Fivvy

Models do the interpretive work, applying what the company calls unique data models to transaction history, mobile application behaviour and device metadata in order to predict churn, identify cross selling openings and surface contextual sales opportunities in real time.

Strip them and a substantial platform remains: the data pipeline connecting into a bank's systems, the credit bureau integration, the mobile delivery layer and the analytics presentation, all of which an institution would still find useful and all of which are the harder part of a three month implementation. This is the Psympl position, with models sharpening a data and delivery platform rather than constituting it.

The scale claim contains the oversight position. The company states it has started millions of hyper personalised conversations without requiring human capital or heavy marketing spend, which describes outbound customer contact generated and delivered automatically, and sales opportunities are presented in real time and contextually rather than assembled by a person.

Nothing published describes a review step before a personalised message reaches a customer, any approval workflow inside the institution, or what constrains the system from repeatedly targeting a customer who does not respond.

No accuracy figure, validation result, error analysis or model documentation was located. The published outcome, a 23 percent reduction in marketing cost through targeted retention, measures commercial efficiency rather than whether the model read a customer correctly, and those diverge precisely where it matters: a targeting model that concentrates on the easiest conversions lowers cost per acquisition while systematically misjudging everyone else. Nothing describes how churn predictions or opportunity identifications are validated against what customers actually did.

Reach is stated in the unit that matters for a personalisation product, with banking clients whose combined customer bases come to around two million end users, and the company describes having started millions of individually personalised conversations. Independent profiling puts headcount near 35 and estimated annual revenue around 7.4 million dollars, which for a company of this type indicates real commercial traction rather than pilots.

One outcome figure is reported, a 23 percent reduction in marketing costs through targeted retention. Backing came from the corporate venture arm of a listed technology services group, committing four million dollars across two years, which is strategic money from a company that also builds for banks. No financial institution is named as a customer, and the revenue figure is an outside estimate rather than a disclosure.

No data boundary statement was located. Behavioural personalisation improves markedly with exposure to more customer populations, and the platform serves banks and credit unions of all sizes that compete directly for the same depositors in the same regional markets, so whether spending patterns and response behaviour observed at one institution inform the models serving another is the material question. Nothing addresses it, and nothing states whether an institution can decline to contribute its customers' behaviour to a shared model.

A stated privacy position exists and it is more specific than most, with the platform described as emphasising ethical data collection and privacy compliance, and delivering insight into user behaviour and preferences without invasive tracking. Scope commitments of that kind are what lift Psympl on this axis too. What sits uneasily beside it is one of the platform's own named inputs: device metadata.

That is the same category of signal that produced this index's sharpest privacy finding at FinBox, because handset and device characteristics correlate with income and circumstance in ways a customer never consents to individually. A claim of non invasiveness alongside device level collection needs reconciling, and no retention schedule, subprocessor list or data processing terms were located.

No attestation, certification, trust centre or enumerated framework was located. Secure transaction processing and privacy compliance are both asserted without naming a control set or an independent assessment. For a platform ingesting the full transaction history of banking customers across multiple institutions and appending bureau data to it, a published assurance set is what each institution's own vendor review will require before any data moves.

Privacy compliance is asserted and no regulation, supervisor or instrument is named anywhere. The gap matters because personalised financial marketing to a bank's own customers sits inside advertising rules that attach separately to credit, deposit and investment products, and inside the prohibition on unfair, deceptive or abusive practices governing how a firm may communicate with consumers. A platform generating that communication at scale operates within all of them and identifies none, which is the same omission recorded for Psympl.

The combination here is sharper than the sum of its parts. Bank customers are profiled from their transaction history, application behaviour and device metadata in order to surface real time, contextual sales opportunities, which means a model deciding which financial products a person is offered and when, optimised for the institution's conversion rather than the customer's outcome.

Device metadata is the aggravating input, since it is the category that produced this index's most invasive finding at FinBox, correlating with income and circumstance through handset and application characteristics. Credit bureau data is appended on top. Against all of that the company markets itself on non invasive tracking, and an explicit non invasiveness claim sitting beside device level collection is the most pointed version of this pattern recorded so far. No fairness testing, no analysis of who is targeted and who is not, no disclosure to the customer and no opt out was located.

No guarantee, indemnity or falsifiable commitment was located. The institution retains responsibility for what it sends its own customers, which is where the regulatory duty properly sits, and it can presumably configure what the platform surfaces.

The customer has nothing: a person profiled from their transactions, application behaviour and device is not told, cannot see what the model concluded about their spending or saving, has no route to correct a wrong inference, and no opt out from behavioural targeting is described anywhere.

The external data dependency that shapes the output is named rather than described generically, with a major credit bureau identified as combined with the company's own technology to automate the identification of opportunities, so a buyer knows whose data supplements their own customer records. The strategic investor is a listed technology services group whose relationship is described as more than financial, which suggests an engineering dependency worth noting. What is not disclosed is the model layer, with no provider named for the analytics or personalisation components, and no subprocessor list or hosting arrangement located.

The critical external dependency is named, with a major credit bureau identified as the data source combined with the company's own stack to automate opportunity identification, which tells a buyer whose file is being appended to their customer records.

Implementation is quantified at three months to move an institution to data driven operation, with integration described as agile against existing systems, and delivery reaches the customer through mobile applications on both major platforms. What is not published is the institutional side: no core banking system, digital banking platform or marketing automation tool is named, so a bank cannot establish what connecting actually involves.

No hosting provider, region selection, residency commitment or private deployment option was located. The question is live because the company is incorporated in the United States with founders and origins in South America and a stated regional focus across the Americas, which spans several data protection regimes, and the material held is the transaction and behavioural history of banking customers.

No rate, tier or charging basis is published. Third party listings describe the pricing structure as notably competitive against comparable products, which is a claim about position rather than a disclosure, and nothing indicates whether cost scales with end users reached, conversations initiated, institution size or data volume. For a product sold to institutions ranging from small credit unions to national banks, the unit is the whole question.

The buyer set spans traditional banks and digital institutions of all sizes, with credit unions named explicitly alongside fintech companies, and the same platform is presented as serving both business and consumer customer bases. Geographic ambition is regional rather than global, with the strategic investor positioning the company as a leading platform across the Americas and the founding team drawn from South America.

The limits are function and language: this is customer intelligence and personalisation rather than any operational process, and the product is supported in English only, which constrains reach across the very markets it targets.