Incomplete or conflicting data: design a workflow that knows when to stop
A six-step method to detect exceptions, protect sensitive actions and request a human decision with the right context.
A reliable workflow does not hide uncertainty. It recognises missing, invalid, stale or conflicting information, preserves useful evidence and stops the action when the required quality is not met.
Define the minimum data contract
List what must be present before any action: required fields, format, freshness, source and quality level. Business and technical teams should understand this contract.
- Required fields and formats
- Acceptable freshness
- Expected source
Distinguish anomaly types
Missing, malformed, stale and conflicting values do not require the same response. Classify the exception before deciding what comes next.
- Missing or incomplete
- Invalid or stale
- Conflicting
Set source precedence without hiding conflict
Define authoritative sources and their order of precedence. If two authorised sources remain incompatible, expose the disagreement instead of resolving it silently.
- Authoritative source
- Documented precedence rule
- Conflict preserved and visible
Link every threshold to a safe action
A score or quality check only helps when it leads to a clear response: continue, prepare a draft, request data, escalate or stop.
- Continue
- Request or prepare
- Escalate or stop
Ask a person with the right context
A decision request should explain what is missing, which sources disagree, what evidence exists, the deadline and possible choices. The reviewer should not have to rebuild the case.
- Clearly stated gap
- Linked evidence and sources
- Choices, impact and deadline
Measure exceptions and improve the process
Track recurring anomalies, resolution time and stops. This evidence helps fix the source, rule or process instead of adding workarounds.
- Rate and recurring causes
- Resolution time
- Updated rules
NOVAMIND
A good workflow does not pretend to know
It protects the action, exposes uncertainty and hands over an actionable exception. This discipline reduces errors, clarifies ownership and steadily improves data quality.
General methodology: adapt thresholds, sources, ownership, retention and mandatory stops to your sector, country, security requirements and compliance obligations.