Strong Match
Identity is verified and seller information aligns strongly with ownership information.
System recommendsRecommend proceeding
Human action: Normal brokerage approval process.
Product Requirements + Risk Decisioning
Listing Fraud Pre-Check: Human-in-the-Loop Risk Decisioning MVP
Turning disconnected identity, ownership, and authority information into a structured risk recommendation—with human review built in.
- Status
- Working MVP / Pilot ready · Synthetic-data public demo
- My role
- Product concept, workflow architecture, decisioning design, requirements, AI-assisted MVP development, pricing & pilot strategy
- Focus
- Product requirements · Risk decisioning · Human review
FIG. 01Product model / verification record
Verification record
- Identity verification Complete
- Ownership record Retrieved
- Seller-to-owner match Partial
- Authority documentation Review required
SYSTEM RECOMMENDATION
Partial Match
Authority review required
The system flags an incomplete relationship. A broker or administrator reviews supporting documents and records the final disposition.
The problem starts before the listing.
Seller impersonation can begin when someone first approaches an agent claiming the right to sell. By the time the listing is marketed, the brokerage may already have invested in agreements, pricing, MLS entry, and showings.
Identity, the owner of record, and the authority to act are separate questions. The product brings those signals into one reviewable workflow for agents, broker/admin teams, MLS or association reviewers, sellers, and entity representatives.
The product question
Can a brokerage build reasonable confidence in a seller’s identity and relationship to a property before marketing the listing?
Can a brokerage build reasonable confidence in a seller’s identity and relationship to a property before marketing the listing?
One workflow. Explicit decision points.
The requirements sequence the checks, preserve uncertainty, and give each result an operational next step.
FIG. 02End-to-end verification workflow
- 01Seller intakeCreate the verification record.
- 02Identity verificationCheck identity before proceeding.
- 03Ownership checkRetrieve records after the identity threshold.
- 04Seller-to-owner matchCompare identity and ownership data.
- 05Entity / authority reviewRoute trusts, estates, and entity cases.
- 06Risk outcomeClassify and recommend.
- 07Audit trailRetain evidence, review, and disposition.
Identity first, property data second. Failed identity checks stop downstream processing. The next stage runs only when the prior stage meets the required threshold.
Decisioning architecture
An identity provider and a property-data provider each answer part of the question. The product translates their summarized evidence into four operational outcomes.
FIG. 03Four-outcome decision matrix
Partial Match
Identity is verified, but the ownership relationship is incomplete, indirect, or ambiguous.
System recommendsFlag for review
Human action: Evaluate authority documentation and record the decision.
No Match
Verified identity does not reasonably correspond with ownership or the claimed authority.
System recommendsRecommend escalation / do not proceed
Human action: Resolve the discrepancy before continuing.
Unable to Verify
Required identity, property, or supporting data cannot produce a reliable result.
System recommendsRequest documentation
Human action: Perform manual verification; preserve the uncertainty.
Human review is part of the product.
FIG. 04Human–system responsibility model
System recommends
- Checks which verification stages completed
- Compares seller and owner information
- Classifies risk and suggests next actions
- Routes exceptions and logs supporting evidence
Shared
decision record
decision record
Broker / admin decides
- Reviews exceptions and supporting documents
- Validates authority for entity representatives
- Resolves or escalates discrepancies
- Records the final listing disposition
Privacy and data controls
Data minimization
Retain what is needed to operate and explain the workflow.
No unnecessary raw ID data
Avoid retaining complete identity documents or raw date-of-birth data when a summarized result is sufficient.
Server-side vendor calls
Keep sensitive vendor interactions and credentials behind server-side services.
Result summaries
Store actionable verification states instead of duplicating full vendor datasets.
Role-based access
Separate normal-user and administrative functions and limit sensitive information by role.
Auditable decisions
Record completed checks, summarized results, reviewer actions, notes, and final disposition.
A focused MVP
I scoped the build around one hypothesis: can identity and ownership signals become an understandable pre-listing risk recommendation that operators can use?
- Seller intake and verification requests
- Sequenced identity and ownership checks
- Seller-to-owner comparison
- Entity and authority review routing
- Four-state risk recommendations
- Admin review and structured audit records
- Concise result reports and role-based access
- Pricing, unit economics, and pilot planning
MVP and public-demo boundary. The case-study MVP uses Next.js / React, Supabase, role-based authentication, and server-side integration patterns. The public demo is a separate, self-contained experience with deterministic sample scenarios. It transmits no real consumer identity or property information and makes no live identity, DMV/DLDV, ATTOM, property-record, or other third-party calls.
Pilot metrics, defined before rollout
The next step is a controlled pilot with a defined set of listings at a brokerage, MLS, REALTOR association, or multi-office organization. These are proposed measures, not reported pilot results.
01 / MEASURETime to verify
02 / MEASUREClean approval rate
03 / MEASUREManual review rate
04 / MEASUREEscalated questionable listings
05 / MEASUREVerification completion rate
06 / MEASUREAgent and reviewer adoption
07 / MEASURECost per verification
What this demonstrates
From operational ambiguity to a testable product.
A working MVP brings together product requirements, decision logic, human-review rules, privacy controls, and a structured audit trail. It demonstrates how I turn a complex risk workflow into something teams can evaluate and pilot.