Make the review decision govern the actual analysis extract

Define the meaning and consequence of each review state

List the states the project actually uses and the evidence required to assign them. Separate received, awaiting clarification, approved and excluded from analysis where those distinctions matter. An unresolved data issue is not a finding of participant misconduct. Specify the authorized reviewer and the reason record needed for a consequential decision.

For each intended analysis product, state which states are included at its cutoff. A fieldwork progress report may count unresolved submissions while a finalized analysis uses only adjudicated records. These can both be legitimate if their purposes are explicit. Do not let a reassuring status name substitute for an agreed membership rule, and do not assume every analyst will infer the same rule from the review interface.

Prove that the extraction applies the intended decision

ODK Central states that review labels do not automatically exclude submissions from exports; filtering is required. It also sets a submission’s state to Edited after an edit. Those behaviors make export membership and amendment handling separate acceptance questions.

Create a synthetic set containing one record in each relevant state and an expected membership list for the planned extract. Have the authorized data owner run the ordinary extraction route. Compare actual record identities with the expected set, including exclusions. A matching total is insufficient if one excluded record replaces one approved record. Keep the test limited to permitted fixtures rather than changing live participant records to demonstrate the mechanism.

Decide what happens after an accepted record changes

Write the conditions under which an amendment requires renewed review. A correction can change the interpretation of a record, so the prior approval should not silently cover every later version. Identify how the review owner and dataset owner will learn that an included record changed. Preserve the original evidence and reason through the approved process rather than overwriting the only account of what was collected.

Test a synthetic amendment and inspect both the review queue and the next extract. Confirm whether the intended updated version is included, withheld or flagged according to the agreed rule. Do not assume the receiving analysis automatically refreshes because the source system changed. If a report has already been issued, route a material change through its correction process instead of quietly replacing the underlying file.

Freeze the extract with its review boundary

Record the extraction time, applied membership rule and unresolved review cases. Explain whether later approvals will produce a revised dataset or belong to a later release. Keep the counts of received, reviewed and included records distinguishable. This lets a reviewer reconcile the analysis population without needing access to every sensitive comment attached to a submission.

Repeat the acceptance fixture when a connector, saved filter or review-state policy changes. Keep the definition of exclusion with the research owner and the implementation evidence with the data owner. A technical filter can faithfully implement a poor methodological decision; it does not make that decision valid. The completed contract establishes that the approved decision actually reaches the dataset, while keeping substantive research judgment and participant treatment separate.

Make the reasoning concrete.

A fictional study of procurement tasks marks one submission as requiring clarification and another as excluded under its approved research rule. The analyst downloads the full table and assumes those labels have already removed both records.

The team chooses explicit membership acceptance and records the cutoff for pending clarification rather than treating every received submission as analysis-ready.

Take the questions into your next brief.

Use these prompts in your own notes. Nothing is submitted from this page.

1. States

   What evidence and owner define each review state?

2. Rule

   Which states belong in this particular analysis extract?

3. Fixture

   Which synthetic records must be included and excluded?

4. Amendment

   What change requires renewed review and receiving-data verification?

5. Freeze

   What cutoff and unresolved cases accompany the dataset?

Watch for the wrong conclusion.

• Assuming a rejected label removes a row automatically.
• Accepting equal totals without comparing membership.
• Letting an earlier approval silently cover later amendments.

Sources & scope

Workflows, worksheets and fictional examples are original editorial analysis. Sources support only specifically attributed principles. Platform guidance can change; legal requirements depend on jurisdiction and context. No measured results or named human reviewer are claimed.

• ODK Docs: Managing Submissions in Central →Primary submission-management documentation · Supports: Supports review-state filtering and Edited status after amendments in Central. The inclusion contract is original analysis.