Baselayer

Machine learning does real work at the top of the stack, clustering identity anomalies, detecting velocity patterns and coordinated fraud rings across the network, and driving agents that run risk checks and produce scored decisions. The foundation beneath it is data aggregation and deterministic matching: secretary of state filings, court and lien records, business registries and direct verification of tax identification numbers against federal records.

Apply the removal test and a comprehensive business data and verification service survives, which is a saleable product in its own right, so this sits with the data and orchestration vendors rather than the model native ones.

The automation boundary is stated in operational terms: low risk applications route through automatically while high risk cases are surfaced to a human with what the company describes as clear, explainable risk signals, and the scoring layer is presented as transparent rather than opaque. That is the right split for an onboarding decision and it lets a reviewer see why a case reached them.

What is not described is the machinery behind it, with no documented review queue, no configurable threshold detail, no sampling or quality control over automated approvals, and no route for a rejected business to have a decision revisited.

Explainability is claimed consistently, with decisions described as explainable and the business verification score as transparent, which is the right instinct for output that supports an onboarding or credit decision. The published measurement points the wrong way.

A 92 percent improvement in auto approval rates describes throughput and says nothing about correctness, and for a verification product the number that matters is how often an approved business turns out to be fraudulent or a rejected one turns out to be legitimate. No accuracy figures, model documentation, evaluation methodology or validation support were located.

Institution count is large, specific and consistent across dated sources, moving from 1,900 to 2,000 to 2,200 over successive periods, which is a better credibility signal than a single static claim. The strongest evidence is a named partnership with a major core processing group's decision solutions arm and its deposit account screening bureau, with a senior executive quoted by name and title describing the combined offering serving hundreds of financial institutions.

Performance is stated at under five seconds per verification and a 92 percent improvement in auto approval rates against legacy vendors. No end customer is named, that improvement figure carries no methodology, and a team reported in the low twenties implies most reach runs through partners rather than direct relationships.

The identity network is the central stewardship question and Baselayer answers half of it, stating that activity connected across institutions is anonymised, which is more than several network operators in this index disclose. The unanswered half is competitive rather than personal.

Inquiry patterns reveal which businesses an institution is evaluating, so a network built on application activity across 2,200 participants is aggregating commercially sensitive pipeline information, and nothing public defines what anonymised means in practice, whether participation is optional, or what a contributing bank's data becomes once it is in the shared layer.

Although the subject of a search is a business, the platform reaches individuals directly, verifying officers, beneficial owners and signers for identity, credit and fraud risk, which brings consumer data and its attendant obligations inside the workflow. It also ingests social media and review data about the business and by extension the people behind it. No published privacy framework, retention schedule, subprocessor list or statement on how officer level personal data is treated differently from entity data was located.

No trust centre, enumerated certification list, attestation scope or audit period was located in this pass. The concentration argues that attestations exist: a shared network holding application activity for more than two thousand institutions including government agencies would not pass those buyers' vendor risk programmes without them, and the core processing partnership would have imposed its own assessment. The grade records what an outside buyer can verify rather than a judgement on the controls.

Baselayer supplies technology and holds no licence, the expected posture, and it holds one data relationship worth noting: tax identification numbers are verified directly against federal tax authority records, which is a gated arrangement rather than a scraped source and gives its core verification a level of authority most competitors approximate. Product scope maps to customer identification obligations for business accounts and to beneficial ownership verification, and the platform is deployed inside government agencies as well as banks.

One marketed capability carries a fairness problem that deserves naming. The platform incorporates website, social media and review data to surface emerging credit risk, which means a business's online presence influences its access to capital, and thin online footprints correlate with age, rurality, language and immigrant ownership rather than with creditworthiness.

Small business lending now sits under expanding data collection and fair lending scrutiny, so the exposure is live rather than theoretical. Officer level identity and credit checks add the familiar name matching asymmetry. No demographic analysis or error rates were located.

The network that makes this product powerful also makes a wrong judgement unusually damaging, and nothing addresses it. A business flagged as synthetic, stacking or high risk can lose an account or a loan, and because inquiry and risk signals propagate across more than two thousand participating institutions, one institution's assessment can follow that business everywhere it applies next.

The affected company is not the customer, is not told a network judged it, and has no described route to see or contest the signal. No accuracy guarantee, remediation commitment or correction process was located.

Data provenance is described more concretely than most vendors manage, naming secretary of state filings, court records, business registries, lien data, direct federal tax authority verification and website, social and review sources, alongside a stated combination of public, proprietary and partner data. The core processing and account screening partnership is named openly, so a buyer can identify a significant part of the chain feeding a decision. What remains undisclosed is the model layer and the subprocessors: no providers are named for the clustering, scoring or agent components, and no subprocessor list is published.

Delivery runs through an enterprise interface and a web application designed to sit inside an existing onboarding flow without adding steps, returning results in seconds so verification does not become a stage of its own. The most consequential integration is commercial rather than technical: embedding inside a major core processing group's decision solutions portfolio places the capability in front of institutions that would never procure a company of this size directly. Public developer documentation, a status page, a changelog and a named connector directory were not located.

Delivery is cloud hosted software as a service serving domestic institutions and verifying domestic entities, so the cross border questions facing global vendors in this index do not arise in the same form. Residency is still undisclosed: no hosting regions, tenancy model, subprocessor chain or data location statement was located, which matters because the platform holds application and inquiry records contributed by thousands of institutions in one shared layer.

No rates, tiers, billing unit or minimum were located. The unit question matters for this product because verification volume scales directly with application flow, and a lender running seasonal small business origination has a very different cost profile from a bank onboarding commercial accounts steadily, with nothing public letting either model it.

Buyers span banks, fintechs, small business lenders with dedicated material of their own, and government agencies, and coverage claims reach effectively every one of the roughly 120 million businesses in the United States, which is the relevant completeness measure for this product.

The boundary is geographic and absolute: this is a domestic platform verifying American entities, with no international coverage, and within financial services there is no separate treatment of credit unions, insurers or capital markets.