Preserve unavailable values in operational count columns — what should the operator record?
A zero is a value, while unavailable describes a limit on what the evidence supplies. Mixing them can distort operational summaries even before substantive interpretation begins. Define how each field represents missingness and test the whole transformation path, including spreadsheet imports that may quietly substitute default values for empty cells.
Prevent missing metrics from becoming false zeroes during table cleanup or aggregation.
Inventory absent-property, explicit-null, empty-string and numeric-zero cases in permitted input. Record source meaning where documented and keep unresolved cases separate rather than assuming every empty-looking cell carries the same meaning across result kinds and fields.
Add field definitions explaining unavailable, not-applicable and observed-zero states. A recipient should be able to tell why a value is absent without guessing whether the table failed to import it or the source actually reported nothing.Inspect actual states
Inventory absent-property, explicit-null, empty-string and numeric-zero cases in permitted input. Record source meaning where documented and keep unresolved cases separate rather than assuming every empty-looking cell carries the same meaning across result kinds and fields.
Define typed output states
Use value plus value_state when the table needs a numeric column. A numeric zero should be retained only when actually returned or validly derived under an explicit rule; unknown values should not enter arithmetic as zero by default.
Test default hazards
Use fixtures with zero, null and missing fields through the actual parser and aggregate function. Verify the count of known values remains separate from the total row count and that empty numeric cells do not become a fabricated measured value.
Review existing derivatives
Inspect sums, differences and ratios for fields that previously used default zeroes. Correct only artifacts supported by the observed conversion error and preserve their revised computation basis rather than asserting every earlier research result was invalid.
Release missingness semantics
Add field definitions explaining unavailable, not-applicable and observed-zero states. A recipient should be able to tell why a value is absent without guessing whether the table failed to import it or the source actually reported nothing.
Keep the boundaries clear.
- Prevent missing metrics from becoming false zeroes during table cleanup or aggregation.
- Operational record: Inspect actual states. Context badges do not certify access.
Start with a read-only check: ask your agent to report actual source coverage before collecting records or exporting an audience.
Primary references
Documentation reviewed 2026-11-15. Product capabilities, client availability and plans can change. Third-party marks identify their products, not an affiliation.