Early Warning Services

The Clearwater shape and a clean instance of it. Certos is the graded product line and its scores are genuinely model driven, with embedded machine learning assessing the likelihood of check and automated clearing house returns within thirty days, and proprietary real time risk scoring applied to payments before funds are released.

But the asset underneath is a consortium database more than three decades old, and stripping the models leaves the National Shared Database, deterministic account ownership and status verification, the negative deposit file and the two payment networks entirely intact, which is most of the company.

Distinguished from the rejected data product class because the deliverable sold is a score rather than a feed: Predict New Account Risk and Deposit Chek exist only as model output, and that output is sold to third parties rather than consumed internally, which is the test that separated the financial data major from the internal copilot rejections earlier in this sweep.

A published scope disclaimer rather than a described control, and it is a shape worth naming. The vendor states in the footer of every product page that its solutions provide insights supporting consistent policy based decisions and that customers retain full control over how those insights are used, and states separately that it provides threat intelligence signals but does not make bank account related or financial access decisions.

The consumer channel repeats it, telling anyone declined an account that the decision was made by the institution that referred them and not by the vendor. That is more specific than the usual assertion because it names precisely what the system will not do and places the decision with the supervised party, the same devolution recorded on the consortium platform elsewhere in this pocket.

Held at B because nothing enforces it: no threshold, no gate, no escalation path and no floor stating what the vendor will not permit an institution to automate. Worth recording that the disclaimer also serves a legal function under the consumer reporting statute, since the party taking adverse action carries the notice obligation, so the governance posture and the liability posture point the same way.

Advanced analytics, embedded machine learning models and proprietary risk scoring, with no documentation, no performance figures, no validation report, no benchmark and no drift policy, for models whose output contributes to bank account denials.

The statutory obligation to follow reasonable procedures assuring maximum possible accuracy attaches to the underlying file rather than to the models, and consumer litigation alleging accounts wrongly reported as fraudulent indicates that obligation is contested in practice. That is an input quality problem rather than a model governance disclosure, and it belongs on this row precisely because the vendor publishes nothing that would let a buyer distinguish the two.

Scale figures are large, current and specific: more than 11.4 trillion dollars in payment and deposit transactions screened in 2025, 124 million new account applications supported, and more than 3 billion dollars in potential fraud loss prevented. Two executives are named and quoted on the product launch. Held at B because no customer is named with an outcome attached, which is the A bar.

One detail cuts in the vendor favour and is worth recording as a positive claim shape: the fraud prevention figure carries a footnote citing the company own internal analysis of database inquiries and stating explicitly that the analysis is unpublished. Most vendors print the number alone. Printing the number and disclosing that it rests on unpublished internal work tells a reader exactly how far it can be trusted. It earns nothing extra, but it is the honest version of a claim this index usually sees dressed up.

Among the better pooled corpus disclosures in the index, and it introduces a distinction the ladder has not carried before. The corpus is named, the custodial role is named, the contribution model is named as give to get, participation is quantified at roughly 5,000 contributors, and anonymisation before redistribution is claimed. That matches the disclosure quality of the consortium platform elsewhere in this pocket, so the grade matches too.

The distinction is sharp: on that platform the pooled material is institutions own transaction data and the contributing party is the described party, whereas here the contributing parties are banks and the described parties are their customers. The consumer whose account history is traded for access to other banks account histories is neither a participant in the arrangement nor able to decline it.

Held at B on the standard ground that disclosing the fact of pooling is not disclosing the terms of it: no use limits, no retention period, no exit treatment and no statement of whether contributed records train the scoring models.

A statute supplies the posture rather than a policy page, the same route that lifted the actuarial consumer report business graded a day earlier. The company is a nationwide specialty consumer reporting agency, states so in its own consumer material, and appears on the federal consumer bureau published register of consumer reporting companies with the access route listed.

Access to a file is restricted by law to entities holding a permissible purpose, an individual may obtain their own file disclosure at no cost, and the vendor publishes specific handling for identity theft plus a named remedy for survivors of human trafficking under the trafficking rule, a level of statutory specificity almost nothing here reaches.

Held off A: no subprocessor list, no residency statement and no processing detail, and the underlying material is bank account history used to exclude people from the banking system, where the accuracy obligations attached to it are actively contested in consumer litigation.

No certification, attestation or trust portal located in the vendor own swept material during this pass, and the product site footer carries only a privacy notice and an accessibility statement. Nothing is credited, because an unverified credential earns nothing and that rule has to apply when it costs a grade.

