Trust center
A serious buyer product is built on evidence, not mystery.
Twellie uses deterministic valuation and readiness logic, limits AI to unverified photo observations, publishes evidence ranges without attaching an unearned probability, discloses data provenance, and refuses price ceilings until current closed-sale benchmarks pass the internal sample, error, coverage, and range-width policy.
Core logic
Rules before words
Readiness status, evidence estimates, ranges, and any permitted ceilings come from stored deterministic outputs. Explanatory text cannot override the gate.
Accuracy policy
No fake percentage
We do not publish a headline accuracy claim merely because a benchmark exists; its error, interval coverage, and interval width must also be disclosed.
Report proof
Source status stays visible
Material inputs are labelled as supported, partial, missing, conflicting, demo, or needing verification when that status is known.
User promise
Explicit readiness
The report withholds price ceilings when identity, comparable-sale evidence, the current ask, the buyer's own limit, or the closed-sale calibration record does not pass the gate.
Reliability roadmap
The operating plan
01
Data moat
Unify live listings, closed comps, public records, tax, permits, hazards, photos, insurance signals, and buyer context into one normalized property graph.
02
Calibration engine
Versioned benchmarks test evidence estimates and interval coverage against unseen closed sales by market, home type, price band, and data quality.
03
Decision intelligence
Convert available evidence into an explicit readiness state, unresolved tasks, and conditional ceilings only when every required gate passes.
04
Proof trail
Every report stores logic version, model IDs, data-source status, comp weights, adjustment reasons, and generated-output timestamps.
05
Workflow capture
Attach verified inspection findings, written quotes, title documents, insurance evidence, lender terms, and closing records to the same property dossier.
06
Distribution
Win trust through public benchmarks, sample reports, buyer education, and partnerships with lenders, inspectors, attorneys, and buyer communities.
Accuracy transparency
What we will publish before claiming accuracy
Benchmark sample
Recent closed sales only; no stale deed prices or active listings.
Coverage split
City, ZIP, property type, price band, comp density, and data-source mode.
Metrics
Median absolute error, p90 absolute error, bias, evidence-range coverage, and median range width.
Failure cases
Rural homes, unusual architecture, thin comps, distressed sales, major unseen renovation.
Trust gates
What must be true before scale
500+ recent closed-sale rows in each launch market before local accuracy copy goes live.
Every report exposes live, hybrid, or demo data status in the header.
Every comp adjustment is visible, weighted, and explainable.
Price ceilings remain unavailable unless live core evidence and the buyer-supplied affordability limit pass.
Every explanation is grounded in stored valuation outputs, not free-form opinion.
Plain-English promise
The report should make a buyer harder to mislead.
That means visible comp evidence, explicit source gaps, known-versus-missing ownership costs, and no walk-away number when the report cannot support one.