Case 04 / Ledger Match
Fictional design exerciseA fictional reconciliation prototype pairs only unambiguous reference-and-amount matches across accounting and operations records, then routes the exceptions to human owners.
Deterministic prototype / invented batches
Amount-only matching can conceal a duplicate, a missing counterpart, or a different reference.
I built the deterministic matcher, exception contracts, rule tests, and proposed reviewer handoffs.
Never force a match when source records are ambiguous.
Two invented batches show unique reference-and-cent pairs and four open exception types, with source IDs and proposed owners.
Ambiguous reference? Keep the exception open.
TXN-205amount varianceOpenTXN-206ambiguous referenceOpenTXN-207missing feedOpenDuplicate and mismatched records are not forced into a pair.
Invented records; inspectable rules; human decisions.
Only a unique reference with equal cents is paired. Ambiguous or unmatched rows remain visible for investigation.
Excerpt from the deterministic rule output on an invented batch. No client ledger data is shown.The two invented batches show exact matches and four exception types. Targeted tests verify the matching rules; they do not validate accounting accuracy or client outcomes.
Disclosure Independently authored fictional exercise, separate from the cleared résumé reconciliation outcome. No client implementation or data is reproduced.
Explore the matching workflow
Switch between two invented batches and inspect the comparison. Each open item retains its source IDs, reason, proposed owner, and next human action.
Reference and exact cents drive the sample results. Reviewers can inspect every pair and exception below.
Ready to replay the fictional matching response.
One counterpart has the same normalized reference and exact cent amount.
Retain the paired source IDs as evidence; reviewer still owns close.
One counterpart has the same normalized reference and exact cent amount.
Retain the paired source IDs as evidence; reviewer still owns close.
Reference and exact cents drive the sample results. Reviewers can inspect every pair and exception below.
Ready to replay the fictional matching response.
One counterpart has the same normalized reference and exact cent amount.
Retain the paired source IDs as evidence; reviewer still owns close.
The reference agrees, but the amounts differ by an exact-cent comparison.
Investigate the source values and document the variance before close.
More than one record shares this reference; an amount match cannot choose the correct counterpart.
Confirm the source records and resolve the duplicate before matching.
No operations-feed row has this reference. Equal amounts under other references are not auto-matched.
Check timing and source completeness; do not create an adjustment from this demo.
This feed row has no ledger reference, even if its amount equals another open item.
Confirm whether the feed item belongs in this period and ledger scope.
THE PROBLEM
When a ledger and an operational feed are compared, a single “matched” label hides the path to that result. A reviewer needs the record pair, the rule used, and a clear queue for items that cannot safely be paired. This exercise makes those handoffs visible with invented records.
FICTIONAL DISCOVERY BRIEF
This is a constructed scenario, not client discovery. The imagined ledger owner needs complete source coverage, the feed owner needs to diagnose missing or duplicated records, and a reviewer needs defensible evidence before closing an item.
Accept one normalized reference with one equal-cent counterpart. Anything ambiguous stays open.
Show a quiet day and an exception day; no bank data, uploads, posting, or source-system connection.
Surface a variance, duplicate, missing counterpart, and extra feed item with source IDs and owners.
A real pilot would first need approved source contracts, accounting rules, period cutoffs, role-based access, retention, and independent validation.
ARCHITECTURE DECISIONS
The demo trims and uppercases references and compares integer cents exactly. It never pairs rows merely because amounts agree, and it does not model netting, split entries, FX, or timing.
Every result names the source IDs, reason, proposed owner, and next human action. Duplicate references stay ambiguous even when amounts match.
The fixed response is advisory. The browser has no ledger write, journal entry, identity, durable decision history, or production close action.
ACCEPTANCE EXAMPLES
The quiet batch yields two exact pairs and no open items. The exception batch yields one pair and four open outcomes: amount variance, duplicate reference, missing feed counterpart, and extra feed row. The extra row carries the same amount as the missing counterpart; the matcher still keeps both open because their references differ.
Four targeted unit tests check those outcomes, reference normalization, and duplicate-ledger handling. They do not validate accounting correctness beyond these fictional rules.
RELATED EXPERIENCE / SEPARATE EVIDENCE
My public résumé separately summarizes a cleared, high-level reconciliation outcome from professional work. This independently authored exercise does not reproduce that client implementation or prove its time savings or financial result.
Read the résuméWHAT THIS DOES AND DOES NOT PROVE
The working rules distinguish a clean pair from several reasons to investigate. They do not establish production reconciliation accuracy, financial impact, compliance, security, or readiness for live data. Those claims require a cleared, separately evaluated system.
Open the focused demoDecision notebook
An unresolved exception is a useful output when it preserves the source records and the reason to investigate.
Bring close exceptions into one reviewer queue.