Two banked checks, both cheap and both likely to move this: the parent domain rather than the product domain may carry a security or trust page that was not swept here, and a company performing account screening and payment services for depository institutions sits on the performing side of the bank service provider line, so federal examination evidence may exist in a form almost no other vendor in this index can produce.

Two independent routes, converging on the same grade. First, consumer reporting agency status is defined by federal statute, carries supervision by the consumer financial bureau and the federal trade regulator, and is confirmed by the company presence on the bureau own published register, which is the actuarial precedent applied to a second vendor.

Second, performing account screening and payment services for depository institutions places the company on the performing side of the bank service provider line, the distinction that moved a core banking incumbent from C to B earlier in this sweep, so it would qualify here independently. Held at B rather than A on the same reasoning as the precedent: this is status acquired by operating in a regulated category rather than a licence granted after examination.

Recorded as context and not as a grade driver, the consumer financial bureau sued the company and three owner banks in December 2024 over fraud handling on its payments network and dismissed the case in 2025, a matter concerning payments rather than screening.

The most conspicuous fairness gap in this build, and the framing makes it worse rather than better. The models contribute to whether a person can open a bank account, which is the best documented site of financial exclusion in the country, with account screening databases long criticised for shutting people out over small unpaid balances.

The vendor markets the opposite as a product benefit, stating that it reduces unnecessary barriers particularly for individuals with limited or nontraditional identity histories and that it expands financial access, a claim repeated across the launch material and the product pages. There is no fairness testing, no protected class analysis, no disparate impact study, no approval rate breakdown and no independent evaluation published anywhere to support it. Selling inclusion as an outcome while publishing nothing that would let anyone verify it is a distinct shape from the plain silence recorded elsewhere in this index.

The second grade above C on this axis in the index, arriving one day after the first and by the identical route. As a nationwide specialty consumer reporting agency the company gives an individual a published, individually exercisable remedy: obtain the file disclosure at no cost, dispute any item, trigger a reinvestigation that must be completed free of charge and generally within thirty days, and have unverifiable information corrected or deleted.

The vendor operates a dedicated consumer channel for people declined an account or a payment, publishes the dispute route, and adds specific statutory handling for identity theft and for survivors of human trafficking. Held firmly at B on the same two grounds as the precedent.

First, it reaches the underlying data record and not the model: there is no route by which a person can contest a risk score, only the facts beneath it, and the vendor states plainly that the adverse decision was made by the institution rather than by itself. Second, the mechanism has a performance record and it is mixed, with a live consumer litigation practice built around accounts wrongly reported as fraudulent and around disputes resolved without meaningful investigation. Two vendors now hold this grade, both obtained it from a consumer reporting statute written decades before this technology existed, and neither obtained it from anything the vendor chose to offer.

No model, provider, architecture or version named anywhere across the product line. The pattern holds that buying forces disclosure while building permits silence, and this is a build. The asymmetry worth recording is the same one found on the actuarial consumer report business: the data supply chain is disclosed at category level, since contributing participants are named as a class and quantified, while the model supply chain is not disclosed at any level at all.

Deep by position rather than by connector catalogue. The products are called synchronously inside the two highest value paths in retail banking, the account opening decision and the payment authorisation moment, and the vendor states that capability bundles are reached through a single interface so an institution implements once and adds further capabilities without new integration work.

One named integration counterparty exists, an indexed identity decisioning platform that resells the suite into community banks and credit unions. Held off A because no core banking, deposit origination or account opening platform is named as a certified connector, so beyond that single partnership the integration story is described by function rather than by counterparty.

Real time delivery of capability bundles through a single interface is stated, and that is the whole of it. Nothing published on hosting, region, tenancy, residency or subprocessing. The operation is domestic by construction, which implies United States residency, but implication is not disclosure and nothing states it. A notable silence for an infrastructure provider holding deposit account history on a large share of the adult banking population.

No price, tier, unit or pricing basis published for any product. The commercial route is a contact form with a picker listing the seven products of interest. Standard for a consortium seller where pricing is negotiated against contributed data volume, and consistent with the rest of this pocket.

Roughly 2,500 banks and credit unions and about 5,000 total participants across six distinct buyer types the vendor addresses separately, each with its own published product mapping: financial institutions, financial services companies, payment processing companies, merchants, government entities and identity theft protection service providers.

Institution size runs from the seven owner banks, among the largest in the country, down to community banks and credit unions, the latter reached through a partnership announced in May 2026 with an indexed identity decisioning platform. Government benefit eligibility determination is a further distinct segment served by a dedicated product.

The one real limit, recorded rather than deducted: this is a single jurisdiction operation covering the United States banking system only, a narrower geographic base than other vendors holding this grade.