Design for temporal data correctness.
Ensure information remains accurate across time and state changes.
Contents 7. Trust, Safety & Responsibility 4. Data Integrity
Whether data can be trusted.
First, ask yourself
“Can users of your system trust their data is safe, accurate, and handled as expected?”
Mission statement
Ensure users trust outcomes by preserving accuracy and handling data reliably.
Key heuristics Data Integrity
- Information must remain accurate across time and state changes.
- Primary records must not be silently overwritten by derived values.
- Data must be protected against duplication, omission, and misapplication.
- State changes must be atomic or reversible.
- Data correctness must persist through retries and interruptions.
- Automated processes must maintain data accuracy and wholeness at scale.
- Users must be able to inspect and confirm data correctness.
- Accumulating data inconsistencies must be detected and addressed proactively.
- Correctness must persist without constant manual monitoring.
Executive brief
The system must preserve data integrity even when users are not watching.
Core questions Data Integrity
“Does the system keep user data complete and accurate over time and during failures?”
“Is data reliable even when the user is absent?”
Focus areas
Custody-first
“What happens to users’ data when they leave the system?”
“Who is responsible for maintaining it at all times?”
Failure-aware
“Does data survive crashes, retries, and partial operations?”
“Are edge cases treated as first-class scenarios?”
Change-aware
“Does data remain valid through updates and migrations?”
“Are transformations reversible or auditable?”
AI-aware
“Are inferred or generated data clearly marked?”
“Does learning ever overwrite the original source of truth?”
UX directives Data Integrity
Protect primary source data.
Prevent silent overwriting of original records by derived values.
Safeguard completeness and scope.
Detect and prevent duplication, omission, or misapplication of data.
Ensure atomic or reversible state changes.
Do not expose users to unstable intermediate states.
Protect data during partial failures.
Design systems to maintain correctness through retries and interruptions.
Enforce integrity in automated processes.
Ensure scale and automation preserve correctness and reliability.
Provide mechanisms for data verification.
Allow users to inspect and confirm accuracy.
Detect and remediate accumulating inconsistencies.
Monitor for cascading data errors proactively.
Design integrity to operate without constant oversight.
Ensure correctness persists without manual monitoring.
Executive summary
- Data Integrity is preserved correctness over time, not temporary accuracy.
- It ensures information remains complete, consistent, and reliable across state changes and automation.
- The system must protect primary data, prevent silent corruption, and avoid unstable intermediate states.
- Integrity mechanisms must withstand interruption, scale, and partial failure without manual oversight.
- Verification and remediation must be built in, not retrofitted after error accumulation.
- Data Integrity succeeds when correctness persists predictably without requiring constant user vigilance.
Success indicators
- Data remains accurate and consistent across actions and time.
- Original records are protected from unintended overwriting.
- Data is not duplicated, lost, or applied incorrectly.
- System changes occur reliably without unstable intermediate states.
- Users can inspect and verify the correctness of important data.
One-line summary
Before users can trust outcomes, they must trust how their data is handled.