73 Strings

Unlike its extraction peers, 73 Strings would still have a working product without its models. Take them out of the extraction module and documents have to be keyed by hand. Take them out of the valuation module and discounted cash flow modeling, market comparables and benchmarking still run, as do scenario periods, version control and the audit trail. That is finance and software engineering, not model work, and valuation practitioners built it.

The company's own chief executive frames it the same way, describing the data substrate as what makes the models trustworthy rather than the other way around. The models compress analyst effort on top of that engine.

Human judgment is designed in, not bolted on, which fits an output that is an opinion of value. The company states that automated outputs, including data sources, peer comparisons and the valuation arithmetic, are presented in standardized formats for human verification. Analysts can override assumptions such as discount rates and growth projections, supporting the reasonableness testing that international valuation standards require. A full audit trail and version control sit underneath. No account is published of who may override, whether overrides are flagged as such downstream, or whether any sampling or independent check runs across them.

Traceability is well covered and validation is not. The company publishes an audit trail, version control, standardised presentation of automated outputs and full data lineage, all of which let a reviewer follow how a number was produced. None of it establishes whether the number is right. No extraction accuracy rate appears, no error rate by document type, no drift monitoring and no model documentation.

The absence that stands out most in this particular domain is back testing: private markets valuations are eventually settled by realised exits, which makes marks uniquely checkable against outcomes, and no published material compares the platform's valuations to what assets actually sold for.

The scale figures describe the buyers, not the results. Two clients are named publicly with their own scale attached: a French listed investment company at roughly thirty five billion euros and a Belgian holding company at over nine billion. The company states its client base collectively manages close to $10 trillion across private equity, growth equity, venture capital and private credit. A case study exists with one of the largest alternative managers in the world.

A global head of financial affairs at a thirty $1 billion private equity fund is quoted on outcomes, though that firm is not named. The speed claim is a round multiple with no client attached. Prestige backing is not outcome evidence.

Nothing published states where client data stops and platform data begins. The company markets a connected client community in which clients benefit from shared insight across some of the world's leading alternative asset managers. It also sells access to market comparables, performance metrics and investor benchmarks inside the same platform. Together those raise the question of whether benchmark and comparable data comes from client portfolios.

What a general partner uploads is the unpublished financial performance of companies it competes to buy and sell. The network effect claim stays unresolved, with no statement on whether contribution is optional or whether a manager can see comparables drawn from a rival's book.

The platform, security and company pages carry no privacy statement on how client and portfolio company information is handled, and nothing on retention periods, deletion on exit or processing locations. The data is portfolio company financials, fund level performance and investor reporting, not consumer financial information, so statutory privacy exposure is largely absent. The gap is one of disclosure, not of category. It matters for a platform holding the unpublished financials of thousands of private companies that never agreed to anything with this vendor.

One attestation is referenced, and it is easy to miss. A product page answering questions about audit readiness states that the extraction audit trails are covered by a SOC Type II certification. That is a named attestation, not a vague assurance. Three limits remain. The reference sits inside a product answer, not on any security or compliance page. No trust center, report request route or scope statement accompanies it, and no second standard appears beside it.

For a company whose clients manage close to $10 trillion, and whose platform holds the unpublished financials of thousands of private companies, a single passing mention is a thin public security position.

No license is held or required, and the company states its regulatory position clearly. It positions the product against international valuation standards, fund reporting obligations and audit scrutiny, and describes outputs as audit defensible. It also claims to be the only solution in its category with no independence or service conflicts.

That is a substantive positioning statement in a field where valuation advisers have long been criticized for selling advice and valuations to the same client. The claim sits awkwardly beside the company's own team of valuation specialists, who deliver expert work alongside the software, and nothing published reconciles the two. Independence of the valuation function remains the manager's duty, not the tool's.

A protected class frame does not fit a valuation platform and forcing one would miss the exposure entirely. The bias that matters here is directional. The party operating the model is the party whose management and performance fees are computed from the resulting net asset value, and the platform openly supports overriding the discount rate and growth assumptions that drive that number. Any systematic drift toward optimistic marks is a transfer from investors to managers.

Nothing published states whether overrides are aggregated, compared against realised exit outcomes, or disclosed to limited partners and auditors. A second and quieter version of the same problem is that comparables drawn from a client pool weighted to large managers will fit small and non United States assets worst.

The published promise is high, and no published commitment stands behind it. Outputs are described as audit ready, defensible and results you can trust. They land in fund financial statements, capital account statements and fee calculations, so an error travels straight into audited numbers and investor reporting. No service level or warranty is published, and no correction obligation or remedy. Nor is the opposite: no disclaimer states that the valuation conclusion remains the manager's own. The audit trail makes an error traceable after the fact, which is useful, but it is not recourse.

The disclosure stops at a general mention: LLMs are acknowledged as the foundation for agentic capabilities on the roadmap. No provider is named. No distinction is drawn between models built in house and models called from an external service. Nothing states whether client documents or portfolio company financials pass to a third party during processing. A close competitor answers the same question explicitly and by name, so the disclosure is commercially possible. Until it is made, an institution cannot tell which counterparties its confidential portfolio data actually reaches.

The integration story is internal rather than external. Extraction, monitoring and valuation run as three modules over one private capital data model, which is a real architectural claim and the reason time series stay consistent from one reporting cycle to the next, and the platform ingests financial and operational inputs from multiple sources and exports into presentation formats used for investment committee and investor reporting.

What is not published is a named connection to anything a client already runs. No fund accounting system, administrator, general partner portal, market data provider or accounting platform is identified as an integration partner, which leaves a buyer unable to judge how the platform reaches the systems holding the rest of its record.

Delivery is cloud hosted software, and the published material says little beyond that. No hosting regions, residency commitments or single tenant options are published, and there is no account of where processing occurs. The gap matters for this client base. It includes European managers with their own data location expectations and Middle Eastern sovereign investors who commonly impose them by contract. The company reorganized in 2026 specifically to serve those regions from a single global model.

Every route through the published material ends at a demo booking. No rates, no tier ladder, no unit of charge and no indication of whether pricing follows seats, funds, portfolio companies or valuations produced, which matters because those units scale very differently for the same firm.

The company also pairs software with a team of valuation specialists, and nothing distinguishes what is licensed from what is delivered as expert work, so a buyer cannot see where the software price ends and the service price begins.

Coverage runs both sides of the private capital relationship and across the asset owner tier above it. The published client base spans pension funds and sovereign wealth funds, plus limited and general partners, including multi strategy alternative managers. Clients sit across North America, Europe and the Middle East, with assets stated at close to $10 trillion. Strategy coverage is listed, not implied: private equity and private credit, venture capital and infrastructure.

In 2026 the company reorganized around a single global operating model to serve those segments consistently across regions. Inside private markets the breadth is real. It does not extend to banks, insurers or retail institutions.