How to record a website report disposition

A reviewed website report needs one clear decision. The investigation result stays immutable, and the disposition records what a human concluded after reading it. Without that decision, later metrics and handoffs become guesswork: nobody can tell which results were usable, which needed more work, and which were unsafe to follow.

This article explains how to choose a disposition, what belongs in the accompanying note, and which nearby actions stay separate from the decision itself.

Keep the result and the decision separate

A review-ready result already separates evidence, a conclusion, unknown facts, a recommended action, and a draft response. That package answers what the investigation found. It does not answer what you decided to do with the finding.

Record the disposition outside the result. Do not rewrite the conclusion to match your later judgment, and do not delete an unknown fact because the next step is already clear. The result remains the evidence package. The disposition remains the human decision about that package. The example report shows the result shape you review before you choose a disposition.

Choose one disposition from a fixed list

Use one decision for each result. A short fixed list keeps outcomes comparable over time:

• Accepted. The result was usable as written. You can act on the conclusion, unknown facts, and recommended action without rewriting the finding.
• Accepted with edits. The direction was right, but you corrected the draft response, narrowed the recommended action, or adjusted wording before you relied on it. The result itself stays unchanged.
• Needs more investigation. The current evidence boundary is not enough for a safe next step. Prefer this when a missing fact could change the action, not when you merely disagree with tone.
• Rejected, wrong conclusion. The evidence does not support the stated cause, or the conclusion is stronger than the citations allow.
• Rejected, unsafe. Following the recommended action, or sending the draft as written, would create risk. Use this when the problem is safety, not only accuracy.
• Reviewed without investigation. You reviewed a stored report that has no investigation result, and you recorded a human decision for that report.

Pick the closest honest label. Do not invent a custom status such as "maybe later" or "waiting on design". Ambiguous labels make weekly review numbers meaningless.

Write a note that explains the decision

An optional note travels with the disposition and can also be added later as an append-only entry. Keep the note short, concrete, and useful to the next reader.

A useful note answers three questions:

1. What did you check after reading the result?
2. Why does that check support this disposition?
3. What happens next?

Good examples:

• "Confirmed the cited path on the registered revision. Accepted the likely conclusion and will ship the focused fix."
• "Draft named the wrong visitor-facing cause. Accepted with edits after rewriting the reply."
• "Needs the missing account-state fact before any code change. Reinvestigation can wait until a better report arrives."

Avoid notes that only restate the label, invent a cause the evidence does not support, or paste private implementation detail into customer-facing language. The same text discipline that applies to a draft response applies to the note: no source paths in visitor language, and no claim stronger than the evidence.

Separate nearby actions from the disposition

Several report actions sit next to the decision and must not replace it.

• Edit and copy the draft response. The draft is owner-facing language for a reply you send yourself. Saving a draft revision does not set a disposition, and IssueProbe does not send that draft to the visitor.
• Request a reinvestigation. A new investigation consumes one monthly credit and uses the current registered source revision. Choose needs_more_investigation when the evidence boundary is the problem, then spend the credit only when the source or the next report actually changed. The guide to unknown conclusions covers that spending decision.
• Archive the report. Archiving removes the report from the active queue after an explicit confirmation. It is not a substitute for recording whether the result was accepted, rejected, or incomplete.
• Add a later note. After the decision, you can still append what you learned from a manual check. The decision stays one value; the note stream keeps the later rationale honest.

Review the evidence before you decide

A disposition is only as good as the review behind it. Before you choose a label, complete the same checks described in how to review an AI-generated bug investigation:

• Confirm the investigation used the registered revision for the report.
• Read the evidence citations before the conclusion summary.
• Match the conclusion label to the strength of the evidence.
• Keep unknown facts visible when they could change the next step.
• Treat the recommended action and the draft response as proposals, not instructions.

If you cannot complete those checks, prefer needs_more_investigation or a rejection with a clear note over an acceptance you cannot defend later.

Use dispositions to keep the process measurable

When every reviewed result receives one disposition, two process questions become answerable: how often a result needed no extra owner research, and how long reports take to reach a ready result that a human then decides. Those definitions are explained in how to measure a website report process.

Without dispositions, a dashboard can show activity and still hide stalled reviews. With them, accepted and accepted-with-edits results become the quality signal, while needs-more and rejection labels show where the process needs better evidence or a safer recommendation.

A practical closing rule

Record the disposition as soon as the review ends, even when the fix will ship later. The decision is about the result you read today. The release schedule is a separate plan. If IssueProbe is already collecting reports for the website, use the same review path in the dashboard or through the API so humans and authorized agents leave one comparable decision history.