Methodology
Three engines, working together.
Quarare turns evidence into an expert report through three signature capabilities. Each is grounded in established law and practice. None disclose proprietary scoring weights. All recompute live as the analyst works.
Engine 1
The Daubert Score™
Your conclusion's defensibility, scored continuously.
The Daubert Score™ is a continuous 0–100 rating of a Finding's defensibility under the four Daubert factors. As the analyst writes, the score recomputes within roughly two seconds of an edit pause. Color-coded green at 80 and above, amber from 60 to 79, red below 60.
Each factor is scored 0–25 and summed. Every factor carries its own reasoning, surfaced in a panel beside the Finding editor. When a factor scores low, Quarare proposes specific improvements — a citation to add, a methodology claim to qualify, a measured phrasing to replace a conclusory one.
Methodology reproducible by independent expert.
Approach cited in Journal of Forensic Accounting (Bremser, 2019).
Source reliability variance disclosed (58–98 range).
Standard ledger reconstruction methodology.
The signature interaction is the way phrasing moves the score. Conclusory language — “X was fraudulent” — scores lower than measured language — “X appears to lack proper authorization based on the records reviewed.” The score teaches defensibility in real time. Analysts internalize the discipline without a separate training pass.
Grounded in the Federal Rules of Evidence and Daubert v. Merrell Dow Pharmaceuticals, Inc. (1993). Proprietary scoring weights are not disclosed.
Engine 2
The Records Reliability Score
Every source, scored before analysis.
The Records Reliability Score answers a single question every expert eventually gets asked: how much weight should this document carry? Quarare scores each source 0–100 at ingestion across six factors, weighted to a composite.
The six factors
- Provenance. Primary source, secondary source, or derived. The lineage from original generator to current form.
- Authentication. Verifiable origin — signature, seal, agency stamp, hash, custody chain.
- Chain of custody. Known custodians from creation to current possession. Gaps surface here.
- Contemporaneity. Was the document created at the time of the events it records, or reconstructed later?
- Completeness. Full document, intact pages, no redaction obscuring material content.
- Corroboration. Supported by other documents in the corpus, or standing alone.
The composite surfaces in the document inventory and in the document detail view. Analysts can override factor scores when their expert judgment differs, with a required justification that is logged in the audit trail. Reliability scores flow downstream into Daubert factor analysis — a Finding built on documents averaging 58 will not score the same as one built on documents averaging 92.
Factor mapping is preserved in the Quarare specification. Proprietary weights are not disclosed.
Engine 3
Category Insights
AI-surfaced patterns, promoted to Findings on your terms.
Category Insights bridge the gap between a fully-ingested corpus and the formal Findings that will appear in the report. An Insight is a hypothesis: it looks like X is happening. A Finding is a substantiated opinion: X is happening, and here is why. Quarare generates the first. The analyst evaluates which deserve to become the second.
Insights are generated automatically after the initial document ingestion completes, and re-triggered on request when new evidence arrives. Each one carries a headline, the supporting documents, the pattern category, a confidence indicator, and the Daubert factor relevance if the analyst promotes it.
Five pattern categories
- Anomalies. Events outside expected patterns.
- Inconsistencies. Contradictions between documents.
- Omissions. Events expected but not documented.
- Concentrations. Clustering of activity around entities or dates.
- Sequences. Chains of events that suggest causation.
Promotion is deliberate. Insights live in their own surface, separate from Findings, until an analyst reviews the supporting documents and decides the hypothesis warrants formal investigation. The AI surfaces; the expert decides.
Prompt structure and confidence calibration are proprietary and not disclosed.
Developed in collaboration with practicing forensic accountants. Methodology grounded in the Federal Rules of Evidence and Daubert v. Merrell Dow Pharmaceuticals, Inc. (1993).