkami paid about $375M for MANTL's account opening in March 2025 [A], nCino bought DocFox, FullCircl and Sandbox Banking [A], and Carlyle, Accel-KKR, Flexpoint Ford and Centerbridge own Abrigo, Baker Hill and MeridianLink. How this differs from the neighbouring records. The mortgage record (5222) is ICE's Encompass world: one-to-four-family loans, servicing and secondary marketing. This record covers everything else a lender books: personal, auto and card loans and small-business credit, plus deposit account opening and the credit decision engine. The core banking record (5221) is the ledger the LOS posts to, and the core vendor is the LOS's first competitor. The financial data and credit analytics record (5239) holds the bureaus and scores that feed the decision; the fraud and identity record (5221) covers application fraud and KYC, which origination vendors now bundle. The lenders themselves, including the 60 on CNBC's 2026 list, are on the fintech topic page. In Canada, Covaleyo (Valeyo and Covarity, merged April 2025) says ProLender is used by 30% of Canadian credit unions [C]; Lendesk's Finmo is mortgage-broker software and belongs with the mortgage record. Incumbent vulnerability decides it. An entrant has to beat the core provider's bundled LOS, then nCino or MeridianLink, then a model-risk review of any new credit model, and the two independent AI-underwriting vendors with real funding (Zest AI, Taktile) are already selling into that gap.

One vendor sits in the path of roughly nine in ten US mortgages and owns both origination and servicing after the Black Knight merger. The most credible challenger has been losing share for three years. Regulatory examination of the origination workflow adds a compliance floor on top. The interesting question in this market is antitrust, not entry.

Well-funded on both sides: entrenched incumbents inside bank procurement and a dense field of venture-backed challengers already competing for the same AML and sanctions-screening budgets. Note the Canadian precedent — Verafin (Killick Capital-backed, and present in this repo's angel dataset) proves the market rewards a winner, and also that the winner has already emerged.

What this category is. The biller's side of a household or business bill: presenting the bill (portal, email, text, PDF), taking the payment by card, ACH, wallet, phone, text, IVR or cash at a retailer, running autopay and payment plans, and posting the money back into the biller's own billing system. The 522320 code (financial transactions processing) is navigation only — the buyers are utilities, telecoms, insurers, hospitals, property managers, lenders, universities and every level of government, so this is cross-industry software rather than a vertical. Paymentus describes exactly that spread: billers 'across a variety of industry verticals, including utilities, financial services, insurance, government, telecommunications, real estate management, education, consumer finance, healthcare, business to business (B2B) and small business' [A, 10-K]. How it differs from the neighbouring records. Utility CIS and metering (2211) is the system of record that issues the bill; this record is the layer that collects it and plugs into that CIS. Embedded payments for vertical software (522390) is the infrastructure a software vendor rents to take payments inside its own product; EBPP vendors sell directly to the biller. Collections (561440) starts where this one stops: the bill is already late. Who holds it. Two public companies publish the numbers. Paymentus (NYSE: PAY; Charlotte) reported FY2025 revenue of $1,196.5M, up from $871.7M, on 724.0M transactions, up 21.3%, with net income of $66.9M and contribution profit of $386.3M; 98.4% of revenue came from the United States [A]. ACI Worldwide's Biller segment — Speedpay, which ACI bought from Western Union for $750M in cash in May 2019 [A, 8-K] — earned $817.7M of revenue in FY2025, up from $726.5M, at $140.7M of segment adjusted EBITDA [A]. Fiserv sells biller-direct, print and in-person cash payment (CheckFreePay, 'over 30,000 retail locations') to billers [C, own site] and CheckFree RXP to banks, whose customers pay billers from online banking [A, 10-K]; FIS sells BillerIQ, which its 2024 product sheet describes as 'a white labeled SaaS biller direct solution' [A, own product sheet]. Neither breaks out a revenue figure for the business. The field has just consolidated again. ACI's own 10-K lists its bill-payment competitors as 'Alacriti, FIS, Fiserv, InvoiceCloud, Kubra, One Inc., Paymentus, PayNearMe, Repay, as well as smaller vertical-specific providers' [A]. Since then REPAY (Nasdaq: RPAY) closed its purchase of KUBRA from Hearst on 1 June 2026 for about $372M in cash, funded partly by a new term loan [A, 8-K]; KUBRA added about $20.8M of revenue in its first month inside REPAY [A, 10-Q], and still tells visitors it reaches 'over 40% of households in the U.S. and Canada' [C]. InvoiceCloud went private inside EngageSmart when Vista Equity Partners bought EngageSmart for about $4.0B in January 2024 [B, Vista release]; the old engagesmart.com address now redirects to invoicecloud.net [A, observed]. On the B2B side, Billtrust (EQT, about $1.7B, December 2022 [B]) and Bottomline (Thoma Bravo, about $2.6B, May 2022 [A, the buyer's release]) were both taken private. One Inc, the insurance-premium specialist, is held in equal stakes by Great Hill Partners and Nordic Capital since March 2024 [A, own release]. The consumer-side consolidator is under an FTC order. doxo, a Seattle bill-pay app that says it reaches 'over 120,000 billers' and 'over 10 million people' [C], agreed in August 2026 to pay $2.1M to settle FTC allegations that it used 'misleading search ads to impersonate consumers' billers' and added undisclosed 'delivery fees' [A, FTC]. That is a warning about the one model that does not need the biller's consent. Pricing is the convenience fee, and scale wins it. Paymentus says it is paid transaction fees 'directly from billers, financial institutions, partners or, in some cases, from consumers as a convenience fee', and that its gross margin falls as it adds 'large, high-volume enterprise billers with lower margins' because card interchange is its main cost [A]. ACI's Biller revenue rose $91.2M in FY2025 while cash operating expense rose $81.7M, 'primarily for payment card interchange and other processing fees' [A]. The economics are a thin spread on interchange that rewards volume, and FIS advertises a 'dynamic and compliant surcharge program' as a feature [A, product sheet]. Incumbent vulnerability decides it. The incumbents are not weak: the largest pure-play grew revenue 37% in a year and is profitable, the next one is a $818M segment of a public payments company, and Fiserv and FIS sell the same product to billers alongside their bank businesses. Each acquisition in the last four years has folded a challenger into a larger book. A new entrant would sell to a regulated biller through an RFP, integrate to its CIS or policy system, and then compete on a fee where the incumbent's interchange volume sets the price.

