The 2017 payday rule built its affordability test on credit data systems the Bureau had yet to register

The CFPB’s October 5, 2017 payday rule had lenders furnish and check loan data through Bureau-registered systems. In our reading, that shared network decided how the rule could run, and lenders never reached the date when they had to use it.

Looking Back revisits past weeks one at a time, up to the most recent, and asks what it meant for the banks, health systems and public agencies we serve. The series was written in 2026, with the benefit of hindsight, and each article says which week it looks back on.

Two-minute summaryThe dek and the first sentence of each section, quoted from the article

The CFPB’s October 5, 2017 payday rule had lenders furnish and check loan data through Bureau-registered systems. In our reading, that shared network decided how the rule could run, and lenders never reached the date when they had to use it.

  1. The week On Thursday, October 5, 2017, the Consumer Financial Protection Bureau issued its final rule on payday, vehicle title and certain high-cost installment loans.
  2. The furnishing duty: what each lender had to send, and when Section 1041.10 of the rule text set the furnishing duty.
  3. The registry: who could become a registered information system, and what it had to prove Section 1041.11 set out who could become a registered information system.
  4. The decision: how a lender used reports, its own records and the fallback The network existed to inform two decisions.
  5. The records: what a lender had to keep, and what the payment rules needed instead Section 1041.12 required a written compliance program and record retention.
  6. How the design reached different lenders and data firms The rule’s coverage was broad, and its data design reached each kind of firm differently.
  7. Objections, in order The argument that the data network was the rule’s decisive and exposed part faces objections.
  8. What a compliance or data leader could check The payday rule’s data design is gone, but its lessons apply to any rule that depends on data from outside the firm.
  9. With hindsight The rule was published in the Federal Register on November 17, 2017, and took effect on January 16, 2018, according to the Bureau’s rule page.

The week

On Thursday, October 5, 2017, the Consumer Financial Protection Bureau issued its final rule on payday, vehicle title and certain high-cost installment loans. It posted the rule text, a press release and a fact sheet the same day. The rule would create a new part of the Bureau’s regulations, 12 CFR part 1041.

The rule had a long run-up. The Bureau issued its proposal on June 2, 2016, and published it in the Federal Register on July 22, 2016. The press release said the final rule followed five years of research and outreach. It said the Bureau had reviewed more than one million comments on the proposal.

The rule had two main parts. The first covered short-term credit and longer-term credit with balloon payments, including payday and vehicle title products. For those, the rule identified it as an unfair and abusive practice to lend without reasonably determining that the consumer could repay. The second part reached a wider set of credit products. It limited repeated attempts to withdraw payment from a consumer’s account.

The affordability test had a name. Under the full-payment test, a lender had to determine that the borrower could make the loan payments and still meet basic living expenses and major financial obligations. The fact sheet said this applied during the loan and for 30 days after the highest payment. Lenders had to verify income and major financial obligations and estimate basic living expenses.

There was an alternative for some short-term credit. Under the principal-payoff option, a consumer could borrow up to $500 without the full-payment test. Up to two extensions were allowed, but only if the borrower paid off at least one-third of the original principal each time. The option was not available where a vehicle title secured the debt.

The rule also exempted some lending. A lender that made 2,500 or fewer covered short-term or balloon-payment loans a year, and drew no more than 10 percent of its revenue from them, did not need either test. Loans that generally met the National Credit Union Administration’s payday alternative loan parameters were excluded. So were certain no-cost advances and employer wage-advance programs.

One requirement concerned data. The fact sheet said lenders had to use credit reporting systems registered by the Bureau to report and obtain information on credit extended under the full-payment test or the principal-payoff option. Companies could apply to be designated as registered information systems. As consumer reporting companies, the fact sheet said, they would be subject to applicable federal laws.

The payment protections were separate. For short-term and balloon-payment credit, and for credit above a 36 percent annual percentage rate with account access, the lender had to give written notice before the first withdrawal attempt. After two straight failed attempts, it could not try again without a new and specific authorization from the borrower.

The timing was staggered. The press release said the rule would take effect 21 months after publication in the Federal Register. It said the provisions allowing registration of information systems would take effect earlier. The rule text set that earlier date at 60 days after publication.

