---
title: "Judge the evidence in the result | Grafana Labs"
description: "Distinguish a completed run from a useful, supported summary."
---

> For a curated documentation index, see [llms.txt](/llms.txt). For the complete documentation index, see [llms-full.txt](/llms-full.txt).

# Completion isn’t proof of a useful answer

A completed checkout summary still needs evidence. The run conversation shows what Assistant did and the result it produced.

Use the intended summary to assess the output:

| Criterion    | Useful result                                  | Reason to revise                                |
|--------------|------------------------------------------------|-------------------------------------------------|
| Scope        | Named service, data sources, and time range    | Unrelated data or the wrong window              |
| Evidence     | Queries, values, or links support the findings | Claims without supporting data                  |
| Completeness | Missing data is identified                     | Missing data is treated as healthy              |
| Usefulness   | Summary answers the morning question           | Follow-up questions are needed to understand it |

## When the result falls short

An ambiguous prompt needs clearer context. Inaccessible data needs an access fix. Neither problem is solved by running the same task more frequently.