Build a research task board that shows what writing still depends on

1. Name an inspectable output instead of a broad activity

Replace a card called “finish paper” with work that another person can check: agree the follow-up definition, run a named analysis on a specified dataset, reconcile a table with its output, or approve a particular manuscript version. Give each card one responsible owner and a completion condition. A due date states an intention; it is not evidence of completion.

Distinguish a contributor who performs the work from a reviewer who checks it and the person authorised to approve the next step. These may be different people. The board can record their responsibilities without granting access to source data or permission to submit the manuscript. Keep confidential files in the team’s approved storage and link to them through the appropriate access controls.

2. Read a six-card fictional snapshot

Imagine a methods paper at version 0.6 on 7 August 2026. The source dictionary does not establish whether its follow-up field refers to 30 or 90 days. The analyst has drafted code but has not run it. A separate paragraph describing the supplied study setting has been drafted and checked against the setting note; it makes no claim about follow-up or results.

The following six cards are the entire teaching board. “Complete” applies only to the checked setting paragraph. It does not approve the whole Methods section. The remaining cards show why a polished Discussion cannot honestly report an effect yet. Owner names below are role labels, not real appointments or evidence that anyone has accepted an assignment.

On small screens, scroll the table horizontally, or focus it and use the arrow keys.

3. Treat a blocked card as a decision to resolve

The follow-up question needs a source and a responsible decision, not an arbitrary value selected to move the board. Describe the ambiguity, identify who can clarify it and record the effect on planned analysis. The analyst can prepare checks or work on independent tasks while that question remains open. They cannot label an unexecuted script as a completed result.

When clarification arrives, attach its date and the version of the definition adopted. Review whether it changes the analysis plan before executing anything. In the fictional snapshot no clarification has arrived, so there is no agreed duration, result table, effect size or completed analysis to report. A calendar reminder can request a decision without manufacturing the answer.

4. Close a card at the level actually checked

For an analysis card, the completion record should identify the input version, code version, executed steps, output and checks relevant to that analysis. For a writing card, it should identify the exact passage, source and reviewer comments resolved. A file uploaded successfully can still be the wrong file. A reviewer opening it is not the same as approving its content.

Use separate states for drafted, awaiting review, changes requested and checked. Keep a short reason when a review fails so the author can correct the problem. Do not promote every card to a publication decision: final author approval, journal submission and editorial acceptance are later events with their own evidence. The fictional setting paragraph is the only closed card here.

5. Reopen downstream work when its evidence changes

Suppose a later authorised correction changes an input used in an already checked table. Record the correction and mark the affected table and interpretation for review. Preserve the earlier output as a superseded version with its date. Moving a card backwards is useful when it reflects changed evidence; hiding the correction to maintain a completion percentage is not.

Trace the dependency through the abstract, Results and Discussion rather than editing only the first number found. Assign each affected passage a reviewer and a current version. Do not send a revised file to a journal automatically because an internal card changed. The team must establish which release is authorised and whether the journal needs a correction or other communication.

6. Record contributions from work performed

NISO’s CRediT guidance supports discussing contributor roles in advance and confirming the roles assigned to contributors. It also distinguishes contribution descriptions from authorship criteria. A card assigned to “analysis” is a plan; the eventual contribution statement should describe what the person actually did and what the team has confirmed.

At each handover, review open decisions, completed artifacts and permissions with the next responsible person. A useful short meeting asks which dependency needs action and which evidence is ready for checking. It does not need a race to close the most cards. Judge the board by whether another researcher can understand and resume the work without guessing its scientific status.

Reference: NISO CRediT: implementation guidance

Questions for a practical final check

• Each card names an output and owner.
• Drafting, execution and approval are separate.
• Blocked work identifies the missing decision.
• Completion points to a checked version.
• Input changes reopen dependent claims.
• Contribution statements reflect performed work.