The press release said the rule covered all lenders who regularly extend credit, including banks, credit unions, nonbanks and their service providers. It applied whether a lender operated online or from storefronts, and whatever state licenses it held. The protections were in addition to existing state or tribal requirements. The rule text said the rule operated as a floor, leaving state and local jurisdictions free to adopt further protections.

The October 2017 payday rule is remembered for its ability-to-repay test, and for being taken apart. Read closely, the rule did something less familiar. It made a lending decision depend on a shared data network that did not yet exist, and it set out who could run that network and what they had to prove.

The registered information systems were the rule’s answer to a specific problem. A lender can see its own loans. It cannot see a borrower’s loans at the storefront down the street or the website across the state line. The full-payment test and the principal-payoff option both depended on knowing a borrower’s recent borrowing. The Bureau chose to build that view through registered consumer reporting systems that every covered lender would feed.

Our argument is that the rule’s data design was the part that decided how it would work, and the part exposed to change. The underwriting tests could not run as written without the network. The network could not exist until companies applied, were approved and had been registered long enough to count. We take the design in four parts:

  1. The furnishing duty: what each lender had to send, and when.
  2. The registry: who could become a registered information system, and what it had to prove.
  3. The decision: how a lender used reports, its own records and the fallback when no system was ready.
  4. The records: what a lender had to keep to show its decisions, and what the payment rules needed instead.

The first two parts describe the network. The third describes its use in each loan. The fourth describes the evidence trail. Together they cover the data path from loan to examiner. After the four parts we look at how the design reached different lenders and data firms, at objections, and at what a compliance or data leader could check.

A note on sources and our reading. We describe the rule from the Bureau’s October 2017 rule text, press release and fact sheet. We describe later events from Bureau statements and the Federal Register. Statements about what lenders owed are our reading of those documents. Organizations should check with counsel. The underwriting and data provisions no longer exist, so we discuss them for what they show about rules that depend on shared data infrastructure.

The furnishing duty: what each lender had to send, and when

Section 1041.10 of the rule text set the furnishing duty. For each covered short-term loan and each covered longer-term balloon-payment loan, the lender had to furnish specified information to each qualifying information system. The duty was not optional and not limited to one system. A lender would furnish to every system that, on the day the loan was made, had been registered or provisionally registered for at least 180 days.

Three moments in the life of a loan

The rule tied furnishing to three moments. At consummation, the lender had to send identifying information about the loan and the consumer, the loan type, the date, and the payment schedule. It had to do so no later than the date the loan was made, or as close to that date as feasible. During the loan, it had to send any update within a reasonable period of the event that made earlier information out of date.

When the loan ended, the lender had to send the date it ceased to be outstanding. It also had to say whether all amounts owed were paid in full. The rule listed what that meant: the amount financed, charges included in the cost of credit and charges excluded from it. In our reading, a loan that ended in a charge-off or settlement would be reported differently from one paid in full.

The fields that mattered

The consummation fields were specific. They included information to identify the loan uniquely and to identify the consumers responsible. They included whether the loan was short-term or balloon-payment, and whether it was made under the full-payment test or the principal-payoff option. For closed-end credit, the lender sent each payment due date and amount. For open-end credit, it sent the credit limit and the minimum payment due on each date.

Those fields describe a loan’s shape, not its price. The rule did not require the lender to send the borrower’s income or living expenses to the system. The systems would hold borrowing history. The affordability analysis would stay inside each lender. That split is a design choice with consequences. Every lender would see the same borrowing record. Each would reach its own judgment on what a borrower could afford.

Identity as the hard field

The weakest point in any shared credit record is matching. The rule asked lenders to furnish information sufficient for the system to identify the specific consumer. It did not prescribe a matching key. In our reading, that left the systems and lenders to agree how a borrower at one lender would be recognized at another. A borrower’s address, phone number or name spelling can differ between two lenders’ records.

A mismatch would cut both ways. A missed match would understate a borrower’s outstanding loans, so a lender might approve a loan the rule meant to block. A false match would overstate them, so a lender might refuse a loan the rule allowed. The rule’s outcome, for each borrower, depended on the quality of a join across companies. That is a data quality problem before it is a legal one.