The thesis behind this category is true, and the proof is on EDGAR — which is exactly why it is a bad place for a new entrant. Toast's FY2025 10-K (year ended 2025-12-31) reports financial technology solutions revenue of $5,037M against subscription services revenue of $936M, on $195.1B of gross payment volume [A]. Payments is 5.4x the software line by revenue and still bigger by gross profit — $1,146M from fintech against $672M from subscriptions, computed from the 10-K's own cost-of-revenue table — even though the fintech margin is about 23% against 72% for software. AppFolio shows the same shape in property management: Value Added Services (electronic payments, tenant screening, risk mitigation) $721.5M against subscription services $211.5M in FY2025 [A]. The atlas's own vertical records reach the same answer from the other side: payments and insurance attach decides self-storage (Storable), marina (DockMaster Payments, Storable Marine), salon (Fresha at zero subscription), medspa (Boulevard, Zenoti) and field-service (Jobber, Housecall Pro) — all of them treat the licence as the cheap part. But the value flows to the software company that owns the merchant, not to the rail it rents. The infrastructure layer underneath is held by two giants. Stripe says businesses on it generated $1.9 trillion of volume in 2025, up 34%, and priced a tender offer at $159B (newsroom, 2026-02-24) [B]. Adyen reported FY2025 net revenue of EUR 2,364.2M on EUR 1,394.3B processed, with issuing volumes up 8x as platforms embedded cards (H2 2025 results, 2026-02-12) [A]. Below them the processors bought their way in: FIS bought Payrix in 2022, and Global Payments completed its purchase of Worldpay on 2026-01-12, serving more than 6 million merchant locations on $3.7 trillion of volume with an 'Integrated & Platforms' channel [A, 8-K]. The venture-funded challengers are small next to that and are selling the same thing: Finix (over $208M raised, Series C led by Acrew), Rainforest ($57.5M, Series B led by Matrix and Infinity Ventures, 2025-09-08), Tilled (nearly $40M), Infinicept ($23M growth round, 2022), Moov ($77.5M). Each one positions itself as cheaper or more vertical-friendly than Stripe — Rainforest's model is a consumption-based cut of each transaction with no platform fee [B] — which means the wedge is price, paid for out of a thin interchange margin. The adjacent layers (BaaS: Unit $169.6M, Treasury Prime ~$73M; embedded lending: Parafin $194M; embedded payroll: Check, $75M Series C led by Stripe) last raised in 2021-24 and none has announced a larger round since in the coverage opened here. An entrant needs a sponsor bank, card-network registration, underwriting and risk reserves before earning its first basis point, then has to win platforms that can renegotiate or switch processors once their volume is large. Entry cost and regulatory drag decide it. The money in this theme is earned by building the vertical software and attaching one of these rails — the right move for a Research Upon vertical, not a market to enter as the rail.