What is an exception?
An exception is created when a transaction from one source has no valid counterpart in another source. Common causes include:
- No candidate found: No transaction in the other source meets the active rule criteria.
- Below confidence threshold: Candidates exist, but score below the minimum confidence (default: 60).
- Duplicate rejection: A previous match was rejected and no alternative candidate remains.
- Source imbalance: One source contains transactions that are missing from the other.
Exception lifecycle
Exceptions move through a simple workflow:
- When Matcher can’t reconcile a transaction, it creates an exception in
OPENstatus. - From there, the exception is assigned to an analyst for investigation (
ASSIGNED). - If the resolution depends on an external system — such as a dispatched JIRA issue or webhook callback — the exception moves to
PENDING_RESOLUTIONuntil the external response arrives. - Once the analyst resolves the exception (force match, adjustment, or external callback), it transitions to
RESOLVED.

Status definitions
| Status | Description | Who Can Transition |
|---|---|---|
OPEN | New exception waiting for assignment | System |
ASSIGNED | Assigned to an analyst for investigation | System, Analyst |
PENDING_RESOLUTION | Awaiting external response (JIRA, webhook callback) | System |
RESOLVED | Closed with an auditable resolution | Analyst, System |
Exception severity
Matcher classifies exceptions by severity so you can work the queue in the right order.
| Severity | Criteria | SLA |
|---|---|---|
| Critical | Amount >= 100,000 OR age >= 120 hours | 24 hours |
| High | Amount >= 10,000 OR age >= 72 hours | 72 hours |
| Medium | Amount >= 1,000 OR age >= 24 hours | 120 hours (5 days) |
| Low | All others | 168 hours (7 days) |
Severity escalation
Severity is re-evaluated as an exception ages. The classification uses OR logic — either the amount or the age threshold is enough to trigger a higher severity:- An exception under 1,000 starts as Low, but escalates to Medium after 24 hours.
- An exception under 10,000 escalates to High after 72 hours.
- Any unresolved exception escalates to Critical after 120 hours.
Resolution methods
You can resolve an exception in four ways.
1. Force match
Manually link transactions when you’ve confirmed they belong together, but the system couldn’t match them. Use Force Match when:- The correct counterpart exists, but variances blocked automatic matching.
- You can clearly explain and document the rationale.
- The variance is expected (fees, timing, rounding).
2. Create adjustment
Create an adjusting entry to account for a variance or to balance an unmatched item. Common adjustment types:| Type | Use Case |
|---|---|
BANK_FEE | Bank charges not recorded in the ledger |
FX_VARIANCE | Currency conversion differences |
TIMING_DIFFERENCE | Settlement timing adjustments |
ROUNDING | Small rounding differences |
CORRECTION | Error corrections |
OTHER | Other documented variances |
3. Write-off
Write off a transaction that has no valid counterpart. This should be rare and typically requires approval. Write-off reasons:| Reason | Description |
|---|---|
DUPLICATE_ENTRY | Transaction was entered twice |
CANCELLED | Transaction was reversed or cancelled |
NOT_APPLICABLE | Doesn’t belong in this reconciliation scope |
BELOW_THRESHOLD | Amount is below the investigation threshold |
APPROVED_VARIANCE | Variance approved by management |
4. Split transaction
Use split when one transaction should match multiple counterparts.Audit requirements
Every resolution creates an audit record. Some resolution types require stronger evidence and approvals.
Required documentation by resolution type
| Resolution | Required Fields | Approval Needed |
|---|---|---|
| Force Match | reason, notes | No (unless amount > threshold) |
| Adjustment | adjustment_type, amount, description | If amount > $1,000 |
| Write-Off | reason, notes, reference_document | Always |
| Split | splits[] with amounts and targets | No |
Bulk operations
When dealing with large volumes of exceptions, bulk endpoints let you process up to 100 exceptions in a single request.
Bulk assign
Assign multiple exceptions to a team member at once:cURL
Bulk resolve
Resolve multiple exceptions with a shared resolution:cURL
succeeded and failed arrays, so you can handle partial failures gracefully.
Bulk dispatch
Dispatch multiple exceptions to an external system:cURL
Exception comments
Comments provide an audit trail of team discussions and investigation notes on exceptions.
Add a comment
cURL
List comments
cURL
Disputes
When an exception requires formal investigation or involves an external party, escalate it to a dispute. Disputes track evidence, state changes, and resolution outcomes.
List disputes
cURL
Retrieve a dispute
cURL
DRAFT → OPEN → PENDING_EVIDENCE → WON or LOST.
Exception resolution workflow
Use this flow to keep reviews consistent and audit-friendly.
Investigate
Use the exception payload to understand what failed and what candidates exist.
- Read
reason_detailsto see why matching failed. - Review
candidatesfor near matches below threshold. - Look for patterns (same counterparty, recurring reference formats).
Resolve
Pick the resolution that best reflects reality and policy.
- Force Match: You found the correct counterpart.
- Adjust: You need an adjusting entry for variance.
- Split: One transaction maps to multiple counterparts.
- Write-Off: No counterpart exists and policy allows it (approval required).
Document
Capture enough detail for someone else to replay your decision later:
- What you checked
- What you concluded
- Links or IDs for supporting evidence
Best practices
Work by severity and SLA
Work by severity and SLA
Start with Critical and High items. They carry the highest risk and the tightest deadlines.
Make decisions auditable
Make decisions auditable
Notes aren’t optional. Treat them as part of the resolution:
- What you checked
- Why this resolution is correct
- Any ticket IDs, statements, or confirmations
Fix patterns at the source
Fix patterns at the source
Repeating exceptions usually point to configuration issues:
- Same counterparty → Normalize names or mapping
- Same date window → Validate ingestion completeness
- Same source → Review field mapping and sign conventions
Treat force matches as exceptions to the rule
Treat force matches as exceptions to the rule
If you force-match regularly, your rules or tolerances need attention.
Route work automatically
Route work automatically
Use assignment rules to reduce triage time and keep ownership clear.
Write-offs are policy calls
Write-offs are policy calls
If you write off often, revisit thresholds, context scope, or upstream data quality.