Format and timing as operating work

The rule said a lender had to furnish information in a format acceptable to each system. That meant one integration per system, not one integration per rule. A lender with several systems in scope would maintain several feeds. Each feed would need monitoring, error handling and reconciliation. A rejected record is a furnishing failure even if the lender sent it.

The timing standard was also demanding. Information at consummation was due by the date the loan was made, or as close as feasible. For a storefront lender, that would mean real-time or same-day transmission from the point of sale. For an online lender, it would mean an application interface called during funding. In our reading, the rule wrote a near-real-time data duty into consumer lending for this product.

Who would own the feed

Inside a lender, the furnishing duty would cross functions. Compliance would own the requirement. Technology would own the integration. Operations would own corrections when a branch entered a wrong date or amount. Finance would care because a loan paid by a third party, or reversed, would change what was reported. A firm that assigned the duty only to compliance would find the data owned elsewhere.

Service providers mattered too. The press release said the rule applied to lenders’ service providers. A lender that used a loan management platform would, in our reading, depend on that platform to produce correct furnishing files. Contracts with such vendors would need to name the furnishing obligation, the format, the timing and who fixes errors. Those are ordinary vendor terms, applied to a new data flow.

The registry: who could become a registered information system, and what it had to prove

Section 1041.11 set out who could become a registered information system. The Bureau would not run the network itself. Private companies would apply. The rule listed eight conditions, and the Bureau would decide whether each applicant met them. The list reads like a specification for critical consumer data infrastructure.

Two technical conditions

The first condition was receiving capability. The entity had to be able to receive furnished information immediately. It had to use reasonable data standards that allowed timely and accurate transmission without unreasonable costs or burdens on lenders. The second was reporting capability. The entity had to generate a consumer report containing the furnished information substantially simultaneously with receiving it.

Those two conditions together describe a real-time system. A lender that furnished a loan at noon should see it reflected in a report requested a minute later at another lender. The rule wrote latency into eligibility.

Two compliance conditions

The third condition was performance. The entity had to perform in a way that facilitated compliance with the rule and furthered its purposes. The fourth was a compliance program. The entity had to maintain a program reasonably designed to ensure compliance with all applicable federal consumer financial laws. The rule specified written policies and procedures, comprehensive training, and monitoring to detect and promptly correct weaknesses.

The fifth condition required an independent assessment of that program, included with the application for registration or provisional registration. Neither this assessment nor the security assessment described below had to accompany an earlier application for preliminary approval. The assessor had to certify that the program was operating with sufficient effectiveness to provide reasonable assurance. The assessor had to be qualified, objective and independent, and free from conflicts of interest. In our reading, this asked an applicant to show audited compliance before it held any data.

Security as a registration condition

The sixth condition was an information security program that complied with the Standards for Safeguarding Customer Information, 16 CFR part 314. Part 314 is the Federal Trade Commission’s Safeguards Rule. The seventh condition required an independent assessment of the security program. It was due with the application and at least every two years after. Each later assessment was due within 60 days after the period it covered.

The security assessment had to set out the administrative, technical and physical safeguards. It had to explain why they fit the entity’s size, activities and the sensitivity of the data. It had to explain how they met or exceeded the Safeguards Rule, and certify that the program was operating effectively. The rule placed a recurring, third-party security attestation at the center of eligibility.

The eighth condition was supervisory. The entity had to acknowledge that it was, or consent to being, subject to the Bureau’s supervisory authority. In our reading, this closed a gap. A small data company might otherwise fall outside the Bureau’s supervision. Registration would bring it inside by agreement.

Clocks that ran before the data counted

The rule set two routes to registration. Before the furnishing duty took effect, an entity could seek preliminary approval, then full registration. The deadline for preliminary approval applications was 90 days from the effective date of section 1041.11. The deadline for the full application was 120 days after preliminary approval. The Bureau could waive these deadlines.

After the furnishing duty took effect, an entity could seek provisional registration. A provisionally registered system would become registered automatically 240 days after the Bureau published notice of its provisional registration. The Bureau would publish each registration, suspension or revocation on its website and in the Federal Register. Registration took effect for these purposes on the date of website notice.

