# Field-change review card — filled example and blank review All records and speakers are fictional. This is a non-writing expected-result sheet, not an executed schema change or an Overskill product test. Nothing in this packet connects to a database or applies the proposal. ## Filled example Proposed field: estimated_visit_minutes, optional nonnegative integral number. Null means not established under this example's policy. The six old records initially have no such key; the proposed field-added stage represents all six as null. That representation is an acceptance target, not proof of a native migration. | Job | Authored evidence | Proposed value | Interpretation | | --- | --- | --- | --- | | J101 | E101 explicitly estimates 30 minutes | 30 | Known estimate | | J102 | E102 explicitly confirms no on-site visit | 0 | Known zero | | J103 | E103 records no reply | null | Unknown | | J104 | E104A says 30; E104B says 45 | null | Hold conflict; no averaging | | J105 | E105 says quick | null | Hold invalid type; no coercion | | J106 | E106 records an empty response | null | Unknown; not zero | Known coverage is 2/6. Their subtotal is 30 minutes. The total for all six remains unknown. Keep the original IDs, client links, labels, notes, whitespace and tags. The comparison files include every row, not just the two populated values. Every compatibility case starts independently from this backfill baseline. An old label-only update should retain J101's 30. An explicit clear should set J101 to null, leaving coverage 1/6 and known subtotal 0 (one confirmed zero, five unresolved). Adding J107 with numeric20 produces coverage3/7 and subtotal50. Reject a negative new estimate without adding or changing a row. An optional unanswered J107 produces coverage2/7 and subtotal30. A required/default-zero design is held if it cannot represent unresolved answers. These are desired app-intent rules. An omitted field and an explicit null may behave differently across update formats; the resulting app's actual request, stored result and old screens need testing. Do not infer transport semantics from these JSON files. ## Blank review - Existing app/table and exact revision: ______ - Authorized isolated test surface and evidence of isolation: ______ (Unknown is not isolated.) - New field, unit, allowed type, unanswered meaning and explicit-zero meaning: ______ - Existing record IDs/count and complete original-value snapshot: ______ - Evidence source for each proposed non-null value; conflicts/unknowns retained: ______ - Known numerator / full denominator / known subtotal / all-record total status: ______ - Exact before/after diff and untouched relationships: ______ - Old screen omits field; explicit clear; valid new row; invalid new row; optional unanswered row: expected ______ / actual **Not tested**. - Required/default decision and unresolved cases: ______ - Reopen/readback and independent comparison result: **Not tested**. Requested next work: review the field meaning, unresolved evidence and non-writing comparison first. Preserve existing rows and related identities; specify and test the old/new screen behavior before relying on the field. Keep native add-field steps with the existing help owner. Preview is not proof of a separate database, and an export is not proof of complete restoration. ## Review the complete worked records These are authored expectations, not an executed migration or observed app results. - [Original six jobs and three clients](visit-jobs-before.json): the new field is absent. - [Proposed optional-field stage](visit-field-added-baseline.json): all six estimates are null; 0/6 established; subtotal unavailable; all-job total unknown. - [Evidence-supported proposal](visit-field-proposal.json): 2/6 established; known subtotal 30 minutes; all-job total unknown. - [Seven fictional evidence records](visit-estimate-evidence.csv). CSV cells contain text; explicit confirmation, units and context support the proposed numeric values. This is not a general import conversion rule. - [Six independent screen cases and their complete expected records](independent-screen-cases.json). ### Expected coverage after each independent case Each case restarts from the six-job evidence-supported proposal. Actual application result: **Not tested** for every case. | Case | Established estimates | Known subtotal | Unresolved | All-job total | Actual app result | | --- | --- | --- | --- | --- | --- | | C01: old screen omits field | 2/6 | 30 minutes | 4 | Unknown | Not tested | | C02: deliberately clear J101 | 1/6 | 0 minutes, from confirmed J102 | 5 | Unknown | Not tested | | C03: new J107 has 20 minutes | 3/7 | 50 minutes | 4 | Unknown | Not tested | | C04: reject negative estimate | 2/6 | 30 minutes | 4 | Unknown | Not tested | | C05: new J107 is unanswered | 2/7 | 30 minutes | 5 | Unknown | Not tested | | C06: hold default-zero redesign | 2/6 | 30 minutes | 4 | Unknown | Not tested | The all-null stage has no established estimate, so its subtotal is unavailable. C02 has a known subtotal of zero because J102 remains explicitly confirmed; it still leaves five unanswered estimates.