Termina

Reading a target's financial and transactional data at a scale no deal team could manage by hand, and turning it into findings on growth quality, is the product. Without the models only the raw data the client already had would remain.

Operators work alongside the models in producing each scan, and every page says the output does not substitute for a subscriber's own analysis or serve as the primary basis for an investment decision. A confidence indication on a benchmark comparison, or a flag for thin data, has no published form.

Definitions come from a metrics glossary and platform documentation, which explain each computed measure without validating it. Error rates, cohort construction and hit rates across scans are not published.

More than 1,000 scans a year, a part in over 10 percent of Series A and later deals and over 20 percent of Series B and later deals, and 20 billion data points processed are the aggregates. Case studies describe findings without naming the investor or company.

Cross client learning is sold as a feature: models recall insights from thousands of previously observed deals, so one investor's target feeds the benchmarks another's is measured against. What is aggregated, at what threshold, and whether a subscriber can opt out are not described.

A target's full transaction history handed over during a competitive process is commercially acute even with little personal data in it. Retention, deletion and the fate of a losing bidder's target data are not published, nor is a subprocessor list.

SOC 2 Type II compliance is stated, with a trust center hosted on Drata and a responsible disclosure route. Report period, and whether the report downloads or must be requested, are not shown.

Securities disclaimers in the footer state that nothing is an offer and that past performance is not indicative of results. The fiduciary and diligence duties that govern how investors use these scans are not addressed, and no supervisor or rule is cited.

The reference distribution reflects the sectors, stages and models subscribers transact in, so a target outside it is measured against norms that may not fit. Cohort size behind a benchmark, and how a thin comparison set is signaled, are unpublished.

Risk is placed on the subscriber, who keeps the underlying data and can reconstruct any finding. Correction of an erroneous scan, or notice if a benchmark proves wrong, has no published process.

Model provider, hosting and subprocessors are all unnamed, and the data side is described only in kind: subscriber records plus the pool of observed deals.

An API with Python and JavaScript SDKs covers benchmarking, financials and deal management, with Excel export and an MCP server for AI clients. No accounting, billing or data room system is named as a source, so data still arrives by upload.

Where target data rests, and for how long, is not stated: no hosting provider, region or private option is named.

Turn key scans sit in a catalog that implies standard prices, yet none is stated. Per scan, per portfolio company or subscription billing all remain unclear.

Venture capital firms, corporate development teams and founders each get pages, with use cases from due diligence to portfolio reviews and benchmark development. Analysis is confined to growth quality in operating businesses, leaving credit, real assets and public markets untouched.