Bring back the answer.
Approved providers compute inside their own environments and return bounded evidence, not broad private-data feeds.
Corroboration
Corroboration asks whether the life behind the file holds together
The point is not to collect more private data. The point is to ask whether source-side patterns support the applicant claim strongly enough for a lender to inspect and defend.
Boundary
A check can be useful without becoming a trail.
A lender may need to know whether a claimed residence, work pattern, or source-side context coheres. It does not need every raw event behind that answer.
Make the pattern inspectable.
The record shows what was checked, what boundary applied, and why the result mattered to the underwriting question.
Keep uncertainty visible.
Missing or contradictory evidence routes to review instead of disappearing inside a synthetic confidence score.
Signal landscape
Everyone has fragments. The question is who binds them into a defensible decision record.
Most players carry one or two of these axes and roll them into a score you cannot open. Kenshiki binds all six as inspectable evidence — what was checked, what boundary applied, what the source returned. The differentiator is the binding architecture, not exclusive data access.
| Axis | Kenshiki | SentiLink | Socure | LexisNexis Risk | Equifax · Experian · TransUnion | Plaid · MX · Finicity | The Work Number · Truv · Pinwheel | Prove · Boku · Telesign |
|---|---|---|---|---|---|---|---|---|
| Spatiotemporal presence Longitudinal location-pattern corroboration | Bound | — | — | — | — | — | — | — |
| Carrier-attested telecom SIM/number via CAMARA, consent-gated | Bound | Adjacent | Adjacent | Adjacent | — | — | — | Vendor signal |
| Life-context corroboration Household trajectory matched to the loan ask | Bound | — | — | Adjacent | — | — | — | — |
| Income at source Payroll-verified at origin | Bound | — | — | — | Adjacent | Adjacent | Vendor signal | — |
| Open-banking cash flow Real account history vs. thin or synthetic | Bound | — | — | — | Adjacent | Vendor signal | — | — |
| Identity-element trace SSN and history continuity over time | Bound | Vendor signal | Vendor signal | Vendor signal | Vendor signal | — | — | Adjacent |
As of June 2026. Based on publicly stated product capabilities. Ratings are qualitative. “Bound” means the signal is admitted into Kenshiki’s evidence record as inspectable evidence. “Vendor signal” means the player offers this as a core product capability. “Adjacent” means the player carries a related or partial capability. ”—” means this axis is not part of their stated offering at time of review.
Static map preview
The map is a product explanation, not a tracking surface.
The static preview compares a coherent pattern with an unsupported flat signal. It is illustrative only: no map service loads here, and no real applicant data is shown.
Static corroboration map preview
Green cells show a coherent source-side pattern. The isolated warning cell shows the kind of flat signal that needs review.
Home pattern
Residence continuity
Work routine
Source-side fit
Life rhythm
Context coherence
- Live pattern
- Multiple bounded source contexts cohere.
- Synthetic pattern
- Activity collapses to one unsupported claim.
Recurrence
A life is not just movement. It is movement that comes back.
At short timescales, ordinary movement can look almost like a Lévy flight: clustered local steps interrupted by occasional long jumps. The difference appears over longer windows. Home, sleep, work, care obligations, and routine return forces beat diffusion.
A true Lévy flight has heavy-tailed jump lengths and never converges; its territory keeps growing without bound. A lived trajectory grows, resets after lawful events, and then bends back toward a bounded envelope. That collapse from power-law expansion to logarithmic recurrence is the signal worth preserving.
This is why corroboration should not ask for a street-corner trail. It should ask whether source-side context shows recurrence, renewal, and ordinary constraints — the statistical shape of a life rather than a pile of coordinates.
Synthetic-risk patterns
Fabricated lives often fail where source-side context should cohere.
A synthetic identity can pass surface checks and still lack ordinary coherence. The product proof keeps those patterns explainable without turning them into secret scoring factors.
- Mailbox address
- A valid-looking address behaves like a receiving agency, not a residence with routine context.
- Impossible velocity
- Signals appear too far apart or too flat to describe one real person moving through ordinary life.
- Ring linkage
- Multiple applications share infrastructure that looks unrelated until the record links repeated patterns.
- Thin signature
- The file is clean because there is little corroborating context behind it, not because the person is clearly low risk.
Thin-file fit
Real missing history and fabricated identity can look similar on paper.
The product proof is built around that distinction. A real borrower can have little bureau-visible history but still show source-side continuity, income, residence, and obligation fit.
Corroboration is not more surveillance. It is a narrower way to decide whether the application story holds together.
FAQ
Common questions
What corroboration collects, what the static map preview proves, and how thin files differ from fabricated ones.
- Does corroboration require Kenshiki to store raw location data?
- No. The public product direction is source-side computation: approved environments answer bounded questions and return evidence, not raw applicant trails.
- What does the static map preview prove?
- It explains the product concept: coherent source-side context can distinguish a lived pattern from a flat or unsupported claim. It is not a production map or real applicant data.
- Can a thin-file applicant still be real?
- Yes. Thin credit history can coexist with real continuity, income, residence, and source-side context. Corroboration is meant to separate real missing history from fabricated identity risk.