Record the source
Preserve the document, page, worksheet or cell reference alongside the reported figure.
Method / comparison rules / boundaries
Evidence first.
Interpretation stays human.
DealTie is designed to make selected financial observations comparable and traceable across a transaction document pack. It does not decide which source is correct. It makes the difference inspectable so a qualified reviewer can resolve it.
01 / Evidence model
One observation. One source trail.
Every observation should carry a metric label, reporting period, numeric value, stated unit, document identity and source location. Extracted candidates are not treated as confirmed facts until an analyst checks them against the source.
The comparison is only as meaningful as the mapping. If two figures refer to different periods, definitions, consolidation perimeters or accounting treatments, a numeric comparison alone cannot make them equivalent.
Check the meaning
Confirm metric, period, unit and definition before placing observations in one comparison group.
Expose the difference
Apply explicit deterministic rules and retain the inputs that produced each classification.
02 / Classification logic
What the workbench can flag.
The current prototype applies narrow, deterministic checks. Review priority orders the analyst's queue; it is not a financial materiality, risk or loss score.
Value mismatch
Comparable, analyst-confirmed observations differ beyond the configured numeric tolerance. The report shows the values and observed spread.
Unit mismatch
Observations use different stated units. The prototype does not silently convert currencies, scales or measurement conventions.
Insufficient evidence
A metric lacks the required source coverage or contains missing or non-numeric observations. The gap remains visible.
Tied
All supplied observations agree within the configured tolerance. This means the supplied values tie—not that the underlying accounting is correct.
Important: A candidate discrepancy is not a confirmed error, misstatement, loss, overpayment or recoverable amount. No source is assumed authoritative. A qualified finance professional must resolve substantive questions.
03 / Acceptance and handoff
A review should be testable, not theatrical.
Agreed comparison set
Before work begins, define up to 15 metrics, their periods, source documents, units and comparison tolerance.
Evidence register
Each finding should expose the underlying observations, source references, comparison values and calculation explanation.
Open questions
Unresolved mappings, missing evidence, limitations and reviewer decisions remain explicit rather than being guessed.
Correction pass
One correction pass checks the agreed deliverable against the agreed acceptance criteria. It is not unlimited diligence or assurance.
04 / Product boundary
Built for a narrow review task.
The prototype does not provide scanned-PDF OCR, semantic metric matching, currency or period conversion, workbook formula evaluation, accounting judgement, source hierarchy inference, secure deal-room hosting, multi-user permissions or audit-grade history. Do not upload confidential deal files to the public prototype.
Request a non-confidential discussion ↗