Digital asset data-quality checklist
====================================

Review identities, units, observation times, incomplete coverage, and the evidence behind every total.

1. State the reporting boundary
List the networks, accounts, resources, and time period covered by the result. Identify excluded or unavailable sources. A partially refreshed account set should not be labeled complete merely because the successfully refreshed accounts returned valid data. Preserve coverage status close to any displayed aggregate.

2. Verify asset identity
Match a network-aware asset reference before comparing or joining amounts. Keep symbols, names, and icons as presentation metadata rather than unique keys. Preserve the original provider identifiers and use explicit mapping rules for unresolved assets, migrations, wrapped representations, and conflicting metadata.

3. Preserve quantity and scale
Carry exact raw quantities and an explicit decimal scale where required. Format only for display, using appropriate arithmetic in the consumer. A missing scale should be unresolved rather than automatically assumed. Keep a token quantity separate from any fiat valuation and its associated methodology.

4. Distinguish the clocks
Record event time, observation time, and fetch time where they answer different questions. A newly loaded page can still show an older observation. Put the relevant date near the value and retain the block or source reference necessary to reproduce the measurement.

5. Represent missing data honestly
Differentiate an empty result from an unavailable provider, unsupported resource, failed authorization, and incomplete history. Do not convert a failure into zero. When a quantity is known but its price is missing, keep the quantity visible and label the valuation unavailable.

6. Test history and duplication
Exercise cursor boundaries, repeated events, out-of-order observations, and imports interrupted during pagination. Make the intended processing effect repeatable without double counting. Confirm that new arrivals do not silently change an import boundary or cause existing records to be omitted.

7. Separate observation from conclusion
Keep the source record, normalization, valuation choice, adjustment, and published result distinguishable. A manual review should leave a reason and evidence. Revisions should preserve the earlier version so an operator can explain what changed rather than seeing only today’s final number.

8. Prove recovery and qualification
Simulate a missing account, delayed event, corrected source, and unavailable dependency. Verify how the gap is detected, which reports become qualified, and how the application reconstructs coverage without duplicate effects. Assign an owner to unresolved exceptions and document when recovery requires additional evidence.

Reference design checklist, not certification.
Read the full guide: https://digitalassetapi.com/blog/digital-asset-api-guide/