The furnishing duty used its own clock. A lender had to furnish to systems registered for 180 days or more, or provisionally registered for 180 days or more. In our reading, a new system would carry data for months before it became one that lenders had to consult. The rule built a ramp so that a system’s records would be reasonably complete before any lender relied on them.

What the conditions would cost an applicant

Each condition has a price. Real-time intake and reporting require engineering and operations staff on call. A compliance program with an independent assessment requires counsel, a compliance function and an outside assessor. A Safeguards Rule program with biennial independent assessments requires security staff and a second outside assessor. Consent to supervision brings examinations. An applicant would bear all of this before the furnishing duty made its service necessary to lenders.

The rule did not set fees or say how systems would charge lenders. In our reading, pricing would be left to the market. A registered system would hold a position every covered lender had to deal with. A small lender would have no choice about using it. That raises questions about fees and service terms that the rule text did not answer.

Suspension, revocation and appeal

The rule gave the Bureau power to suspend or revoke a registration if a system stopped meeting the conditions or posed an unreasonable risk to consumers. Except in cases of willfulness or where the public interest required otherwise, the Bureau would give written notice and a chance to fix the problem first. A revocation would take effect for underwriting purposes five days after website notice.

An entity could appeal a denial, suspension or revocation within 30 business days. For lenders, the suspension rules meant that the list of systems they must use could change. A lender’s furnishing and reporting integrations had to follow the Bureau’s published list. In our reading, the registry would be a live configuration item in each lender’s systems, not a fixed fact.

The decision: how a lender used reports, its own records and the fallback

The network existed to inform two decisions. Under the full-payment test, a lender had to determine whether the borrower could afford the loan. Under the principal-payoff option, a lender had to confirm that the borrower was eligible for a lighter path. Both depended on borrowing history.

The full-payment test

The fact sheet set out the inputs. The lender had to verify income and major financial obligations and estimate basic living expenses for the month with the highest sum of payments. The lender could reasonably rely on stated income if further verification was unavailable. It could reasonably rely on stated rental housing expenses. The rule also capped short-term loans made in quick succession at three, followed by a 30-day cooling-off period.

The borrowing history fed two parts of this. It helped show major financial obligations, because other covered loans are obligations. It also showed whether a borrower was in a sequence of loans that triggered the cap and cooling-off period. Without a view across lenders, a borrower could take a third loan at one store and a fourth at another. The network was designed to stop that.

The principal-payoff option

The principal-payoff option depended even more on the network. A lender could not offer it to a consumer with recent or outstanding short-term or balloon-payment loans. It could not offer it if the consumer had already had more than six short-term loans, or been in debt on short-term loans for more than 90 days, in a rolling 12-month period. Those are cross-lender facts.

The fact sheet drew the consequence. If no registered information systems were available, lenders could still make loans under the full-payment test without a system report. They still had to check their own records. But principal-payoff loans could not be made if no registered information systems were available. In our reading, the lighter path depended entirely on the network’s existence.

The fallback and its incentive

The fallback created an incentive that is easy to miss. If no system qualified, the full-payment test still worked on a lender’s own records and verification. The principal-payoff option did not work at all. A lender that relied on the lighter path would want the network to exist. A lender that relied on the full-payment test might be indifferent. In our reading, the rule’s data design shaped which lenders had a stake in the registry’s success.

What a report could and could not tell

A report from a registered system would show covered loans furnished to it. It would not show loans outside the rule’s coverage. It would not show loans by exempt lenders that did not furnish, or loans from before the system existed. A lender reading a clean report would know only that no covered, furnished loans appeared. In our reading, a careful lender would treat a report as one input, with its own records and the borrower’s statements.

The rule asked lenders to obtain a report from a national consumer reporting agency as well, as the Bureau’s later revocation described. That gave two views: the specialty network for covered loans, and a national file for other obligations. Each had different coverage and freshness. A decision engine would need rules for conflicts between them.

