Turn missing evidence into an actionable research question list
An unresolved-question list is useful when each item states what is unknown, what evidence could answer it, and what to do next. “Understand the company better” is too broad to execute. “Which legal entity signed the facility agreement?” points to a document and a specific check.
The W3C provenance overview is background for connecting research objects and actions. SEC search is one possible source route. Neither prescribes the question-ticket design below; it is a proposed manual method for controlling unfinished research.
Classify the gap before assigning work
Separate an unknown fact, inaccessible source, identity ambiguity, and interpretation disagreement. An unknown fact needs evidence. An inaccessible document needs retrieval work. Identity ambiguity needs matching. An interpretation disagreement may remain even after everyone reads the same source. Mixing them creates tasks that appear complete without resolving the actual problem.
Write the question in a form that can receive a bounded answer. Name the entity, period, and relevant event where possible. Avoid asking two questions in one ticket. If “was approval obtained and did construction finish?” requires different sources, create two connected tickets.
Hypothetical question ticket
Assume fictional Silver Marsh announces a project subject to approval. All details are hypothetical.
| Field | Example |
|---|---|
| Question | Has the described approval been obtained? |
| Existing evidence | Issuer notice says approval is pending |
| Gap type | Fact not yet established |
| Target evidence | Approval record or explicit later issuer update |
| Next action | Search the relevant official publication route |
| Close condition | Source directly confirms status, or task closed as unresolved with coverage noted |
The ticket should not close merely because another news story repeats the plan. That story may be a lead, but it does not answer the approval question. A closed-unresolved outcome is different from a confirmed approval and must remain labeled accordingly.
Choose actions that can produce progress
An action should be small enough to perform: inspect one cited attachment, verify one issuer identifier, or search a defined date range. “Research more” does not specify a result. Add an owner if working with others, or a next review date for solo research, while preserving the evidence requirement as the real closure test.
Prioritize questions by their effect on the current research task. An identity uncertainty blocking every citation should precede a minor historical detail. This is workflow prioritization, not an investment ranking. Record why a question matters so a later reviewer can decide whether it remains relevant.
Verify closure explicitly
When closing a ticket, attach the source, locator, inspection date, and answer. If the source only narrows the uncertainty, update the question rather than forcing a binary result. For example, confirming which authority handles approval does not establish that approval was granted.
Keep the original question text and a short resolution history. This prevents a later reader from mistaking a revised question for the one that was answered. If the answer becomes stale after a new disclosure, reopen the ticket with the new reason rather than erasing the earlier resolution.
Limits and common mistakes
A question list does not guarantee that public information will resolve every issue. Missing schedules, future events, and ambiguous terms may remain open. Define a reasonable search boundary and record it honestly. The goal is visible uncertainty and directed work, not a falsely complete checklist.
Common mistakes include closing on a weak proxy, leaving no next action, mixing several entities in one question, and retaining obsolete questions forever. Another is treating an unanswered question as negative evidence. A knowledge gap should remain a knowledge gap until evidence supports a stronger claim.
A final record check
Review open tickets for answer substitution. Finding who is responsible for an approval is not the same as finding approval status; locating an agreement is not the same as checking its terms. If the latest action answered a nearby question, preserve that progress but leave the original question open. This prevents administrative completion from becoming a false factual resolution.
FAQ
Can I close a question without finding the answer?
Yes, as closed-unresolved with the search boundary and reason recorded. That is an administrative decision, not a factual answer.
How many open questions should a dossier contain?
There is no universal number. Keep the list tied to current tasks, split executable questions, and archive items whose purpose has ended with a visible reason.