Resolve missing or insufficient evidence

A result is empty, limited, pending, or marked as not applicable. Before rerunning analysis, determine which state you are seeing and what additional evidence would help.

Your outcome: a specific recovery action or a documented limit on the conclusion you can make.

1. Read the actual state

Use the result's status, coverage, and evidence information. The terms below describe situations; exact labels depend on the screen.

Situation	Appropriate next step
Configured but not yet observed	Start the intended observation or wait for the scheduled execution, as appropriate.
Execution failed or collection stopped	Inspect run or measurement health and resolve the issue before judging performance.
Observation exists but relevant coverage is insufficient	Identify the missing system, questions, or period.
Metric is not applicable to the question type	Use the measurement suitable for that question; more identical executions will not fix applicability.
A valid measured result is zero	Inspect scope and evidence to understand the observed absence.
Only source content is inaccessible	Keep the response/citation observation, but limit claims about the source's full content.

2. Narrow the recovery to the missing information

Use the supported retry, investigation, or configuration path for the issue. For a limited question group, investigate that scope where available instead of automatically rerunning an entire report.

Before committing usage, state the decision another observation could change. Check the displayed usage and preserve the intended Context. If the problem is a source connection, fix the source; if the metric is inapplicable, select the right measurement.

Do not replace missing data from one AI system with all-system results. A narrower, clearly labeled conclusion is more useful than an apparently complete but mismatched result.

3. Decide whether to proceed or keep the conclusion open

You may have enough evidence for a narrow action even when a broad comparison remains unavailable. A verified incorrect product fact can support correcting that fact. It cannot establish a market-wide visibility deficit.

If recovery is blocked, retain the observation reference, status, intended scope, and attempted action for support. Set a review checkpoint only when an expected event could provide new information; avoid repeated executions with no decision purpose.

You are done when you can state what is measured, what remains missing, and the next action. If a valid result establishes that the brand is absent, continue to Find out why your brand is missing from AI answers.