Accuracy and disputes

The fact sheet said registered systems, as consumer reporting companies, would be subject to applicable federal laws. In our reading, that brought the Fair Credit Reporting Act’s accuracy and dispute framework to the network. A consumer refused a loan because of a reported loan could seek to correct the record. The lender that furnished the loan would then be asked to verify it.

That creates a loop between lenders. A furnishing error at one lender could cause a denial at another. The second lender would not control the record. The first would not see the denial. In our reading, the network’s fairness would depend on fast corrections, and lenders would need a process to handle dispute requests about loans they had furnished.

Decision logic as a system

For lenders, the rule turned underwriting into a defined data process. Each application would call one or more registered systems, a national agency, and internal records. It would calculate residual income or a debt-to-income ratio. It would check sequence counts and cooling-off periods. It would choose between the full-payment test and the principal-payoff option. Each step depended on data from outside the firm.

That has a resilience side. If a registered system was down, the lender would need a rule for what to do. The fact sheet’s fallback applied when no systems were available in the regulatory sense. It did not address a technical outage at a registered system. In our reading, lenders would need their own procedures for outages, documented in advance, and counsel’s view on how the rule applied.

The records: what a lender had to keep, and what the payment rules needed instead

Section 1041.12 required a written compliance program and record retention. The compliance program had to be reasonably designed to ensure compliance and appropriate to the lender’s size and complexity. Records had to be kept for 36 months after a covered loan ceased to be outstanding.

Images of the evidence

For loans under the full-payment test and the principal-payoff option, the lender had to retain or be able to reproduce the loan agreement and documentation obtained to originate the loan. The rule listed the consumer report from a registered information system, verification evidence and the consumer’s written statement. In our reading, an examiner could ask to see the exact report a lender saw when it approved a loan.

Tabular records of the calculation

The rule went further than images. For loans under the full-payment test, a lender had to keep electronic records in tabular format. They had to show its projection of the consumer’s net income and major financial obligations for the relevant month. They had to show the calculated residual income or debt-to-income ratio, estimated basic living expenses, and other consumer-specific information considered.

For all covered short-term and balloon-payment loans, the tabular records had to include the furnished loan information, whether the lender took vehicle security, and the loan’s number in a sequence. They also had to include the days past due for any late payment, up to 180 days. In our reading, the rule specified a data model for supervision. An examiner could load these tables and test a lender’s decisions at scale.

Evidence that depends on outside data

The records requirement shows the rule’s dependency again. A lender’s evidence of compliance would include reports from registered systems. If those systems did not exist, the evidence would consist of internal records and borrower statements. The rule anticipated this for the full-payment test. It did not make the network optional for the principal-payoff option.

The payment rules used data the lender already had

The payment protections worked differently. A lender had to give notice before the first withdrawal attempt, and notice of unusual attempts. After two consecutive failed attempts, it needed a new authorization. Every input to those duties already sat in the lender’s own payment systems: the schedule, the attempt, the return code, the authorization. No outside network was needed.

The notice duty shows the same point. The fact sheet said the first notice had to give the timing, amount and channel of the coming payment transfer. If a later transfer differed in amount, timing or channel from what the consumer might expect, the lender had to send a specific notice of the change. A lender’s payment schedule already holds each of those fields. Building the notice is a matter of reading the lender’s own data and sending a message before the debit.

The two-attempt limit works the same way. An unsuccessful attempt, the fact sheet said, includes a debit returned unpaid or declined for insufficient funds. The lender receives those returns. Counting two in a row and blocking a third is logic on internal data.

In our reading, that difference is the clearest lesson in the design. The underwriting provisions depended on a shared system that others had to build, register and season. The payment provisions depended on data the lender generated itself. When the rule’s fortunes changed, the two halves had different prospects.

What the records cost to build

The tabular records requirement would affect lenders unevenly. A lender with a modern loan platform could add fields and export tables. A lender with spreadsheets or a legacy point-of-sale system would need new tooling. The rule did not prescribe a system. It prescribed the content. In our reading, the cost fell on data engineering, not on legal drafting.

How the design reached different lenders and data firms

The rule’s coverage was broad, and its data design reached each kind of firm differently. The fact sheet and press release said the rule applied to banks, credit unions, nonbanks and their service providers. The exemptions and exclusions changed who carried the data load.

