Industries / Finance and business services / Finance and banking / Credit unions
Core systems
The core is the system of record that holds every member's shares and loans and posts each transaction; nearly every credit union now rents it, and the computers it runs on, from an outside firm.
5
Milestones achieved, from 1998 to 2025.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1998
Vendor oversight granted, then lost
In March 1998 Congress passed the Examination Parity and Year 2000 Readiness for Financial Institutions Act, which let the National Credit Union Administration (NCUA), the US credit union regulator, examine the firms that ran credit unions' computer systems, so it could check their readiness for the year 2000 date change. The power expired on 31 December 2001 and has not been restored. NCUA says it reviewed vendors only with their consent until 2013, that several vendors declined or refused to fix network security problems it found, and that in 2016 the banking agencies stopped letting it join their joint examinations of technology providers. Bank regulators have had this power over bank service firms since 1962.
Lessona power granted for one emergency lapsed with it, and a regulator that depends on vendors' consent cannot make them fix what it finds.
Credit
US Congress; NCUA, which documents the gap; GAO, which has recommended restoring the power in three reports since 1999
2007
Hand-kept ledgers disappear
An economist studying NCUA data for 1992 to 2005 found that manual record keeping "has essentially disappeared", that systems a credit union built for itself had become very rare, and that nearly every credit union outsourced its data processing by 2005. Vendor systems cost more than home-built ones but offered many more products, and credit unions on vendor systems were much more likely to survive than those on manual or in-house systems, a gap that was larger in 2005 than in the early 1990s.
Credit
Victor Stango, Dartmouth College, for the Filene Research Institute
2023
A forced switch goes wrong
UNI Financial Cooperation, a New Brunswick credit union with about 170,000 members, moved to a new banking platform in early July 2023. Members reported pay deposited late, payments taken on loans already paid off and transfers that failed, with waits of two to three hours in branches. Its chief executive, Camille Thériault, said the platform "wasn't ready", that UNI had to move because Desjardins would not continue to supply its technology, and that UNI had lost 2,600 members since July.
Lessona deadline set by a departing supplier pushed a platform live before it was ready.
Credit
UNI Financial Cooperation and Pinnacle InfoSys, the platform's developer; reported by Global News
2023
Ransomware at a shared host
On 26 November 2023 criminals used ransomware, software that locks a victim's files until a ransom is paid, against Ongoing Operations, a firm owned by Trellance that hosts computer systems for credit unions. NCUA said about 60 credit unions had outages; FedComp, a core vendor for small credit unions that Trellance also owns, reported a nationwide outage. On 12 December NCUA said the affected credit unions were "fully operational". A security researcher linked the attack to a widely exploited flaw in Citrix network devices, which the companies did not confirm.
Lessonone hosting firm, two steps removed from members, took dozens of credit unions offline at once.
Credit
NCUA, which counted the credit unions affected; Ongoing Operations
2025
Clients drift from the leader
Market share figures compiled by Callahan & Associates for mid-2025 show Fiserv, across its core systems, serving 1,155 credit unions, 25.9% of them, down 1.18 percentage points in a year. Cores built on Jack Henry's Symitar software served 699 credit unions through Jack Henry and two firms owned by credit unions, Member Driven Technologies and Synergent. Corelation's KeyStone core, the fastest gainer, served 211 credit unions, 4.7%, up 0.67 points. Of 27 core vendors with at least $400 million in client assets, 7 gained clients and 16 lost them.
Credit
Callahan & Associates, from data the vendors submitted
Watch for
One cloud core, one region
League Data, a co-operative that supplies technology to the credit unions of Atlantic Canada, is moving them all from a legacy core that reached the end of its life in June 2025 to Mambu's core, which runs on Amazon Web Services. Mambu says 33 conversions are complete with four left, and that the whole Atlantic system will move in just over a year.
Achieved whenall the region's credit unions run on the new core and an outside party, such as a provincial regulator or deposit insurer, reports the conversion complete and the system's outage record.
Claimed by
Mambu and League Data, 2026 (League Data's case study is dated May 2026)
Open
01
Map the vendors behind vendors
An open table, one row per US credit union, of the outside firms it depends on in four layers: its core, its online and mobile banking, its card processing and the firm that hosts its systems. Each cell carries the public evidence and the date it was seen: a vendor's press release or client list, the web address of the login page, the developer named in the app store, a job posting or an annual report. From the table, compute how many credit unions and members would lose service if each firm, including a firm one step removed, went down.
