Trusting Social

The removal test leaves no assessable population, since the entire proposition concerns people conventional scoring cannot evaluate. Proprietary machine learning over alternative social, web and mobile signals produces risk scores where no credit file exists, applied across more than a billion consumers, and the same modelling extends to identity verification, fraud prevention and predicting residential and office addresses from indirect evidence. The founders describe the enterprise as rebuilding scoring technology from the ground up rather than adapting existing methods.

No oversight model is described in either direction. Scores and signals are supplied to institutions that make lending decisions, which implies the decision stays with the lender, and nothing published states whether scores feed automated approval, what thresholds apply, whether a human reviews adverse outcomes, or what happens when the model and conventional evidence disagree. For a company whose scores reach the top of several national banking systems, the absence of any stated position on automated decisioning is a substantive gap.

No accuracy, lift, default reduction or validation figure appears anywhere despite thirteen years of operation and a billion consumers scored, which is a striking omission for a company whose product is a prediction. The founding team's credentials are relevant and stated, including a doctorate in econometrics and prior responsibility for global credit risk at a major international bank, and credentials are not measurement. Nothing describes model governance, revalidation cadence, or how an institution assesses whether a score performs on its own portfolio.

Thirteen years of operation with more than 130 institutional customers across four countries, over a billion consumers scored, and the sharpest single proof point available in this category: a regional executive states that six of the ten largest banks in the Philippines are existing clients already running the credit scoring system, alongside more than 40 institutional clients in Vietnam and a named digital bank partnership.

More than 200 million dollars has been raised, a national central bank selected the company for a supervised programme, a global card network admitted it to its fintech partner programme, and it won an artificial intelligence award at a major industry festival. Institutional customer count has declined from a previously reported 170.

No boundary statement was located. Models trained across a billion consumers serve more than 130 institutions that compete directly, including six of one country's ten largest banks, so a shared scoring layer necessarily carries learning between rivals. Nothing states whether a lender's repayment outcomes improve models scoring its competitors' applicants, what a customer contributes by using the service, or how long consumer profiles persist once scored.

No data protection agreement, consent framework, retention schedule, subprocessor list or lawful basis statement was located, and this is among the largest unaddressed privacy footprints in the index. More than a billion people are described as scored using social, web and mobile data, and a further product infers where individuals live and work from alternative signals, across four jurisdictions with materially different and mostly recent data protection regimes. An information security certification exists and addresses how data is protected rather than whether its collection and use are consented to.

Certification to the principal international information security management standard is held across operations, which is a third party assessed control set rather than a self declaration and puts this ahead of most of the index on this axis. It is also what a top ten bank's information security review would require before admitting a scoring provider. Held at B because no trust centre, penetration testing summary, subprocessor register or further framework such as a service organisation control report was located.

The strongest signal is direct engagement rather than a compliance claim: a national central bank selected the company for participation in a supervised programme, which means a regulator examined the approach and admitted it. Admission to a global card network's partner programme adds a second gatekeeper assessment, and information security is certified to a recognised international standard across operations. Held at B because no statute, rule or supervisory obligation is named for any of the four markets, and the credit reporting and alternative data rules in each differ considerably.

The inclusion footprint is the largest in this index and the need is documented independently, with research cited showing more than seven in ten adults across Southeast Asia are underbanked, meaning limited or no access to financial services. Scoring a billion people who conventional systems cannot assess is the substance of the claim rather than a framing of it. The counterweight is the input set and it is the most contested in the whole roster: social, web and mobile data.

Inferring creditworthiness from someone's social and web footprint imports whatever inequality is already encoded there, and the address prediction product extends inference to where a person lives, which correlates strongly with wealth and ethnicity in every market served. No fairness testing, subgroup analysis or adverse outcome data is published.

No guarantee, indemnity or correction process was located, and the individual's position is among the weakest in this index. A person can be scored from social, web and mobile signals, and have their home and work locations inferred, by a company they have never dealt with, in markets where no general right to an explanation of automated decisions exists. Nothing describes notification, access to the inputs used, or any route to challenge an inference or have it corrected once a lender has acted on it.

Input categories are named as social, web and mobile data alongside a general reference to alternative sources, and no individual provider, platform, telecommunications partner or data licensor is identified. For a company whose scores depend entirely on externally sourced behavioural signals, that leaves a buyer unable to assess whether access is contractual or scraped, how durable it is, or what happens to coverage if a major platform restricts access, which has occurred repeatedly in this data category.

Integration extends beyond data delivery into commercial structure, with co-lending arrangements connecting banks to consumer brands so credit is offered at the point of purchase, and a platform partnership with a large retail conglomerate intended to reach tens of millions of households. Membership of a card network's partner programme provides a distribution route into card issuers. That is deeper embedding than an interface alone. No named core banking, origination or bureau system appears and no developer documentation was located.

No hosting provider, region selection, residency commitment or private deployment option was located. Operating across four countries that have each introduced or tightened data localisation expectations for financial and personal information, while holding profiles on a billion consumers, makes residency a first order question for any institutional buyer, and published material does not address it.

No pricing appears, and this is nonetheless the most financially transparent vendor in the index because the parent files public accounts. Revenue is disclosed at 23.1 million dollars for 2025 against 21.2 million in 2024 and 18.1 million in 2023, with a net loss of 87.5 million for 2025 following a 33.8 million profit in 2024 and a 31.9 million loss in 2023, the swing attributed largely to a revaluation of financial liabilities rather than trading.

Cost lines are visible, with distribution and marketing falling to 8.6 million and administrative expenses rising 20 percent to 24.1 million. A buyer can therefore assess durability directly, and should note that revenue growth of roughly nine percent is modest against more than 200 million dollars raised.

More than 130 financial institutions across four large emerging markets, reaching commercial banks, digital banks, consumer finance companies and retail brands, with the top tier of at least one national banking system substantially covered. Functional breadth extends past scoring into identity verification, fraud prevention, address inference, co-lending and embedded finance, so the company participates across origination rather than supplying one signal. The four markets together represent well over a billion people, most of them outside conventional credit systems.