Storefront payday and title lenders

For storefront lenders, the rule meant connecting each point of sale to the registered systems. A store would need to request a report before lending and furnish the loan at consummation. A lender with several locations would need a central integration. Title lenders could not use the principal-payoff option, so their path ran only through the full-payment test, which could operate on internal records if no system existed.

Online lenders

Online lenders would integrate through application interfaces. In our reading, the furnishing and reporting calls would sit inside the funding flow. Identity matching would be a sharper problem, because applicants might provide data that differs from what storefront lenders hold. Where an online lender collects through account debits, the payment protections would also reach its collection systems.

Banks with deposit advance products

The press release named deposit advance products among the covered loans. A bank offering such a product would face the same furnishing and reporting duties for covered loans. For a bank that already furnishes to national consumer reporting agencies, the registered systems would be a new, separate destination with its own format and timing. In our reading, a bank would treat it as a new data obligation, reviewed through its normal change process.

Credit unions and community banks

The fact sheet said loans from lenders making 2,500 or fewer covered loans a year are usually small personal loans made by community banks or credit unions to existing customers or members. The exclusion for loans meeting the payday alternative loan parameters also mattered to credit unions. In our reading, a small institution that stayed inside these lines would avoid the furnishing duty for those loans.

Those institutions still had a data question. To rely on the exemption, a lender would need to count its covered loans and measure revenue share each year. Crossing the line would bring the full regime. The fact sheet also said the penalty-fee prevention measures did not apply to banks or credit unions lending to their own customers where the loans could not generate overdraft or insufficient funds fees.

Data companies that might register

For data companies, the rule was a product specification and a supervisory bargain. A company would need real-time intake and reporting, a federal consumer financial law compliance program with an independent assessment, a Safeguards Rule security program with biennial independent assessments, and consent to Bureau supervision. In return, it would become a system every covered lender had to use. That is a strong market position, with heavy obligations.

Loan platforms and other service providers

Software vendors serving lenders would carry much of the work. Their products would need to produce furnishing files, call reports, store tabular records and handle payment notices. The press release said service providers were covered. In our reading, vendors would face questions from lender clients and possibly from examiners about whether their products produced compliant data.

Employers and wage-advance programs

The rule excluded certain wage-advance programs offered by employers or their business partners. The fact sheet described the conditions, including that the employer could deduct the advance from the next paycheck, or the partner could debit it on payday, with no other collection allowed. Firms structuring such programs would need to stay inside those conditions to avoid the furnishing and testing regime.

Objections, in order

The argument that the data network was the rule’s decisive and exposed part faces objections. We take the strongest six in order.

1. The underwriting tests were the substance; the network was plumbing

The tests were the substance of consumer protection. But the rule wrote the tests so that some of them could not run without the network. The principal-payoff option needed it outright, and the full-payment test lost its cross-lender view without it. Plumbing that a rule cannot work without is part of the rule’s substance.

2. Credit reporting already existed, so the network added little

National consumer reporting agencies existed. The rule asked for something different: covered-loan data furnished at consummation and reported substantially simultaneously. In our reading, the rule’s own eligibility conditions suggest the Bureau did not think existing reporting met that need. Whether a national agency could have met the conditions is a separate question the rule left to applicants.

3. The registry would have filled once the rule took effect

Perhaps it would have. The rule gave companies an incentive to register, since every covered lender would have to use them. But registration required independent assessments, consent to supervision and real-time systems before any revenue arrived. The rule also allowed only a waivable 90-day window for preliminary applications before the furnishing duty took effect. The sequence asked a lot of a market that did not yet exist.

The Bureau’s 2020 revocation said it was based on a re-evaluation of the legal and evidentiary bases for the provisions. We do not argue the data design caused the revocation. Our point is narrower. The data design made the underwriting provisions slow to start and dependent on third parties, while the payment provisions were not.

5. Shared data networks raise privacy and security risks

They do. A real-time registry of short-term loans would hold sensitive information about financially stressed consumers. The rule’s answer was a Safeguards Rule program, independent assessments, Bureau supervision and Fair Credit Reporting Act status. Whether that was enough is open. The risk supports our argument that the network was the hardest part to stand up.