The briefCredit · unclaimed02
A rehearsal kit for conversions
An open, made-up credit union of about 50,000 invented members, holding the awkward cases a conversion must carry over: joint owners, accounts for minors, share certificates (term deposits) with dividends accrued but not yet paid, loans with escrow and partial payments, holds, dormant accounts and closed accounts with history. Export it in two different record layouts, and write a checker that compares source and converted data: counts, balances, accrued dividends and interest, payment schedules and owner links, listing every mismatch. A language model can draft the field mapping between the two layouts, and the checker scores the draft.
The briefCredit · unclaimed
Open 01 · the brief
Map the vendors behind vendors
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open table, one row per US credit union, of the outside firms it depends on in four layers: its core, its online and mobile banking, its card processing and the firm that hosts its systems. Each cell carries the public evidence and the date it was seen: a vendor's press release or client list, the web address of the login page, the developer named in the app store, a job posting or an annual report. From the table, compute how many credit unions and members would lose service if each firm, including a firm one step removed, went down.
Why it is still open
NCUA collects vendor information from credit unions, but its public quarterly data files carry one technology field, "Credit Union Data Processing System Type", and it reads 0 for all 4,299 credit unions in the June 2026 file, by a count made for this entry. Callahan & Associates publishes core vendors' market shares from data the vendors submit, as totals per vendor, without the hosting firms beneath them. Lists per credit union are sold, not published: CUCollaborate sells a spreadsheet of the core processor, online and mobile banking and loan systems of every US credit union, which does not mention card processors or hosting firms, and a 2024 Kansas City Fed study of the core market had only survey and Callahan totals, from 2020, for credit unions. The 2023 outage ran through such a link: Ongoing Operations hosted systems for credit unions, and its owner also owns the core vendor FedComp.
ncua.govcreditunions.comcucollaborate.comkansascityfed.orgcybersecuritydive.com
Built with what exists
NCUA's quarterly call report files for the list of credit unions with their assets and members; a browser agent built on the open-source Browser Use library to visit each credit union's website and record the login page's address and the app's developer (see the Note "Deleting 80% to start over"); a language model to read press releases and job postings into fields, each with a quoted excerpt (see the Note "The jobs agents repeat"); SQLite or a spreadsheet.
First step
Take the 100 smallest credit unions in one state, fill the four layers by hand from public pages, and time each lookup to learn which kinds of evidence work.
Done looks like
All federally insured credit unions listed, with evidence-backed entries in at least two of the four layers for 80% of them; a second person checks 200 random cells against their evidence and the error rate is published; a ranked list of the 20 firms whose failure would cut off the most members; the table released under an open license and refreshed each quarter.
Open 02 · the brief
A rehearsal kit for conversions
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open, made-up credit union of about 50,000 invented members, holding the awkward cases a conversion must carry over: joint owners, accounts for minors, share certificates (term deposits) with dividends accrued but not yet paid, loans with escrow and partial payments, holds, dormant accounts and closed accounts with history. Export it in two different record layouts, and write a checker that compares source and converted data: counts, balances, accrued dividends and interest, payment schedules and owner links, listing every mismatch. A language model can draft the field mapping between the two layouts, and the checker scores the draft.
Why it is still open
Failed conversions keep hurting members. After VyStar Credit Union's May 2022 cutover some features stayed down for more than six months, and the CFPB called the problems foreseeable; UNI's 2023 switch took payments on loans already paid off. Test-data tools that cover core migrations, such as GenRocket, are sold to individual firms, and a search found no open test credit union or open conversion checker.
Built with what exists
Python with the Faker library for invented names and addresses; Apache Fineract, an open-source core banking system under the Apache 2.0 license, as one target system; a simple flat-file layout modeled on a legacy export as the other; DuckDB or pandas for the comparison; a model such as Claude or Gemini to draft mappings.
First step
List 30 conversion faults from public accounts of failed conversions, then generate 1,000 invented members across the ten commonest account types.
Done looks like
The made-up credit union and the checker published under an open license, with 30 seeded conversion faults (for example a dropped joint owner, or dividends accrued for the wrong number of days); the checker finds all 30; a model-drafted mapping scored by the mismatches it leaves before and after one round of human correction.