6. Lenders could have built the cross-lender view themselves

Lenders could, in principle, share data through private arrangements. The rule did not rely on voluntary sharing. It required furnishing to every qualifying system. In our reading, that choice reflected the gap the Bureau saw. A lender acting alone cannot see a competitor’s loans.

What a compliance or data leader could check

The payday rule’s data design is gone, but its lessons apply to any rule that depends on data from outside the firm. Banks, credit unions and fintech lenders face such rules in other areas. The design suggests a short set of checks:

  • List each compliance duty that depends on an outside data source, and what happens to the duty if the source is unavailable.
  • For each outside source, record the matching key and test how often internal and external records fail to match.
  • Confirm that furnishing obligations name an owner in technology and in operations, as well as in compliance.
  • Check that vendor contracts name the data each platform must produce for regulators, its format and its timing.
  • Keep decision records in a form an examiner can load and test: inputs, calculations and outcome for each decision.
  • Separate controls that run on the firm’s own data from controls that depend on others, and rank their fragility.

The checks also give a leader something concrete to report. A board or examiner can be told which obligations depend on outside data, how matching performs and where the fallback is. Those are measurable facts. In our view, a rule’s data dependencies deserve the same review as its legal text, because they decide whether the rule can run on day one.

With hindsight

The rule was published in the Federal Register on November 17, 2017, and took effect on January 16, 2018, according to the Bureau’s rule page. Sections 1041.2 through 1041.10, 1041.12 and 1041.13 had a compliance date of August 19, 2019. On the effective date, the Bureau issued a statement saying it intended to engage in rulemaking to reconsider the rule.

The same statement addressed the registry directly. It said the effective date established April 16, 2018, as the deadline to apply for preliminary approval as a registered information system. It noted the Bureau could waive the deadline. Recognizing that the deadline might cause some entities to prepare applications, the Bureau said it would entertain waiver requests from any potential applicant.

Litigation followed. The Bureau’s 2020 revocation recounts that a challenge to the rule was filed in the Western District of Texas on April 9, 2018. On November 6, 2018, the court stayed the August 19, 2019 compliance date pending further order. According to the Bureau’s rule page, it issued two proposals on February 6, 2019. One would revoke the mandatory underwriting provisions. The other would delay their compliance date.

On June 6, 2019, the Bureau issued a final rule delaying the compliance date for the mandatory underwriting provisions, according to its rule page. The 2020 revocation gives the delayed date as November 19, 2020. On July 7, 2020, it issued a final rule revoking the mandatory underwriting provisions. The revocation was published on July 22, 2020, and took effect on October 20, 2020.

The revocation removed more than the tests. It revoked the furnishing provisions and the process for registering information systems, sections 1041.10 and 1041.11. It revoked the related recordkeeping provisions and the principal-payoff exemption. The Bureau said it was acting on a re-evaluation of the legal and evidentiary bases for those provisions. In our reading of the stay, the delay and the revocation, lenders never reached a date when they had to furnish to a registered system.

The payment provisions took a different path. They stayed in the rule. In a January 10, 2025 blog post, the Bureau said that after litigation, the court of appeals had rejected the lenders’ claims, affirmed the rule and upheld the Bureau’s unfairness finding. It said lenders must begin complying with the payment requirements on March 30, 2025.

On March 28, 2025, the Bureau announced it would not prioritize enforcement or supervision actions seeking penalties or fines under the payment withdrawal and payment disclosure provisions once they became operative. It said it was contemplating a proposal to narrow the rule’s scope. As checked on November 15, 2026, those are the latest Bureau statements on the payment provisions that we cite.

What held up is the distinction the design implied. The provisions that depended on a shared network that others had to build were delayed and then removed before any lender had to use the network. The provisions that ran on each lender’s own payment data survived court challenge and took effect, even if enforcement priorities changed. In our reading, the lesson for rule-dependent data systems is direct. A control is only as available as the data it needs.

Spotted an error? Tell us and we will check it and correct the article where needed.

  • payday lending
  • CFPB
  • registered information systems
  • consumer reporting
  • ability to repay
  • data governance